Вы находитесь на странице: 1из 3121

PROGRAM • COGSCI 2015

37th Annual Meeting of the


Cognitive Science Society

Mind, Technology, and Society

Pasadena, California, 23-25 July 2015

Organized by

Rick Dale
Carolyn Jennings
Paul P. Maglio
Teenie Matlock
David C. Noelle
Anne Warlaumont
Jeff Yoshimi

Cognitive and Information Sciences


University of California, Merced

! 1
Table of Contents

Introduction and Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xli

Committees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xliii

Conference and Proceedings Information ........................................................ xlvi

Conference Awards .................................................................................... liii

Invited Presentations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . lvi

Workshops

Applying for National Science Foundation Funding in Cognitive Science: Cognition, Computation,
Development, Education, and Neuroscience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Anne Cleary, Hector Avila-Munoz, Evan Heit, Chris Hoadley, Laura Namy, Alumit Ishai, Betty

Tuller

Physical and Social Scene Understanding .............................................................. 3

Tao Gao, Yibiao Zhao, Lap-Fai Yu

Language & common sense: Integrating across psychology, linguistics, and computer science ........ 5

Joshua Hartshorne, Joshua Tenenbaum

Evidence Accumulation Modeling: Bayesian Estimation using Dierential Evolution ................ 7

Andrew Heathcote, Brandon Turner Turner, Scott Brown

Workshop on Optimizing Experimental Designs: Theory, Practice, and Applications ................ 9

Jay Myung, Mark Pitt, Maarten Speekenbrink

Tutorials

Quantifying the Dynamics of Interpersonal Interaction: A Primer on Cross-Recurrence


Quantication Analysis using R . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Moreno I. Coco, Rick Dale

Programming online experiments with jsPsych . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

Josh de Leeuw

Tutorial: Bayesian data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

John Kruschke

Full Day Tutorial on Quantum Models of Cognition and Decision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Jennifer Trueblood, James Yearsley, Zheng Wang, Jerome Busemeyer

Symposia

Connecting learning, memory, and representation in math education ............................... 19

Martha Alibali, Chuck Kalish, Timothy Rogers, Christine Massey, Phil Kellman, Vladimir

Sloutsky, James L. McClelland, Kevin Mickey

Causality and Agency Across Cultures and Languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Sieghard Beller, Andrea Bender, Jürgen Bohnemeyer, Annelie Rothe-Wulf, York Hagmayer,

Rita Astuti

i
Communicating Cognitive Science: Improving Awareness and Understanding Among People Who
are Not Ourselves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

Kevin Gluck, Wayne Gray, Marsha Lovett, Art Markman, Jim Spohrer

Generative and Discriminative Models in Cognitive Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Brad Love, Michael Ramscar, Tom Griths, Matt Jones

Analogical Processes in Language Learning .......................................................... 27

Bozena Pajak, Micah Goldwater, Dedre Gentner, Adele Goldberg, Ruxue Shao

The Relevance of Alternative Possibilities throughout Cognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Jonathan Phillips, Joshua Knobe, Andrew Shtulman, Charles Kalish, Anne Riggs,

Christopher Hitchcock

Papers

Eye-tracking situated language comprehension: Immediate actor gaze versus recent action events . . 31

Dato Abashidze, Pia Knoeferle, Maria Nella Carminati

Eect of heaviness on the cognitive evaluation process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

Keiga Abe

A Computational Approach to Modelling the Perception of Pitch and Tonality in Music . . . . . . . . . . . 42

Kat Agres, Carlos Cancino, Maarten Grachten, Stefan Lattner

Do potential past and future events activate the Lateral Mental Timeline? . . . . . . . . . . . . . . . . . . . . . . . . . . 48

Roberto Aguirre, Julio Santiago

A Rational Model for Individual Dierences in Preference Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

Sheeraz Ahmad, Angela Yu

Motion event expressions in language and gesture: Evidence from Persian . . . . . . . . . . . . . . . . . . . . . . . . . . 60

Niloofar Akhavan, Nazbanou Nozari, Tilbe Goksun

Daxing with a Dax: Evidence of Productive Lexical Structures in Children . . . . . . . . . . . . . . . . . . . . . . . . . . 66

Sara Al-Mughairy, Ruthe Foushee, David Barner, Mahesh Srinivasan

Visuo-Spatial Memory Processing and the Visual Impedance Eect ................................. 72

Rebecca Albrecht, Holger Schultheis, Wai-Tat Fu

Change your Mind: Investigating the Eects of Self-Explanation in the Resolution of


Misconceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Laura Allen, Danielle McNamara, Matthew McCrudden

Go shing! Responsibility judgments when cooperation breaks down . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

Kelsey Allen, Julian Jara-Ettinger, Tobias Gerstenberg, Max Kleiman-Weiner, Josh

Tenenbaum

Cognition in reach: continuous statistical inference in optimal motor planning . . . . . . . . . . . . . . . . . . . . . 90

Santiago Alonso-Diaz, Jessica F. Cantlon, Steven T. Piantadosi

Social Cues Aect Grasping Hysteresis in ASD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

Joseph Amaral, Heidi Kloos, Veronica Romero, Mike Richardson

How Grammatical Gender Aects Perspective Taking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

Elena Andonova, Zornitsa Savcheva, Gergana Todorova

ii
Teaching Children to Attribute Second-order False Beliefs: A Training Study with Feedback ...... 108

Burcu Arslan, Rineke Verbrugge, Niels Taatgen, Bart Hollebrandse

The London Underground Diagram as an example of cognitive niche construction . . . . . . . . . . . . . . . . . 114

Pedro Atã, João Queiroz

Adults Track Multiple Hypotheses Simultaneously during Word Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

Suzanne Aussems, Paul Vogt

Applying Pattern-based Classication to Sequences of Gestures .................................... 124

Suzanne Aussems, Mingyuan Chu, Sotaro Kita, Menno van Zaanen

Explaining Injustice in Speech: Individualistic vs. Structural Explanation ......................... 130

Saray Ayala, Nadya Vasilyeva

Does Training of Cognitive and Metacognitive Regulatory Processes Enhance Learning and
Deployment of Processes with Hypermedia? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

Roger Azevedo, Amy Johnson, Candice Burkett

Landmarks in motion: Unstable entities in route directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

Adriana Alexandra Baltaretu, Emiel Krahmer, Alfons Maes

A dynamic neural eld model of self-regulated eye movements during category learning . . . . . . . . . . . 148

Jordan Barnes, Mark Blair, Paul Tupper, R Calen Walshe

Moral Dynamics In Everyday Life: How Does Morality Evolve In Time? ......................... 154

Albert Barque-Duran, Emmanuel Pothos, James Yearsley, James Hampton

Improving Science Writing in Research Methods Classes Through Computerized Argument


Diagramming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

Brendan Barstow, Christian Schunn, Lisa Fazio, Mohammad Falakmasir, Kevin Ashley

Not by number alone: The eect of teachers' knowledge and its value in evaluating "sins of
omission" . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

Ilona Bass, Daniel Hawthorne, Noah Goodman, Hyowon Gweon

Humans predict liquid dynamics using probabilistic simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

Christopher Bates, Peter Battaglia, Ilker Yildirim, Josh Tenenbaum

The special status of color in pragmatic reasoning: evidence from a language game ............... 178

Peter Baumann

Folk Judgments of Normality: Part Statistical, Part Evaluative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

Adam Bear, Joshua Knobe

Modeling Lexical Acquisition Through Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

Nicole Beckage, Ariel Aguilar, Eliana Colunga

Predicting a Child's Trajectory of Lexical Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196

Nicole Beckage, Michael Mozer, Eliana Colunga

Crowdsourcing elicitation data for semantic typologies ............................................. 202

Barend Beekhuizen, Suzanne Stevenson

The Role of Executive Functions for Structure-Mapping in Mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

Kreshnik Begolli, Lindsey Richland, Susanne Jaeggi

iii
A model-based theory of omissive causation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

Paul Bello, Sangeet Khemlani

A model for full local image interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220

Guy Ben-Yosef, Liav Assif, Daniel Harari, Shimon Ullman

Extremely costly intensiers are stronger than quite costly ones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

Erin Bennett, Noah Goodman

The Power of the Representativeness Heuristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232

Sudeep Bhatia

Constraint-Based Parsing with Distributed Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238

Peter Blouw, Chris Eliasmith

Incorporating Background Knowledge into Text Classication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244

Reihane Boghrati, Justin Garten, Aleksandra Litvinova, Morteza Dehghani

Phonological Neighborhood Density Modulates Errors In Spoken Word Recognition . . . . . . . . . . . . . . . . 250

Mona Roxana Botezatu, Jon-Frederick Landrigan, Qi Chen, Daniel Mirman

Developmental Changes in the Relationship Between Grammar and the Lexicon . . . . . . . . . . . . . . . . . . . 256

Mika Braginsky, Daniel Yurovsky, Virginia Marchman, Michael Frank

Staying aoat on Neurath's boat  Heuristics for sequential causal learning . . . . . . . . . . . . . . . . . . . . . . . . 262

Neil Bramley, Peter Dayan, David Lagnado

The Eect of Probability Anchors in Moral Decision Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268

Chris Brand, Mike Oaksford

Hybrid-Logical Reasoning in the Smarties and Sally-Anne Tasks: What Goes Wrong When
Incorrect Responses are Given? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273

Torben Braüner

Incremental Object Perception in an Attention-Driven Cognitive Architecture . . . . . . . . . . . . . . . . . . . . . 279

Will Bridewell, Paul Bello

Minimal Requirements for Productive Compositional Signaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285

Thomas Brochhagen

Formalizing Risky Choice with a Logistic Model of Fuzzy Trace Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291

David Broniatowski, Valerie Reyna

Neural Correlates of Purchasing Behavior in the Prefrontal Cortex: An Optical Brain Imaging
Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297

Murat Perit Cakir, Tuna Cakar, Yener Girisken

Language evolution in the lab tends toward informative communication ........................... 303

Alexandra Carstensen, Jing Xu, Cameron Smith, Terry Regier

Eectiveness of Learner-Regulated Study Sequence: An in-vivo study in Introductory Psychology


course . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309

Paulo Carvalho, David Braithwaite, Josh de Leeuw, Benjamin Motz, Rob Goldstone

The perception of stroke-to-stroke turn boundaries in signed conversation ......................... 315

Marisa Casillas, Connie de Vos, Onno Crasborn, Stephen C. Levinson

iv
Eye Movement Pattern in Face Recognition is Associated with Cognitive Decline in the Elderly . . 321

Cynthia Y.h. Chan, Antoni B. Chan, Tatia M.c. Lee, Janet H. Hsiao

Eye to I: Males Recognize Own Eye Movements, Females Inhibit Recognition . . . . . . . . . . . . . . . . . . . . . 327

Sanjay Chandrasekharan, Geetanjali Date, Prajakt Pande, Jeenath Rahaman, Rakh

Shaikh, Anveshna Srivastava, Nisheeth Srivastava, Harshit Agrawal

Chunking in Working Memory and its Relationship to Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

Mustapha Chekaf, Nicolas Gauvrit, Alessandro Guida, Fabien Mathy

Learning and Generalizing Cross-Category Relations Using Hierarchical Distributed


Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339

Dawn Chen, Hongjing Lu, Keith Holyoak

Analyzing chunk pauses to measure mathematical competence: Copying equations using


`centre-click' interaction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345

Peter Cheng

Statements of equivalence can imply dierences: Asymmetries in directional comparisons . . . . . . . . . 351

Eleanor Chestnut, Carla Remulla, Ellen Markman

Probing the mental number line: A between-task analysis of spatial-numerical associations ....... 357

Chi-Ngai Cheung, Vladislav Ayzenberg, Rachel F. L. Diamond, Sami Yousif, Stella F.

Lourenco

Complex Mental Addition and Multiplication Rely More on Visuospatial than Verbal Processing . . 363

Tommy Kwun Leuk Cheung, Janet Hui Wen Hsiao

Algebraic reasoning in 3- to 5-year-olds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369

Pierina Cheung, Mathieu Le Corre

Promoting Comprehension of Health Information among Older Adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375

Jessie Chin, Jessica Johnson, Darcie Moeller, Elise Duwe, James Graumlich, Michael

Murray, Elizabeth Stine-Morrow, Daniel Morrow

Diagrams benet symbolic problem solving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381

Junyi Chu, Emily Fyfe, Bethany Rittle-Johnson

The Role of Certainty and Time Delay in Students' Cheating Decisions during Online Testing . . . 387

Chia-Yuan Chuang, Scotty D. Craig, John Femiani

Hidden Markov model analysis reveals better eye movement strategies in face recognition . . . . . . . . . 393

Tim Chuk, Antoni B. Chan, Janet Hsiao

Expertise modulates hemispheric asymmetry in holistic processing: Evidence from Chinese


character processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399

Harry K. S. Chung, Jacklyn C. Y. Leung, Janet H. Hsiao

Incidental Memory for Naturalistic Scenes: Exposure, Semantics, and Encoding . . . . . . . . . . . . . . . . . . 405

Moreno I. Coco, Nicholas Duran

Are Biases When Making Causal Interventions Related to Biases in Belief Updating? . . . . . . . . . . . . 411

Anna Coenen, Todd Gureckis

That went over my head: Constraints on the visual vocabulary of comics . . . . . . . . . . . . . . . . . . . . . . . . . . 417

Neil Cohn, Beena Murthy

v
The Bi-directional Relationship Between Source Characteristics and Message Content . . . . . . . . . . . . 423

Peter Collins, Ulrike Hahn, Yvlva von Gerber, Erik Olsson

Why Build a Virtual Brain? Large-scale Neural Simulations as Test-bed for Articial Computing
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429

Matteo Colombo

A Dissociation between Categorization and Similarity to Exemplars ............................... 435

Nolan Conaway, Kenneth Kurtz

More than Meets the Eye: Gesture Changes Thought, even without Visual Feedback .............. 441

Kensy Cooperrider, Elizabeth Wakeeld, Susan Goldin-Meadow

What is Lost in Translation from Visual Graphics to Text for Accessibility . . . . . . . . . . . . . . . . . . . . . . . . 447

Peter Coppin, Peter Coppin

Reading Words Hurts: The impact of pain sensitivity on people's ratings of pain-related words . . . . 453

Erica Cosentino, Markus Werning, Kevin Reuter

Gesture Production under Instructional Context: The Role of Mode of Instruction . . . . . . . . . . . . . . . . 459

Melda Coskun, Cengiz Acartürk

Time Course of Metaphor Comprehension in the Visual World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465

Seana Coulson, Tristan Davenport, Pia Knoeferle, Sarah Creel

A Dynamic Approach to Secondary Processes in Associative Recognition . . . . . . . . . . . . . . . . . . . . . . . . . . 471

Gregory Cox, Richard Shirin

Children's early perceptual and late-emerging social sensitivity to accented speech ................. 477

Sarah Creel, Emilie Seubert

Odor naming is dicult, even for wine and coee experts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483

Ilja Croijmans, Asifa Majid

The Eects of Racial Similarity and Dissimilarity on the Joint Simon Task . . . . . . . . . . . . . . . . . . . . . . . 489

Steve Croker, J. Scott Jordan, Daniel Schloesser, Vincent Cialdella

Embodied cognition and passive processing: What hand-tracking tells us about syntactic processing
in L1 and L2 speakers of English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495

Scott Crossley, Youjin Kim, Tiany Lester, Samuel Clark

A Bayesian Latent Mixture Approach to Modeling Individual Dierences in Categorization Using


General Recognition Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501

Irina Danileiko, Michael Lee, Michael Kalish

Neural Eects of Childhood Language Deprivation on Picture Processing: Insights from Adolescent
First-Language Learners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507

Tristan Davenport, Naja Ferjan Ramirez, Matthew Leonard, Rachel Mayberry, Eric Halgren

The suggestible nature of apparent motion perception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513

Nicolas Davidenko, Yeram Cheong, Jacob Smith

Evidence for widespread thematic structure in the mental lexicon. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518

Simon De Deyne, Steven Verheyen, Amy Perfors, Daniel Navarro

vi
Behaviorist Thinking in Judgments of Wrongness, Punishment, and Blame . . . . . . . . . . . . . . . . . . . . . . . 524

Julian de Freitas, Samuel Johnson

Memory constraints aect statistical learning; statistical learning aects memory constraints . . . . . 530

Josh de Leeuw, Rob Goldstone

Using a Task-Filled Delay During Discrimination Trials to Examine Dierent Components of


Learned Visual Categorical Perception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536

Joshua de Leeuw, Jan Andrews

Savvy software agents can encourage the use of second-order theory of mind by negotiators . . . . . . . 542

Harmen de Weerd, Eveline Broers, Rineke Verbrugge

Wonky worlds: Listeners revise world knowledge when utterances are odd ......................... 548

Judith Degen, Michael Henry Tessler, Noah D. Goodman

Jack is a True Scientist: On the Content of Dual Character Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . 554

Guillermo Del Pinal, Kevin Reuter

Modeling Relational Priming and Multiplicative Reasoning with Rational Numbers . . . . . . . . . . . . . . . . 560

Melissa Dewolf, Miriam Bassok, Keith Holyoak

Implicit Understanding of Arithmetic with Rational Numbers: The Impact of Expertise . . . . . . . . . . . 566

Melissa Dewolf, Ji Son, Miriam Bassok, Keith Holyoak

The Sound of Valence: Phonological Features Predict Word Meaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 572

Karlijn Dinnissen, Max M. Louwerse

Learning to reason about desires: An infant training study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 578

Tiany Doan, Stephanie Denison, Christopher Lucas, Alison Gopnik

When high pitches sound low: Children's acquisition of space-pitch metaphors . . . . . . . . . . . . . . . . . . . . . 584

Sarah Dolscheid, Sabine Hunnius, Asifa Majid

Connecting rule-abstraction and model-based choice across disparate learning tasks ............... 590

Hilary Don, Micah Goldwater, A. Ross Otto, Evan Livesey

Examining the Bilingual Advantage on Conict Resolution Tasks: A Meta-Analysis .............. 596

Seamus Donnelly, Patricia Brooks, Bruce Homer

Sound-Symbolism is Disrupted in Dyslexia: Implications for the Role of Cross-Modal Abstraction


Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 602

Linda Drijvers, Lorijn Zaadnoordijk, Mark Dingemanse

Tracking the Response Dynamics of Implicit Partisan Biases ...................................... 608

Nicholas Duran, Stephen Nicholson, Rick Dale

Explaining Choice Behavior: The Intentional Selection Assumption ............................... 614

Kelley Durkin, Leyla R. Caglar, Elizabeth Bonawitz, Patrick Shafto

Pathways of Conceptual Change: Investigating the Inuence of Experimentation Skills on


Conceptual Knowledge Development in Early Science Education . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620

Peter Edelsbrunner, Lennart Schalk, Ralph Schumacher, Elsbeth Stern

When Less Can Be More: Dual Task Eects on Speech Fluency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626

Naomi Eichorn, Klara Marton

vii
Behavioral Dynamics of a Collision Avoidance Task: How Asymmetry Stabilizes Performance ... 632

Brian Eiler, Rachel Kallen, Steven Harrison, Elliot Saltzman, Richard Schmidt, Mike

Richardson

Making moves: How sex and race are detected from biological motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638

Brian Eiler, Rachel Kallen, Mike Richardson

Navigation with Learned Spatial Aordances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644

Susan L. Epstein, Anoop Aroor, Matthew Evanusa, Elizabeth Sklar, Simon Parsons

Common object representations for visual recognition and production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 650

Judith E. Fan, Daniel L. K. Yamins, Nicholas B. Turk-Browne

A computational model of bilingual semantic convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656

Shin-Yi Fang, Benjamin Zinszer, Barbara Malt, Ping Li

Matching arti[U+FB01]cial agents' and users' personalities: designing agents with regulatory-focus
and testing the regulatory [U+FB01]t eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 662

Caroline Faur, Jean-Claude Martin, Céline Clavel

Conict Sensitivity and the Conjunction Fallacy: Eye-tracking Evidence for Logical Intuitions in
Conjunction Probability Judgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 668

Jenny Faure-Bloom, Gaëlle Vallée-Tourangeau, Sabira Mannan

Production is biased to provide informative cues early: Evidence from miniature articial
languages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674

Maryia Fedzechkina, T. Florian Jaeger, John Trueswell

Music familiarity modulates mind wandering during lexical processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 680

Shi Feng, Gavin Bidelman

Temporal Binding and Internal Clocks: Is Clock Slowing General or Specic? . . . . . . . . . . . . . . . . . . . . . 686

Richard Fereday, Marc Buehner

Visual abstract rule learning by 3- and 4-month-old infants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 692

Brock Ferguson, Sandra Waxman

Learning Exceptions in Phonological Alternations .................................................. 698

Sara Finley

Frequency Eects in Morpheme Segmentation ...................................................... 704

Sara Finley

ACT-R and LBA Model Mimicry Reveals Similarity Across Modeling Formalisms . . . . . . . . . . . . . . . . 710

Christopher Fisher, Matthew Walsh, Leslie Blaha, Glenn Gunzelmann

How Physical Interaction Helps Performance in a Scrabble-like Task .............................. 716

Morgan Fleming, Paul Maglio

Framing eects and the folk psychiatry of addiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 722

Stephen Flusberg, Michael Dellavalle, Paul Thibodeau

Visual-motor coordination in natural reaching of young children and adults . . . . . . . . . . . . . . . . . . . . . . . 728

John Franchak, Chen Yu

viii
Cumulative Contextual Facilitation in Word Activation and Processing: Evidence from
Distributional Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734

Diego Frassinelli, Frank Keller

Turn, Turn, Turn: Perceiving Global and Local, Clockwise and Counterclockwise Rotations ...... 740

Bob French, Helle Lukowski-Duplessy, Cory Rieth, Gary Cottrell

Defaulting eects contribute to the simulation of cross-linguistic dierences in Optional Innitive


errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746

Daniel Freudenthal, Julian Pine, Gary Jones, Fernand Gobet

Physiological entrainment and behavioral coordination in a collective, creative construction task . . 752

Riccardo Fusaroli, Johanne Bjørndahl, Andreas Roepstor, Kristian Tylén

A Study and Preliminary Model of Cross-Domain Inuences on Creativity . . . . . . . . . . . . . . . . . . . . . . . . 758

Liane Gabora, Nicole Carbert

Listen, Look, Go! The Role of Prosody and Gaze in Turn-End Anticipation . . . . . . . . . . . . . . . . . . . . . . . 764

Chiara Gambi, Torsten Kai Jachmann, Maria Staudte

Defeasible Reasoning with Quantiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 770

Lupita Estefania Gazzo Castaneda, Markus Knau

Phrase similarity in humans and machines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 776

Samuel Gershman, Josh Tenenbaum

How, whether, why: Causal judgments as counterfactual contrasts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 782

Tobias Gerstenberg, Noah Goodman, David Lagnado, Josh Tenenbaum

Responsibility judgments in voting scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 788

Tobias Gerstenberg, Joseph Halpern, Josh Tenenbaum

What causes category-shifting in human semi-supervised learning? ................................ 794

Bryan Gibson, Timothy Rogers, Chuck Kalish, Xiao jin Zhu

Mental states are more important in evaluating moral than conventional violations ............... 800

Carly Gin, Tania Lombrozo

The Eect of Disrupted Attention on Encoding in Young Children . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 806

Karrie Godwin, Anna Fisher

A Spiking Neural Model of the n-Back Task ........................................................ 812

Jan Gosmann, Chris Eliasmith

Robustness of semantic encoding eects in a transfer task for multiple-strategy arithmetic


problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 818

Hippolyte Gros, Jean-Pierre Thibaut, Emmanuel Sander

A Hierarchical Cognitive Threshold Model of Human Decision Making on Dierent Length


Optimal Stopping Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 824

Maime Guan, Michael Lee, Joachim Vandekerckhove

The Inuence of Language on Memory for Object Location ........................................ 830

Harmen Gudde, Kenny Coventry, Paul Engelhardt

ix
A Dual-process Model of Framing Eects in Risky Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 836

Lisa Guo, Jennifer S. Trueblood, Adele Diederich

Interactivity, Expertise and Individual Dierences in Mental Arithmetic . . . . . . . . . . . . . . . . . . . . . . . . . . . 842

Lisa Guthrie, Charlotte Harris, Frederic Vallee-Tourangeau

Knowing what he could have shown: The role of alternatives in children's evaluation of
under-informative teachers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 848

Hyowon Gweon, Mika Asaba

How do adults reason about their opponent? Typologies of players in a turn-taking game . . . . . . . . . 854

Tamoghna Halder, Khyati Sharma, Sujata Ghosh, Rineke Verbrugge

Finding the return path: allo- versus egocentric perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 860

Kai Hamburger, Florian Röser

Think again? The amount of mental simulation tracks uncertainty in the outcome . . . . . . . . . . . . . . . . 866

Jessica Hamrick, Kevin Smith, Tom Griths, Ed Vul

Making Sense of Time-Series Data: How Language Can Help Identify Long-Term Trends . . . . . . . . 872

Jordan Harold, Kenny R. Coventry, Irene Lorenzoni, Thomas F. Shipley

Why do you ask? Good questions provoke informative answers. ................................... 878

Robert X.d. Hawkins, Andreas Stuhlmüller, Judith Degen, Noah D. Goodman

So good it has to be true: Wishful thinking in theory of mind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 884

Daniel Hawthorne-Madell, Noah Goodman

Why do people fail to consider alternative hypotheses in judgments under uncertainty? . . . . . . . . . . . 890

Brett Hayes, Guy Hawkins, Ben Newell

Inferring the Tsimane's use of color categories from recognition memory . . . . . . . . . . . . . . . . . . . . . . . . . . 896

Pernille Hemmer, Kimele Persaud, Celeste Kidd, Steven Piantadosi

New space-time metaphors foster new mental representations of time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 902

Rose Hendricks, Lera Boroditsky

Quantifying the time course of similarity ........................................................... 908

Andrew Hendrickson, Daniel Navarro, Chris Donkin

A Computational Model of Mind Wandering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 914

Laura Hiatt, Greg Trafton

Teaching with Rewards and Punishments: Reinforcement or Communication? . . . . . . . . . . . . . . . . . . . . 920

Mark Ho, Michael Littman, Fiery Cushman, Joseph Austerweil

Sources of developmental change in pragmatic inferences about scalar terms . . . . . . . . . . . . . . . . . . . . . . . 926

Alexandra Horowitz, Michael Frank

Exploring Individual Dierences via Clustering on Capacity Coecients .......................... 932

Joseph Houpt, Leslie Blaha

Ideas in Dialogue: The Eects of Interaction on Creative Problem Solving ........................ 938

Christine Howes, Patrick Healey, Pietro Panzarasa, Thomas Hills

Analogical comparison aids false belief understanding in preschoolers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 944

Christian Hoyos, William Horton, Dedre Gentner

x
Preschoolers' and Chimpanzees' Use of Source Reliability on Action-Based Tasks . . . . . . . . . . . . . . . . . 950

Melissa Hrabic, Bethany MacDonald, Michael Beran, Rebecca Williamson

Can children balance the size of a majority with the quality of their information? . . . . . . . . . . . . . . . . . 956

Jane Hu, Andrew Whalen, Daphna Buchsbaum, Tom Griths, Fei Xu

A Resource-Rational Approach to the Causal Frame Problem ...................................... 962

Thomas Icard, Noah Goodman

Beliefs about desires: Children's understanding of how knowledge and preference inuence choice. 968

Julian Jara-Ettinger, Emily Lydic, Josh Tenenbaum, Laura Schulz

The naïve utility calculus: Joint inferences about the costs and rewards of actions ................ 974

Julian Jara-Ettinger, Laura Schulz, Josh Tenenbaum

Task-General Object Similarity Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 980

Gavin Jenkins, Larissa Samuelson, John Spencer

The Standard Theory of Conscious Perception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 986

Carolyn Jennings

Mere Newness Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 992

Yun Jie, Ye Li

Similarity and Variation in the Distribution of Spatial Expressions Across Three Languages . . . . . . 997

Kristen Johannes, Jenny Wang, Anna Papafragou, Barbara Landau

Predictions from Uncertain Beliefs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1003

Samuel Johnson, Thomas Merchant, Frank Keil

Belief Utility as an Explanatory Virtue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1009

Samuel Johnson, Greeshma Rajeev-Kumar, Frank Keil

Argument Scope in Inductive Reasoning: Evidence for an Abductive Account of Induction . . . . . . . 1015

Samuel Johnson, Thomas Merchant, Frank Keil

Probabilistic Versus Heuristic Accounts of Explanation in Children: Evidence from a Latent


Scope Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1021

Angie Johnston, Samuel Johnson, Marissa Koven, Frank Keil

Interruption-recovery training transfers to novel tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1027

Winston Jones, Jarrod Moss

Organizing Metacognitive Tutoring Around Functional Roles of Teachers . . . . . . . . . . . . . . . . . . . . . . . . . 1033

David Joyner, Ashok Goel

Analyzing the Predictability of Lexeme-specic Prosodic Features as a Cue to Sentence


Prominence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1039

Sofoklis Kakouros, Okko Johannes Räsänen

Young Children's Understanding of the Successor Function ....................................... 1045

Jennifer Kaminski

Let's talk (ironically) about the weather: Modeling verbal irony . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1051

Justine Kao, Noah Goodman

xi
What is the Role of Conceptual Analysis in Cognitive Science? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1057

Liam Kavanagh, Christopher Suhler

Pronominal Reference and Pragmatic Enrichment: A Bayesian Account . . . . . . . . . . . . . . . . . . . . . . . . . 1063

Andrew Kehler, Hannah Rohde

Resolving Rogers' Paradox with Specialized Hybrid Learners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1069

Milad Kharratzadeh, Marcel Montrey, Alex Metz, Thomas Shultz

Causal relations from kinematic simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1075

Sangeet Khemlani, Geo Goodwin, Phil Johnson-Laird

Domino eects in causal contradictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1081

Sangeet Khemlani, Phil Johnson-Laird

Development of Numerosity Estimation: A Linear to Logarithmic Shift? . . . . . . . . . . . . . . . . . . . . . . . . . 1087

Dan Kim, John Opfer

The number of times a motion repeats inuences sentence processing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1093

Lucy Kyoungsook Kim, Elsi Kaiser

Eects of Emotional Prosody and Attention on Semantic Priming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1099

Seung Kyung Kim, Meghan Sumner

Comparison and Function in Children's Object Categorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1105

Katherine Kimura, Samuel Hunley, Laura Namy

Voice-specic eects in semantic association ...................................................... 1111

Ed King, Meghan Sumner

When Do Nonspecic Goals Help Learning? An Issue of Model Quality . . . . . . . . . . . . . . . . . . . . . . . . . . 1117

Saskia Kistner, Bruce Burns, Regina Vollmeyer, Kortenkamp Ulrich

Inference of Intention and Permissibility in Moral Decision Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1123

Max Kleiman-Weiner, Tobias Gerstenberg, Sydney Levine, Josh Tenenbaum

Supervised and unsupervised learning in phonetic adaptation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1129

Dave F. Kleinschmidt, Rajeev Raizada, T. Florian Jaeger

2-year-olds use syntax to infer actor intentions in a rational-action paradigm . . . . . . . . . . . . . . . . . . . . 1135

Melissa Kline, Jesse Snedeker

Scene Inversion Slows the Rejection of False Positives through Saccade Exploration During
Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1141

Kathryn Koehler, Miguel P. Eckstein

Expertise in Cognitive Task Analysis Interviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1147

Danny Koh, Kenneth Koedinger, Carolyn Rose, David Feldon

On the interplay between spontaneous spoken instructions and human visual behaviour in an
indoor guidance task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1153

Nikolina Koleva, Sabrina Hoppe, Mohammad Mehdi Moniri, Maria Staudte, Andreas

Bulling

xii
Investigating Ways of Interpretations of Articial Subtle Expressions Among Dierent
Languages: A Case of Comparison Among Japanese, German, Portuguese and Mandarin
Chinese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1159

Takanori Komatsu, Rui Prada, Kazuki Kobayashi, Seiji Yamada, Kotaro Funakoshi, Mikio

Nakano

The better part of not knowing: Virtuous ignorance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1165

Jonathan F. Kominsky, Philip Langthorne, Frank C. Keil

Hierarchical Reasoning with Distributed Vector Representations .................................. 1171

Cody Kommers, Volkan Ustun, Abram Demski, Paul Rosenbloom

Exploring Complexity in Decisions from Experience: Same Minds, Same Strategy . . . . . . . . . . . . . . . . 1177

Emmanouil Konstantinidis, Nathaniel J. S. Ashby, Cleotilde Gonzalez

Language and Gesture Descriptions Aect Memory: A Nonverbal Overshadowing Eect . . . . . . . . . 1183

Mark Koranda, Maryellen MacDonald

Can You Repeat That? The Eect of Item Repetition on Interleaved and Blocked Study . . . . . . . . . 1189

Abigail Kost, Paulo Carvalho, Rob Goldstone

Can Modern Neuroscience Change Our Idea of the Human? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1195

Boris Kotchoubey

Emergent Collective Sensing in Human Groups ................................................... 1201

Peter Krat, Robert X.d. Hawkins, Alex Pentland, Noah Goodman, Josh Tenenbaum

Semantically underinformative utterances trigger pragmatic inferences . . . . . . . . . . . . . . . . . . . . . . . . . . . 1207

Ekaterina Kravtchenko, Vera Demberg

Animation Facilitates Source Understanding and Spontaneous Analogical Transfer . . . . . . . . . . . . . . . 1213

James Kubricht, Hongjing Lu, Keith Holyoak

A diusion model account of the transfer-of-training eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1219

Colin Kupitz, Martin Buschkuehl, Susanne Jaeggi, John Jonides, Priti Shah, Joachim

Vandekerckhove

Computational evolution of decision-making strategies ............................................ 1225

Peter Kvam, Joseph Cesario, Jory Schossau, Heather Eisthen, Arend Hintze

The fan eect in overlapping data sets and logical inference ...................................... 1231

Kam Kwok, Robert West, Matthew Kelly

The learnability of Auditory Center-embedded Recursion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1237

Jun Lai, Emiel Krahmer, Jan Sprenger

Deep Neural Networks Predict Category Typicality Ratings for Images ........................... 1243

Brenden Lake, Wojciech Zaremba, Rob Fergus, Todd Gureckis

More than true: Developmental changes in use of the inductive strength for selective trust ...... 1249

Asheley Landrum, Joshua Cloudy, Patrick Shafto

Assessing Claims of Metaphorical Salience Through Corpus Data ................................ 1255

Jenny Lederer

xiii
Semantic Alignment of Fractions and Decimals with Discrete Versus Continuous Entities: A
Cross-national Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1261

Hee Seung Lee, Melissa Dewolf, Miriam Bassok, Keith Holyoak

The Roles of Knowledge and Memory in Generating Top-10 Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1267

Michael Lee, Emily Liu, Mark Steyvers

Convincing people of the Monty Hall Dilemma answer: The impact of solution type and
individual dierences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1273

Joanne Lee, Bruce Burns

Independent Recognition of Numerosity Requires Attention ....................................... 1279

Saebyul Lee, Vladimir Sloutsky

Measuring Time Gestures with the Microsoft Kinect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1285

Daniel Lenzen

If at First You Don't Succeed: The Role of Evidence in Preschoolers' and Infants' Persistence. . 1290

Julia Leonard, Laura Schulz

Visual Working Memory as Decision Making: Compensation for Memory Uncertainty in Reach
Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1296

Rachel Lerch, Chris Sims

The reliability of testimony and perception: connecting epistemology and linguistic evidentiality . 1302

Claire Lesage, Nalini Ramlakhan, Ida Toivonen, Chris Wildman

Linguistic input overrides conceptual biases: When goals don't matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1308

Nicholas Lester

Word order in a grammarless language: A `small-data' information-theoretic approach . . . . . . . . . . 1314

Nicholas Lester, Fermin Moscoso del Prado Martin

Constructional paradigms aect visual lexical decision latencies in English . . . . . . . . . . . . . . . . . . . . . . . 1320

Nicholas Lester, Fermin Moscoso del Prado Martin

The Smell of Jazz: Crossmodal Correspondences Between Music, Odor, and Emotion . . . . . . . . . . . 1326

Carmel Levitan, Sara Charney, Karen Schloss, Stephen Palmer

Structured priors in visual working memory revealed through iterated learning ................... 1332

Timothy Lew, Ed Vul

Conceptual complexity and the evolution of the lexicon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1338

Molly Lewis, Mike Frank

Music Reading Expertise Modulates Hemispheric Lateralization in English Word processing but
not in Chinese Character Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1344

Tze Kwan Li, Janet H. Hsiao

Constraints on Learning Non-Adjacent Dependencies (NADs) of Visual Stimuli . . . . . . . . . . . . . . . . . 1350

Jia Li, Toben Mintz

Evaluating Human Cognition of Containing Relations with Physical Simulation . . . . . . . . . . . . . . . . . . 1356

Wei Liang, Yibiao Zhao, Yixin Zhu, Song-Chun Zhu

xiv
When to use which heuristic: A rational solution to the strategy selection problem . . . . . . . . . . . . . . . 1362

Falk Lieder, Tom Griths

Children and adults dier in their strategies for social learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1368

Falk Lieder, Zi Lin Sim, Jane C. Hu, Tom Griths, Fei Xu

The Exemplar Confusion Model: An Account of Biased Probability Estimates in Decisions from
Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1374

Deborah Lin, Christopher Donkin, Ben Newell

Stepping Up to the Blackboard: Distributed Cognition in Doctor-Patient Interactions . . . . . . . . . . . . 1380

Katherine Lippa, Valerie Shalin

Creating New Sign Systems from Scratch: Gesture has the Upper Hand . . . . . . . . . . . . . . . . . . . . . . . . . . 1386

Casey J. Lister, Nicolas Fay, T. Mark Ellison, Jeneva Ohan

Linguistic Modality Aects the Creation of Structure and Iconicity in Signals . . . . . . . . . . . . . . . . . . . . 1392

Hannah Little, Kerem Erylmaz, Bart de Boer

Can experience with dierent types of writing system modulate holistic processing in speech
perception? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1398

Tianyin Liu, Janet Hsiao

Symbolic Integration, Not Symbolic Estrangement, For Double-Digit Numbers . . . . . . . . . . . . . . . . . . . . 1404

Allison Liu, Christian Schunn, Julie Fiez, Melissa Libertus

Action-Oriented Representations in the Motor Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1410

Daniel Hsi-wen Liu

Tracking Relations: The Eects of Visual Attention on Relational Recognition . . . . . . . . . . . . . . . . . . . 1416

Katherine Livins, Leonidas Doumas, Michael Spivey

Piece of Mind: Long-Term Memory Structure in ACT-R and CHREST . . . . . . . . . . . . . . . . . . . . . . . . . . 1422

Martyn Lloyd-Kelly, Fernand Gobet, Peter Lane

Time after Time in Words: Chronology through Language Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1428

Max Louwerse, Susanne Raisig, Richard Tillman, Sterling Hutchinson

Partitioning the Firing Patterns of Spike Trains by Community Modularity . . . . . . . . . . . . . . . . . . . . . . 1434

Hu Lu, Xing Hao Huang, Yu Qing Song, Hui Wei

Exploring the Concept of Utility: Are Separate Value Functions required for Risky and
Inter-temporal Choice? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1440

Ash Luckman, Chris Donkin, Ben R. Newell

Variability in Human Response Time Reects Statistical Learning and Adaptive Decision-Making 1446

Ning Ma, Angela Yu

Referential cues modulate attention and memory during cross-situational word learning . . . . . . . . . 1452

Kyle MacDonald, Daniel Yurovsky, Michael Frank

Memory Strategically Encodes Externally Unavailable Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1458

Carla Macias, Amanda Yung, Pernille Hemmer, Celeste Kidd

Evaluating contingencies by a dual system of learning the structure and the parameters of the
environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1464

Tamas Madarasz, Joseph Ledoux, Joshua Johansen

xv
Modelling Political Source Credibility of Election Candidates in the USA . . . . . . . . . . . . . . . . . . . . . . . . . 1470

Jens Koed Madsen

Investigation on Using 3D Printed Liver during Surgery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1476

Akihiro Maehigashi, Kazuhisa Miwa, Hitoshi Terai, Tsuyoshi Igami, Yoshihiko Nakamura,

Kensaku Mori

Quit while you're ahead: Preschoolers' persistence and willingness to accept challenges are
aected by social comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1482

Rachel Magid, Laura Schulz

Ignorance-Based Chance Discovery. Beyond Dark Events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1488

Lorenzo Magnani, Selene Arni, Tommaso Bertolotti

Universals on natural language determiners from a PAC-learnability perspective . . . . . . . . . . . . . . . . . 1494

Giorgio Magri, Giorgio Magri

Explaining the Number Hierarchy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1500

Robert Malouf, Farrell Ackerman, Scott Seyfarth

The mental number-line spreads by gestural contagion ............................................ 1506

Tyler Marghetis, Luke Eberle, Benjamin Bergen

Modeling choice and search in decisions from experience: A sequential sampling approach . . . . . . . 1512

Douglas Markant, Timothy Pleskac, Adele Diederich, Thorsten Pachur, Ralph Hertwig

Both symbolic and embodied representations contribute to spatial language processing; Evidence
from younger and older adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1518

Ioanna Markostamou, Kenny Coventry, Chris Fox, Lynn McInnes

Intellectualism and Psychology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1524

Jack Marley-Payne

Naïve Beliefs About Intervening on Causes and Symptoms in the Health Domain . . . . . . . . . . . . . . . . 1529

Jessecae Marsh, Andrew Zeveney

Priming bicultural bilingual Latino-Americans as Latino or American modulates access to the


Spanish and English meaning of interlingual homographs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1535

Benjamin Marsh, Jean-Paul Snijder, Marina Pulver, Veronica Johnson, Janna Schirmer,

Hyun Seo Lee, Ashley Horiuchi, Natalie Koskela, Brandon Reynoso, Raul Fajardo

Illusory inferences: disjunctions, indenites, and the erotetic theory of reasoning . . . . . . . . . . . . . . . . 1541

Salvador Mascarenhas, Philipp Koralus

Distributional determinants of learning argument structure constructions in rst and second


language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1547

Yevgen Matusevych, Afra Alishahi, Ad Backus

Individual Dierences in Chunking Ability Predict On-line Sentence Processing . . . . . . . . . . . . . . . . . . 1553

Stewart M. McCauley, Morten H. Christiansen

During category learning, top-down and bottom up processes battle for control of the eyes ....... 1559

Caitlyn McColeman, Mark Blair

Sound to Meaning Mappings in the Bouba-Kiki Eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1565

Kelly McCormick, Jee Young Kim, Sara List, Lynne C. Nygaard

xvi
An Integrated Account of Explanation and Question Answering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1571

Ben Meadows, Richard Heald, Pat Langley

Reasoning About Diverse Evidence in Preference Predictions ..................................... 1577

Rachel Meng, Stephanie Y. Chen, Daniel M. Bartels

A Bayesian Framework for Learning Words From Multiword Utterances ......................... 1583

Stephan Meylan, Thomas Griths

A latent-mixture quantum probability model of causal reasoning within a Bayesian inference


framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1589

Percy Mistry, Jennifer Trueblood, Joachim Vandekerckhove, Emmanuel Pothos

Reconstructing the Bayesian Adaptive Toolbox: Challenges of a dynamic environment and partial
information acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1595

Percy Mistry, Jennifer Trueblood

The Role of Outcome Divergence in Goal-Directed Choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1601

Prachi Mistry, Mimi Liljeholm

Towards semantically rich and recursive word learning models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1607

Francis Mollica, Steven Piantadosi

The perceptual foundation of linguistic context .................................................... 1613

Francis Mollica, Steven Piantadosi, Michael Tanenhaus

Personal Change and the Continuity of Identity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1619

Sarah Molouki, Daniel Bartels

An ACT-R Model of the Choose-Short Eect in Time and Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1625

Jung Aa Moon, John Anderson

The Antecedents of Moments of Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1631

Gregory Moore, Ryan Baker, Sujith Gowda

Cognitive Factors and Representation Strategies In Sketching Math Diagrams . . . . . . . . . . . . . . . . . . . . 1637

Damian Morden-Snipper, Ting Dai, Julie Booth, Briana Chang, Jennifer Cromley, Nora

Newcombe

What the Baldwin Eect aects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1643

Thomas Morgan, Thomas Griths

Modeling idiosyncratic preferences: How generative knowledge and expression frequency jointly
determine language structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1649

Emily Morgan, Roger Levy

Syntactic Alignment is an Index of Aective Alignment: An Information-Theoretical Study of


Natural Dialogue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1655

Fermín Moscoso del Prado Martín, John W. Du Bois

Does the Frequency of Pedagogical Agent Intervention Relate to Learners' Self-Reported Boredom
while using Multiagent Intelligent Tutoring Systems? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1661

Nicholas Mudrick, Roger Azevedo, Michelle Taub, François Bouchet

A non-monotonic extension of universal moral grammar theory .................................. 1667

Gert-Jan Munneke, Jakub Szymanik

xvii
The Eect of Facial Emotion and Action Depiction on Situated Language Processing . . . . . . . . . . . . 1673

Katja Münster, Maria Nella Carminati, Pia Knoeferle

Beyond Magnitude: How Math Expertise Guides Number Representation . . . . . . . . . . . . . . . . . . . . . . . . . 1679

April Murphy, Timothy Rogers, Edward Hubbard, Autumn Brower

Inuence of Excitation/Inhibition Imbalance on Local Processing Bias in Autism Spectrum


Disorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1685

Yukie Nagai, Takakazu Moriwaki, Minoru Asada

Mediators vs. Confounds: Exploring Dierent Intuitions about Causal Mechanisms . . . . . . . . . . . . . . 1691

Jonas Nagel, Simon Stephan

Eect is sure, but explanation is unsure:Closer investigation of the foreign language eect with
Japanese participants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1697

Kuninori Nakamura

Investigating Strategy Discovery and Coordination in a Novel Virtual Sheep Herding Game
among Dyads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1703

Patrick Nalepka, Cristopher Riehm, Carl Bou Mansour, Anthony Chemero, Michael J.

Richardson

Count on Diversity: The Cognitive and Mathematical Proles of Children in Early Elementary
School . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1709

Adam Newton, Marcie Penner-Wilger

Gestures Prime Temporal Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1715

Melvin Ng, Winston Goh, Melvin Yap, Chi-Shing Tse, Wing Chee So

Children's Trust in Technological and Human Informants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1721

Nicholaus Noles, Judith Danovitch, Patrick Shafto

The Symbolic Working Memory: memory accommodations for schematic processing of symbolic
information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1727

Nader Noori

Schematic Processing in Working Memory Tasks Relies on Learning and Long-Term Memory
Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1733

Nader Noori

The pragmatics of negation across contexts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1739

Ann Nordmeyer, Michael Frank

Implementation of selective attention in sequential word production .............................. 1745

Nazbanou Nozari, Gary Dell, Kyle Schneck, Barry Gordon

Response Dominance Predicts Garden-Path Comprehension: An ERP Study . . . . . . . . . . . . . . . . . . . . . 1751

Polly O'Rourke, Gregory Colesh

Support for a Deliberative Failure Account of Base-Rate Neglect: Prompting Deliberation


Increases Base-Rate Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1757

Natalie Obrecht, Dana Chesney

Agency concepts across cultures: How intuitive is folkpsychology? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1763

bethany ojalehto, Douglas Medin, Salino Garcia

xviii
A Comparison of Small Crowd Selection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1769

Henrik Olsson, Jane Loveday

Near-misses sting even when they are uncontrollable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1775

Desmond Ong, Noah Goodman, Jamil Zaki

Is statistical learning trainable? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1781

Luca Onnis, Matthew Lou-Magnuson, Hongoak Yun, Erik Thiessen

Causal reasoning in a prediction task with hidden causes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1787

Pedro A. Ortega, Daniel D. Lee, Alan A. Stocker

Getting From Here to There! : Testing the Eectiveness of an Interactive Mathematics


Intervention Embedding Perceptual Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1793

Erin Ottmar, David Landy, Robert Goldstone, Erik Weitnauer

You 're special, but it doesn't matter if you 're a greenhorn: Social recommender strategies for
mere mortals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1799

Pantelis P. Analytis, Daniel Barkoczi, Stefan Herzog

Upsetting the contingency table: Causal induction over sequences of point events . . . . . . . . . . . . . . . . 1805

Michael Pacer, Tom Griths

Assessing a Bayesian account of human gaze perception .......................................... 1811

Peter C. Pantelis, Daniel P. Kennedy

How Sharing Contexts Inuence Purchase Amounts: The Case of Food Choices ................. 1817

Jerey Parker, Nita Umashankar, Martin Schleicher

Memory distortions resulting from a choice blindness task ........................................ 1823

Philip Pärnamets, Lars Hall, Petter Johansson

Active learning as a means to distinguish among prominent decision strategies . . . . . . . . . . . . . . . . . . . 1829

Paula Parpart, Eric Schulz, Maarten Speekenbrink, Brad Love

Self-Directed Information Gathering Improves Learning in Young Children ...................... 1835

Eric Partridge, Matthew McGovern, Amanda Yung, Celeste Kidd

Learning mode and comparison in relational category learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1841

John Patterson, Kenneth Kurtz

Communicative Eciency and Miscommunication: The Costs and Benets of Variable Language
Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1847

Alexandra Paxton, Jennifer Roche, Michael Tanenhaus

Congenitally Deaf Children Generate Iconic Vocalizations to Communicate Magnitude .......... 1853

Marcus Perlman, Jing Paul, Gary Lupyan

Iconicity in English Vocabulary and its Relation to Toddlers' Word Learning . . . . . . . . . . . . . . . . . . . . . 1859

Lynn Perry, Marcus Perlman, Gary Lupyan

Anticipatory and Locally Coherent Lexical Activation Varies as a Function of Language


Prociency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1865

Ryan Peters, Theres Grüter, Arielle Borovsky

xix
Waynding and restructuring in a novel city: an insight problem solving task . . . . . . . . . . . . . . . . . . . . 1871

Judit Petervari, Amory H. Danek, Virginia L. Flanagin

Cognitive architecture and second-order systematicity: categorical compositionality and a


(co)recursion model of systematic learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1877

Steven Phillips, William Wilson

Computational principles underlying people's behavior explanations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1883

Aj Piergiovanni, Alan Jern

Speaker-specic generalization of pragmatic inferences based on prenominal adjectives . . . . . . . . . . . 1889

Amanda Pogue, Chigusa Kurumada, Michael Tanenhaus

Flexible Use of Phonological and Visual Memory in Language-mediated Visual Search . . . . . . . . . . . 1895

Daniel F. Pontillo, Anne Pier Salverda, Michael K. Tanenhaus

Mathematical Model of Developmental Changes in Number Cognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1901

Richard Prather

Shifting Covert Attention to Spatially Indexed Locations Increases Retrieval Performance of


Verbal Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1907

Anja Prittmann, Agnes Scholz, Josef Krems

Attacker and Defender Counting Approach for Abstract Argumentation . . . . . . . . . . . . . . . . . . . . . . . . . . 1913

Fuan Pu, Jian Luo, Yulai Zhang, Guiming Luo

Learning Additive and Substitutive Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1919

Ting Qian, Joseph Austerweil

Why do readers answer questions wrongly after reading garden-path sentences? . . . . . . . . . . . . . . . . . . 1925

Zhiying Qian, Susan Garnsey

Lateral Inhibition Overcomes Limits of Temporal Dierence Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1931

Jacob Rafati, David Noelle

Preferred Inferences in Causal Relational Reasoning: Counting Model Operations ............... 1937

Marco Ragni, Stephanie Schwenke, Christine Otieno

Generating Hyperdimensional Distributed Representations from Continuous-Valued Multivariate


Sensory Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1943

Okko Räsänen

Cross-situational cues are relevant for early word segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1949

Okko Räsänen, Heikki Rasilo

Computational evidence for eects of memory decay, familiarity preference and mutual
exclusivity in cross-situational learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1955

Heikki Rasilo, Okko Johannes Räsänen

Cognitive consequences of interactivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1961

Nan Renner

Eye Movements Reveal Sensitivity to Sound Symbolism Early and Late in Word Learning . . . . . . . 1967

Kate Pirog Revill, Laura Namy, Lynne Nygaard

The Attentional Learning Trap and How to Avoid It . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1973

Alexander Rich, Todd Gureckis

xx
What denes a category? Evidence that listeners' perception is governed by generalizations ..... 1979

Rachael Richardson, Naomi Feldman, William Idsardi

Transfer eects of prompted and self-reported analogical comparison and self-explanation . . . . . . . . 1985

J. Elizabeth Richey, Cristina D. Zepeda, Timothy J. Nokes-Malach

How do dierent training tasks modulate our perception and hemispheric lateralization in the
development of perceptual expertise? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1991

Tso Ricky Van-yip, Au Terry Kit-fong, Hsiao Janet Hui-wen

Eects of Complementary Control on the Coordination Dynamics of Joint-Action . . . . . . . . . . . . . . . 1997

Lillian Rigoli, Veronica Romero, Kevin Shockley, Gregory Funke, Adam Strang, Michael

Richardson

Development of selective attention in category learning ........................................... 2003

Samuel Rivera, Vladimir Sloutsky

Auditory Stimuli Slow Down Responses and First Fixations: Support for Auditory Dominance in
Adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2009

Christopher Robinson, Wesley Barnhart, Samuel Rivera

Capturing Social Motor Coordination: A comparison of the Microsoft Kinect, Video-Motion


Analysis and the Polhemus Latus Motion Tracking System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2015

Veronica Romero, Joseph Amaral, Paula Fitzpatrick, Richard Schmidt, Mike Richardson

Can Joint Action be Synergistic? Studying the Stabilization of Interpersonal Hand Coordination 2021

Veronica Romero, Rachel Kallen, Michael Riley, Mike Richardson

Do Markov Violations and Failures of Explaining Away Persist with Experience? ............... 2027

Benjamin Rottman, Reid Hastie

How Causal Mechanism and Autocorrelation Beliefs Inform Information Search . . . . . . . . . . . . . . . . . 2033

Benjamin Rottman

Children search for information as eciently as adults, but seek additional conrmatory evidence 2039

Azzurra Ruggeri, Tania Lombrozo, Tom Griths, Fei Xu

Restoring the Context of Interrupted Work with Desktop Thumbnails . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2045

Adam Rule, Aurélien Tabard, Karen Boyd, James Hollan

Representing and Learning a Large System of Number Concepts with Latent Predicate Networks 2051

Joshua Rule, Eyal Dechter, Joshua Tenenbaum

Helping students understand posterior probabilities: research with a digital learning environment
on the Monty Hall dilemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2057

Lore Saenen, Mieke Heyvaert, Wim Van Dooren, Patrick Onghena

The Moral Rhetoric of Climate Change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2063

Eyal Sagi, Timothy Gann, Teenie Matlock

Some Probability Judgments may Rely on Complexity Assessments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2069

Antoine Saillenfest, Jean-Louis Dessalles

How People Estimate Eect Sizes: The Role of Means and Standard Deviations . . . . . . . . . . . . . . . . . 2075

Motoyuki Saito

xxi
How do children construct the color lexicon? : Restructuring the domain as a connected system 2080

Noburo Saji, Michiko Asano, Midori Oishi, Mutsumi Imai

Tactile Experience Is Evoked by Visual Image of Materials:Evidence from Onomatopoeia . . . . . . . . 2086

Maki Sakamoto, Tatsuki Kagitani, Ryuichi Doizaki

Highlighting the Causal Meaning of Causal Test Questions in Contexts of Norm Violations . . . . . 2092

Jana Samland, Michael Waldmann

The inuence of hand or foot responses on response times in investigating action sentence
processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2098

Franziska Schaller, Sabine Weiss, Horst M. Müller

Gaze is not Enough: Computational Analysis of Infant's Head Movement Measures the
Developing Response to Social Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2104

Lars Schillingmann, Joseph Burling, Hanako Yoshida, Yuki Nagai

Large-scale investigations of variability in children's rst words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2110

Rose Schneider, Dan Yurovsky, Mike Frank

Assessing the Perceived Predictability of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2116

Eric Schulz, Josh Tenenbaum, David Reshef, Maarten Speekenbrink, Samuel Gershman

Learning and decisions in contextual multi-armed bandit tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2122

Eric Schulz, Emmanouil Konstantinidis, Maarten Speekenbrink

Motion perception of biological swarms ............................................................ 2128

Adriane Seiert, Sean Hayes, Caroline Harriott, Julie Adams

Task-set inhibition, conict, and the n-2 repetition cost: A computational model of task switching 2134

Nicholas Sexton, Richard Cooper

Neuronal Dynamics and Spatial Foraging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2140

Timothy Shea, Anne Warlaumont, Chris Kello, David Noelle

Examining the role of Inhibitory control in bilingual language switching . . . . . . . . . . . . . . . . . . . . . . . . . . 2146

Alison Shell, Jared Linck, L. Robert Slevc

Deliberate Practice Revisited: Complexity and Creativity in the Practice Process in Breakdance . 2152

Daichi Shimizu, Takeshi Okada

The Dynamics of Spoken Word Recognition in Second Language Listeners Reveal Native-Like
Lexical Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2158

Henna Shin, Brian Bauman, Imola MacPhee, Jason Zevin

Learning a Center-Embeddding Rule in an Articial Grammar Learning Task ................... 2164

Won Jae Shin, Kathleen Eberhard

Modelling Causal Reasoning under Ambiguity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2170

Yiyun Shou, Michael Smithson

Moral Reasoning as Probability Reasoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2176

Yiyun Shou, Fei Song

Predicting Meme Success with Linguistic Features in a Multilayer Backpropagation Network . . . . 2182

Keith Shubeck, Stephanie Huette

xxii
Tetris[U+0097]: Exploring Human Performance via Cross Entropy Reinforcement Learning
Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2188

Catherine Sibert, Wayne Gray, John Lindstedt

Children Learn Better When They Select Their Own Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2194

Zi L. Sim, Michelle M. Tanner, Nina Y. Alpert, Fei Xu

Toddlers Learn with Facilitated Play, Not Free Play . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2200

Zi L. Sim, Fei Xu

Memory Capacity Limits in Processing of Natural Connected Speech: The Psychological Reality
of Intonation Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2206

Heather Elizabeth Simpson, Fermin Moscoso del Prado Martin

Attention and Pattern Consciousness Reorganize the Cortical Topography of Event-Related


Potential Correlates of Visual Sequential Learning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2212

Sonia Singh, Jerome Daltrozzo, Chistopher Conway

Statistical and Chunking Processes in Adults' Visual Sequence Learning . . . . . . . . . . . . . . . . . . . . . . . . . . 2218

Lauren Slone, Scott Johnson

Understanding deverbal nominals: World knowledge or lexical semantics? . . . . . . . . . . . . . . . . . . . . . . . . 2224

Anastasia Smirnova

Prospective uncertainty: The range of possible futures in physical prediction ..................... 2230

Kevin Smith, Edward Vul

Let's Get Physical: Thinking with Things in Architectural Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2236

Daniel Smithwick, David Kirsh

Reading and writing direction eects on the aesthetic perception of photographs . . . . . . . . . . . . . . . . . . 2242

Chahboun Sobh, Flumini Andrea, Carmen Pérez González, I. Chris McManus, Julio

Santiago

Multiple Language Gender Identication for Blog Posts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2248

Juan Soler-Company, Leo Wanner

Elemental Causal Learning from Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2254

Kevin Soo, Benjamin Rottman

Exploring the processing costs of the "exactly" and "at least" readings of bare numerals with
event-related brain potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2260

Maria Spychalska, Jarmo Kontinen, Ira Noveck, Ludmila Roesch, Markus Werning

Attention dynamics in multiple object tracking .................................................... 2266

Nisheeth Srivastava, Ed Vul

Choosing fast and slow: explaining dierences between hedonic and utilitarian choices . . . . . . . . . . . 2272

Nisheeth Srivastava, Ed Vul

Watch out! - An instruction raising students' epistemic vigilance augments their sourcing
activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2278

Marc Stadtler, Johanna Maria Paul, Silke Globoschütz, Rainer Bromme

Social Situation Awareness: Empathic Accuracy in the Aircraft Cockpit . . . . . . . . . . . . . . . . . . . . . . . . . . 2284

Irene Stepniczka, Livia Tomova, Dominik Niedermeier, Markus Peschl, Claus Lamm

xxiii
Human behavior in contextual multi-armed bandit problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2290

Hrvo je Stojic, Pantelis P. Analytis, Maarten Speekenbrink

Toddlers Always Get the Last Word: Recency biases in early verbal behavior . . . . . . . . . . . . . . . . . . . . . 2296

Emily Sumner, Erika Deangelis, Mara Hyatt, Noah Goodman, Celeste Kidd

Generating Functions in Neural Learning of Sequential Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2302

Yanlong Sun, Hongbin Wang

How learners use feedback information: Eects of social comparative information and
achievement goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2308

Masayuki Suzuki, Tetsuya Toyota, Yuan Sun

Inhibition Failure is Mediated by a Disposition Toward Flexible Thinking ........................ 2314

Alexander Swan, Russell Revlin

M3 - Situating Embodied Learning: Embedding Gestures in Narratives to Learn Mathematical


FrActions in a digital tablet environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2320

Michael Swart, Benjamin Friedman, Sorachai Kornkasem, John B. Black, Jonathan Vitale

Memory foraging in a spatial domain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2326

Janelle Szary, Chris Kello, Rick Dale

Formation of an art concept: A case study using quantitative analysis of a contemporary artist's
interview data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2332

Kikuko Takagi, Akihiro Kawase, Sawako Yokochi, Takeshi Okada

Memory Processes of Sequential Action Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2338

Frank Tamborello, Gregory Trafton, Erik Altmann

The Eects of Criticism on Creative Ideation ..................................................... 2344

Yuko Tanaka, Yasuaki Sakamoto, Noboru Sonehara

Adaptive Perceptual Learning in Electrocardiography: The Synergy of Passive and Active


Classication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2350

Khanh-Phuong Thai, Sally Krasne, Philip Kellman

Improving Lexical Memory Access and Decision Making Processes Using Cognitive Word Games 2356

Kejkaew Thanasuan, Shane Mueller

Systemic Metaphors Promote Systems Thinking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2362

Paul Thibodeau, Anna Winneg, Cindy Frantz, Stephen Flusberg

Comparing Metaphors Reveals their Persuasive Capacity ......................................... 2368

Paul Thibodeau, Karlyn Gehring

Metaphors Aect Reasoning: Measuring Eects of Metaphor in a Dynamic Opinion Landscape . 2374

Paul Thibodeau, Peace Iyiewaure, Lera Boroditsky

The perception and memory of object properties: The role of attention, intention, and
information detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2380

Brandon Thomas, Michael Riley

An Account of Associative Learning in Memory Recall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2386

Robert Thomson, Aryn Pyke, Laura Hiatt, Greg Trafton

xxiv
Representations of Time Aect Willingness to Wait for Future Rewards ......................... 2392

Robert Thorstad, Aiming Nie, Phillip Wol

Building the mental timeline: Spatial representations of time in preschoolers . . . . . . . . . . . . . . . . . . . . . 2398

Katharine Tillman, Nestor Tulagan, David Barner

How Sharp is Occam's Razor? Language Statistics in Cognitive Processing ...................... 2404

Richard Tillman, Sterling Hutchinson, Max Louwerse

Addressee Backchannels Can Bias Third-Party Memory and Judgment . . . . . . . . . . . . . . . . . . . . . . . . . . . 2410

Jackson Tolins, Jean E Fox Tree

What drives Unconscious Multi-Attribute Decision-Making? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2416

Sabine Topf, Eddy Davelaar

ERP indices of situated reference in visual contexts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2422

Elli Tourtouri, Francesca Delogu, Matthew Crocker

Childhood SES aects anticipatory language comprehension in college-aged adults . . . . . . . . . . . . . . . . 2428

Melissa Troyer, Arielle Borovsky

Hypothesis-Space Constraints in Causal Learning ................................................. 2434

Pedro Tsividis, Josh Tenenbaum, Laura Schulz

Social Eye Cue: How Knowledge Of Another Person's Attention Changes Your Own . . . . . . . . . . . . 2440

Miles Tut, Matthias Gobel, Daniel Richardson

Constructing meaning: Material products of a creative activity engage the social brain . . . . . . . . . . . 2446

Kristian Tylén, Johanne Stege Bjørndahl, Andreas Roepstor, Riccardo Fusaroli

Why Stickiness is not Enough to Explain Persistence of Counterintuitive Religious Concepts . . . . 2452

M. Afzal Upal

Insight and cognitive ecosystems ................................................................... 2457

Frederic Vallee-Tourangeau, Sune Vork Steensen, Gaelle Vallee-Tourangeau, Angeliki

Makri

Predicting Lexical Norms Using a Word Association Corpus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2463

Hendrik Vankrunkelsven, Steven Verheyen, Simon De Deyne, Gert Storms

Goals Aect the Perceived Quality of Explanations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2469

Nadya Vasilyeva, Daniel Wilkenfeld, Tania Lombrozo

Explanations and Causal Judgments are Dierentially Sensitive to Covariation and Mechanism
Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2475

Nadya Vasilyeva, Tania Lombrozo

Emergence of systematic iconicity: transmission, interaction and analogy . . . . . . . . . . . . . . . . . . . . . . . . 2481

Tessa Verhoef, Sean Roberts, Mark Dingemanse

Manipulating the Contents of Consciousness: A Mechanistic-Manipulationist Perspective on


Content-NCC Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2487

Alfredo Vernazzani

Spatial Perception is Continuously Constrained by Goals and Memories . . . . . . . . . . . . . . . . . . . . . . . . . . 2493

David Vinson, Jerome Scott Jordan, Alycia Hund

xxv
Processing Overt and Null Subject Pronouns in Italian: a Cognitive Model . . . . . . . . . . . . . . . . . . . . . . . 2499

Margreet Vogelzang, Petra Hendriks, Hedderik van Rijn

Individual Belief Revision Dynamics in a Group Context ......................................... 2505

Igor Volzhanin, Ulrike Hahn, Martin Jonsson, Erik Olsson

Pattern Probabilities for Non-Dichotomous Events: A New Rational Contribution to the


Conjunction Fallacy Debate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2511

Momme von Sydow

The Tragedy of Inner-Individual Dilemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2517

Momme von Sydow

Verbal Synchrony in Large Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2523

Jorina von Zimmermann, Daniel Richardson

Executive Functions and Conceptual Change in Science and Mathematics Learning . . . . . . . . . . . . . . 2529

Stella Vosniadou, Dimitrios Pnevmatikos, Nikos Makris, Kalliopi Eikospentaki, Despina

Lepenioti, Anna Chountala, Giorgos Kyrianakis

Cross-Cultural Comparison of Peer Inuence on Discovery Rate during Play .................... 2535

Shirlene Wade, Celeste Kidd

Why is Number Word Learning Hard? Evidence from Bilingual Learners . . . . . . . . . . . . . . . . . . . . . . . . . 2541

Katie Wagner, Katherine Kimura, Pierina Cheung, David Barner

The Fundamental Attribution Error is rational in an uncertain world. . . . . . . . . . . . . . . . . . . . . . . . . . 2547

Drew Walker, Kevin Smith, Ed Vul

The Role of Working Memory in Melodic Perception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2553

Maegen Walker, Ahnate Lim, Scott Sinnett

The early emergence and puzzling decline of relational reasoning: Eects of prior knowledge and
search on inferring same and dierent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2559

Caren Walker, Sophie Bridgers, Alison Gopnik

Disambiguation Across the Senses: The Role of Discovery-Based Interference . . . . . . . . . . . . . . . . . . . . 2565

Jenna Wall, William Merriman

Condence Judgments and Eye Fixations Reveal Adults' Fractions Knowledge . . . . . . . . . . . . . . . . . . . 2571

Jenna Wall, Clarissa Thompson, Bradley Morris

Infant Locomotion, the Language Environment, and Language Development: A Home


Observation Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2577

Eric Walle, Anne Warlaumont

Power-law uctuations in eye movements predict text comprehension during connected text
reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2583

Sebastian Wallot, Beth O'Brien, Charles Coey, Damian Kelty-Stephen

Verbal Reports Reveal Strategies in Multiple-Cue Probabilistic Inference . . . . . . . . . . . . . . . . . . . . . . . . . . 2589

Matthew Walsh, Michael Collins, Kevin Gluck

Musical improvisation: Multi-scaled spatiotemporal patterns of coordination . . . . . . . . . . . . . . . . . . . . . . 2595

Ashley Walton, Mike Richardson, Peter Langland-Hassan, Anthony Chemero, Auriel

Washburn

xxvi
Modeling the Object Recognition Pathway: A Deep Hierarchical Model Using Gnostic Fields .... 2601

Panqu Wang, Garrison Cottrell, Christopher Kanan

Statistical Structures in Articial languages Prime Relative Clause Attachment Biases in English 2607

Felix Wang, Mythili Menon, Elsi Kaiser

Characterizing the Dierence Between Learning about Adjacent and Non-adjacent Dependencies 2613

Felix Wang, Toben Mintz

Interpersonal Anticipatory Synchronization: The Facilitating Role of Short Visual-Motor


Feedback Delays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2619

Auriel Washburn, Rachel Kallen, Charles Coey, Kevin Shockley, Michael Richardson

Reasoning about sentience and animacy: Children's and adults' inferences about the properties of
unseen entities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2625

Kara Weisman, Ellen Markman, Carol Dweck

A Computational Model for Learning Structured Concepts From Physical Scenes ................ 2631

Erik Weitnauer, David Landy, Robert Goldstone, Helge Ritter

Reducing overcondence in forecasting with repeated judgement elicitation ....................... 2637

Matthew Welsh, Steve Begg

Using Ground Truths to Improve Wisdom of the Crowd Estimates ............................... 2643

Andrew Whalen, Saiwing Yeung

When killing the heavy man seems right. Making people utilitarian by simply adding options to
moral dilemmas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2649

Alex Wiegmann, Karina Meyer

Transitivity is Not Obvious: Probing Prerequisites for Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2655

Eliane Wiese, Rony Patel, Jennifer Olsen, Ken Koedinger

We Readily Anchor Upon Others, But it is Easier to Anchor on the Self . . . . . . . . . . . . . . . . . . . . . . . . . 2661

Daniel Willard, Art Markman

A Domain-Independent Model of Open-World Reference Resolution .............................. 2667

Tom Williams, Matthias Scheutz

Visuo-spatial Working Memory and the Comprehension of Iconic Gestures . . . . . . . . . . . . . . . . . . . . . . . 2673

Ying Choon Wu, Bonnie Chinh, Seana Coulson

Contingent Labeling after Infants' Pointing Helps Infants Learn Words .......................... 2679

Zhen Wu, Julie Gros-Louis

A ne-grained understanding of emotions: Young children match within-valence emotional


expressions to their causes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2685

Yang Wu, Paul Muentener, Laura Schulz

Assessing Masked Semantic Priming: Cursor Trajectory versus Response Time Measures . . . . . . . 2691

Kunchen Xiao, Takashi Yamauchi, Casady Bowman

General Language Ability Predicts Talker Identication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2697

Xin Xie, Emily Myers

xxvii
A Computational Evaluation of Two Laws of Semantic Change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2703

Yang Xu, Charles Kemp

Semantic chaining and ecient communication: The case of container names . . . . . . . . . . . . . . . . . . . 2709

Yang Xu, Terry Regier, Barbara Malt

An adaptive cue combination model of spatial reorientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2715

Yang Xu, Terry Regier, Nora Newcombe

Assessing Emotions by Cursor Motions: An Aective Computing Approach . . . . . . . . . . . . . . . . . . . . . . 2721

Takashi Yamauchi, Hwaryong Seo, Yoonsuck Choe, Casady Bowman, Kunchen Xiao

Learning of Time Varying Functions is Based on Association Between Successive Stimuli . . . . . . . 2727

Lee-Xieng Yang, Tzu-Hsi Lee

Towards an empirical test of realism in cognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2733

James Yearsley, Emmanuel Pothos

Diagnosticity: Some theoretical and empirical progress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2739

James Yearsley, Emmanuel Pothos, Albert Barque-Duran, James Hampton

Learning of bimodally distributed quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2745

Saiwing Yeung, Andrew Whalen

Ecient analysis-by-synthesis in vision: A computational framework, behavioral tests, and


modeling neuronal representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2751

Ilker Yildirim, Tejas Kulkarni, Winrich Freiwald, Joshua Tenenbaum

Children's Online Processing of Ad-Hoc Implicatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2757

Erica J. Yoon, Yunan Charles Wu, Michael C. Frank

Linking Joint Attention with Hand-Eye Coordination  A Sensorimotor Approach to


Understanding Child-Parent Social Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2763

Chen Yu, Linda Smith

Understanding young children's imitative behavior from an individual dierences perspective . . . . 2769

Yue Yu, Tamar Kushnir

Signatures of Domain-General Categorization Mechanisms in Color Word Learning ............. 2775

Daniel Yurovsky, Katie Wagner, David Barner, Michael Frank

Rethinking the Conceptual History of the Term `Cognitive' ....................................... 2781

Nicholas Zautra

Consistency in Brain Activation Predicts Success in Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2787

Qiong Zhang, John R. Anderson, Robert E. Kass

Statistical Word Learning is a Continuous Process: Evidence from the Human Simulation
Paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2793

Yayun Zhang, Daniel Yurovsky, Chen Yu

No One Left Behind: How Social Distance Aects Life-Saving Decision Making . . . . . . . . . . . . . . . . . 2799

Yufeng Zhang, Haotian Zhou, Mo Luan, Hong Li

A Bayesian hierarchical model of local-global processing: Visual crowding as a case-study ....... 2805

Shunan Zhang, Man Song, Angela Yu

xxviii
In Search of Triggering Conditions for Spontaneous Visual Perspective Taking . . . . . . . . . . . . . . . . . . . 2811

Xuan Zhao, Corey Cusimano, Bertram F. Malle

The Impact of Granularity on Worked Examples and Problem Solving . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2817

Guojing Zhou, Thomas Price, Collin Lynch, Tiany Barnes, Min Chi

More Than a Blood Pump: An Experimental Enquiry of the Folk Theory of the Heart . . . . . . . . . . . 2823

Haotian Zhou, Cacioppo John

You say potato, I say t


udòu: How speakers of dierent languages share the same concept . . . . . . . . 2829

Benjamin Zinszer, Andrew Anderson, Olivia Kang, Thalia Wheatley, Ra jeev Raizada

Publication-Based Presentations

A Communal Exchange-based Framework for Cultural Evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2835

Liane Gabora

The Eco-Cognitive Model of Abduction (EC-Model). Is Abduction Really Ignorance-Preserving? 2837

Lorenzo Magnani

How the curse of intractability can be cognitive science's blessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2839

Iris van Rooij

Member Abstracts

Multiscale clustering of vocalizations during naturalistic infant-caregiver interactions . . . . . . . . . . . . 2841

Drew Abney, Anne Warlaumont, D. Kimbrough Oller, Sebastian Wallot, Chris Kello

Priming Dynamic-Kinematic Routines Using Spatial Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2842

Deanne Adams, Christopher Galeucia, Jennifer Kolesari, Laura Carlson, Kenny Coventry

Searching for the best functional comparison to isolate neural processes related to response
inhibtion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2843

Jacobo Albert, Alberto Sánchez-Carmona, Gerardo Santaniello, Sara López-Martín, Jose

Antonio Hinojosa

A Theory of Information Processing for Large-Scale Brain Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2844

Xerxes Arsiwalla, Paul Verschure

Cognitive representations of form in pop music: A probabilistic grammars approach . . . . . . . . . . . . . 2845

Richard Ashley

Cognitive Flexibility in Mathematics: Bilingual Children Show Cognitive Advantages . . . . . . . . . . . . 2846

Natsuki Atagi, Catherine Sandhofer

Tense systems across languages support ecient communication ................................. 2847

Geo Bacon, Yang Xu, Terry Regier

Modelling insight: The case of the nine-dot problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2848

Adrian Banks, Thomas Ormerod, Patrice Rusconi, Jim MacGregor

Interpreting Visualizations of Uncertainty on Smartphone Displays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2849

Trevor Barrett, Mary Hegarty, Grant McKenzie, Michael Goodchild

Cognitive productivity: Can cognitive science improve how knowledge workers' use IT to learn
from source material? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2850

Luc Beaudoin, Geneviève Gauthier, Phil Winne

xxix
It's all in the eye: multiple orders of motor planning in gaze control ............................. 2851

Anna Belardinelli, Martin V. Butz

Watching Fictive Motion in Action: Discourse Data from the TV News Archive . . . . . . . . . . . . . . . . . 2852

Till Bergmann, Teenie Matlock

Valence vs. Value in Decision-Making in Depression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2853

Nathaniel Blanco, W. Todd Maddox

Bridging the communicative gap between robots and humans, by analogy . . . . . . . . . . . . . . . . . . . . . . . . . 2854

Mark Blokpoel, Todd Wareham, J.p. de Ruiter, Pim Haselager, Ivan Toni, Iris van Rooij

Strategy dierences do not account for gender dierence in mental rotation . . . . . . . . . . . . . . . . . . . . . . 2855

Alexander Boone, Mary Hegarty

Individual Dierences in Coordinating Between Graphs and Equations of Functions: Eects of


CMR Facilitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2856

Julie Booth, Jennifer Cromley, Theodore Wills, Walt Stepnowski, Thomas Shipley, William

Zahner, Jessica Rossi

Assessing Two Dimensions of Gender Essentialism in Monolingual and Bilingual Adults . . . . . . . . 2857

Jacob Brodsky, Kevin Holmes

Perceptual Learning in Mathematics Produces Durable Encoding Improvements . . . . . . . . . . . . . . . . . . 2858

Carolyn Buord, Philip Kellman

Creating You-Are-Here Maps: Mapping location and orientation using photographs . . . . . . . . . . . . . . 2859

Heather Burte

Education, not age, predicts variable plural production in Yucatec Maya . . . . . . . . . . . . . . . . . . . . . . . . . . 2860

Lindsay Butler, Rosa Couoh Pool

An Automatized Heider-Simmel Story Generation Tool ........................................... 2861

Martin V. Butz, Robert Geirhos, Jan Kneissler

A Computational Modeling Approach to Understanding Gender Dierences in the Iowa Gambling


Task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2862

Kaileigh A. Byrne, Darrell A. Worthy

Selecting landmarks when giving directions to dierent addressees on campus . . . . . . . . . . . . . . . . . . . . 2863

Laura Carlson, Jennifer Kolesari, Christopher Galeucia, Deanne Adams

The space of spatial relations: An extended stimulus set .......................................... 2864

Alexandra Carstensen, Yang Xu, Charles Kemp, Terry Regier

A Puzzle for your thoughts: Information about the diculty of one task inuences preschoolers'
exploratory play with a novel toy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2865

Amanda Castro, Elizabeth Bonawitz

Using Wordless Picture Books during Shared Reading Boost Language Production in Preschoolers
2866

Leydi Chaparro-Moreno, Florencia Reali, Carolina Maldonado-Carreño

The Breadth and Depth of E-reading and Paper-reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2867

Jenn-Yeu Chen, Wan-Hsin Lee

xxx
Analyze Chinese Lexicon Project in the Chinese Character norms of traditional scripts . . . . . . . . . . 2868

Sau-chin Chen, Chung-Ching Wang, Jon-Fan Hu

Implicit Association in Mathematics and Science . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2869

Yuliya Chernykhovskaya, Carolyn Jennings, Maryam Tabatabaeian

Age dierences in information search: An exploration-exploitation tradeo model . . . . . . . . . . . . . . . . 2870

Jessie Chin, Evan Anderson, Chieh-Li Chin, Wai-Tat Fu

An Embodied Cognition Approach to Studying Emotional Words: The Impact of Positive Facial
Experiences on Semantic Properties Judgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2871

Ching Chu, Chi-Lin Yu, Ya-Yun Chuang, Yueh-Lin Tsai, Jon-Fan Hu

Conceptual Combination Modulated by Action using Tangible Computers ........................ 2872

Timothy Clausner, Mary Lou Maher, Berto Gonzales

Violence Metaphors in Presidential Debates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2873

Chelsea Coe, Till Bergmann, Teenie Matlock

Optimal stopping in a natural sampling task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2874

Anna Coenen, Todd Gureckis

Exemplar models can't see the forest for the trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2875

Nolan Conaway, Kenneth Kurtz

That's not the whole story: The role of reliability and credibility in evidential reasoning . . . . . . . . . 2876

Saoirse Connor Desai, David Lagnado

A holistic advantage in face drawing: higher accuracy when drawing upright faces . . . . . . . . . . . . . . . 2877

Jennifer Day, Nicolas Davidenko

Cultural Dierences in Fluid Collaboration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2878

Andrew Dayton, Barbara Rogo

Informative Transitions: A Heuristic for Conditionalized Causal Strength Learning . . . . . . . . . . . . . . 2879

Cory Derringer, Benjamin Rottman

A test of the somnolent mentation theory and the cognitive shue insomnia treatment . . . . . . . . . . 2880

Nancy Digdon, Luc Beaudoin

Natural language quantiers are exclusively linked to exact number skills . . . . . . . . . . . . . . . . . . . . . . . . . 2881

Sarah Dolscheid

Acoustic Correlates of Speaker Condence: Can They Tell I Don't Know? . . . . . . . . . . . . . . . . . . . . . . . 2882

Krystal Duchi, Alison Kristo, Schea Fissel, Jennifer Roche

Inuences of task diculty on initiation time and overall use of an external strategy . . . . . . . . . . . . 2883

Timothy Dunn, Evan Risko

Body-centric and world-centric components of the large-scale horizontal-vertical illusion . . . . . . . . . 2884

Frank Durgin, Zhi Li, Brennan Klein

Distinguishing the Recent Past from the Complicated Present in Recognition Memory ........... 2885

Melody Dye, Rich Shirin

A Computational Model of Emotion and Personality in Mastery Motivational Oriented Students 2886

Somayeh Fatahi, Hadi Moradi, Ali Nouri Zonoz

xxxi
Can priming intuitions about the logic of sets promote logical evaluations of conjunctive
probability judgments? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2887

Jenny Faure-Bloom, Gaëlle Vallée-Tourangeau, Frédéric Vallée-Tourangeau

Interactions of emoticon valence and text processing .............................................. 2888

Laurie Feldman, Kit W. Cho, Cecilia Aragon, Judith Kroll

A Triple-Stopping Threshold System For a Sequential Decision Task: A Cast-Net Stopping Rule
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2889

Mario Fic, Marcus Buckmann

16-month-olds use language to generate expectations about the visual world ...................... 2890

Allison Fitch, Patricia Ganea, Paul Harris, Zsuzsa Kaldy

Text Analytic Techniques in Survey Questionnaire Development and Analysis ................... 2891

John Ford

Wisdom of Randomly Assembled Small Crowds ................................................... 2892

Mirta Galesic, Daniel Barkoczi, Konstantinos Katsikopoulos

Using Advance Organizers to Improve Learning from Video . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2893

Emma Geller, James Stigler

The eects of spatial anxiety on memory for spatio-temporal scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2894

Devin Gill, Jeanine Stefanucci, Sarah Creem-Regehr, Erica Barhorst

Analogical reasoning performance and organization is inuenced by the type of semantic


distractors: an investigation with adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2895

Yannick Glady, Bob French, Jean-Pierre Thibaut

Individual Dierences, Conrmation, and the Consideration of Alternative Causes . . . . . . . . . . . . . . 2896

Kelly Goedert, Michelle Ellefson, Victoria Kerns

The Increased Use of Tablets In Education: Why Physical Learning Is Sometimes Better . . . . . . . . 2897

Sara Goodman, Travis Seymour, Barrett Anderson

Exploring the mechanism of context-dependent memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2898

Chelsea Gordon, Michael Spivey

Economic Behavioral and Semantic Analysis of Generosity and Fairness in L'Arche Caregivers . 2899

Mark Graves, Kevin Reimer, Andrea Beckum, Shaina Smith, Remya Nair, Michael Spezio,

Warren Brown, Steven Quartz

Do infants compare ratios or use simpler heuristics in probabilistic inference? . . . . . . . . . . . . . . . . . . . 2900

Samantha Gualtieri, Elizabeth Bonawitz, Stephanie Denison

How are interaction between human and an autonomous agent aected by embodiments and
voice?: Investigation with age groups comparison. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2901

Etsuko, T. Harada, Riko Hasegawa, Wataru Kayano, Hirotaka Osawa

Which way to present product information is best for higher purchase intention . . . . . . . . . . . . . . . . . . 2902

So-eun Her, Kwanghee Han

Social Inuences on the Spatial Perspective-Taking Abilities of Males and Females .............. 2903

Nahal Heydari, Mary Hegarty, Margaret Tarampi

xxxii
Twelve-month-olds dierentiate between typical and atypical conversational timing .............. 2904

Elma E. Hilbrink, Marisa Casillas, Imme L. Lammertink, Stephen C. Levinson

Eects of lined traces and hand motion in underlining sentences on comprehension . . . . . . . . . . . . . . 2905

Misaki Horie, Sachiko Kiyokawa

The inuence of an inherence heuristic on scientic explanation ................................. 2906

Zachary Horne, Andrei Cimpian

The Relationship Between Empathy and Humor use in Adolescents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2907

Yong-Ru Hsiao, Yueh-Lin Tsai, Hsueh-Chih Chen, Jon-Fan Hu

The Relationship between Theory of Mind Abilities and Humor Comprehension . . . . . . . . . . . . . . . . . . 2908

Jon-Fan Hu, Yueh Lin Tsai, Yong-Ru Hsiao, Yu-Chen Lin, Liang-Yu Shen, Li Tsao,

Yu-Chen Chan, Hsueh-Chih Chen

Understanding the Cone of Uncertainty: Non-expert interpretations of hurricane forecast


uncertainty visualizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2909

Ian T. Ruginski, Alexander P. Boone, Lace M.k. Padilla, Mary Hegarty, William B.

Thompson, Donald H. House, Sarah H. Creem-Regehr

The Color of Music: Synesthesia or emotion-mediated cross-modal associations? . . . . . . . . . . . . . . . . 2910

Erin Isbilen, Carol Lynne Krumhansl

The Eects of Art Experience, Competence in Artistic Creation, and Methods of Appreciation on
Artistic Inspiration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2911

Chiaki Ishiguro, Takeshi Okada

Eect of language on discrimination between warm and cold color hues. . . . . . . . . . . . . . . . . . . . . . . . . . 2912

Kirill Istomin, Irina Ilina, Oleg Uliashev

The specicity of the labeling eect on memory: what kinds of labels improve retrieval? . . . . . . . . . 2913

Anja Jamrozik, Dedre Gentner

Dogmas of Understanding in Western Art Music Performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2914

Linda T. Kaastra

Understanding developmental bottlenecks in active inquiry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2915

George Kachergis, Marjorie Rhodes, Todd Gureckis

Use of Lexical Statistics for Compound Word Recognition and Segmentation in Turkish ......... 2916

Ozkan Kilic

NARS as a Normative Model of Cognition ........................................................ 2917

Ozkan Kilic, Pei Wang

Social categories as `excluders': Explaining stereotyping with connectionist modeling . . . . . . . . . . . . . 2918

Andre Klapper, Ron Dotsch, Iris van Rooij, Daniel Wigboldus

A Computational Model of Jazz Improvisation Inspired by Language ............................. 2919

Cody Komers, Alan Yuille

Using Real-Time Computational Modeling to Individually Optimize Speech Category Learning .. 2920

Seth Koslov, Nathaniel Blanco, Bharath Chandrasekaran, Todd Maddox

Figurative and Literal Action-Sentence Compatibility Eect in Japanese . . . . . . . . . . . . . . . . . . . . . . . . . 2921

Soichi Kozai, Katsunori Kotani, Markane Sipraseuth

xxxiii
Finding Meaning in Neuroaesthetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2922

Alexander Kranjec, Julia Sienkewicz, Corey Robinson, Amanda Buchheit

A Case-Based Reasoning Approach to Providing High-Quality Feedback on Computer


Programming Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2923

Angelo Kyrilov, David Noelle

Linear Versus Non-Linear Policy Capturing in a Dynamic Classication Task . . . . . . . . . . . . . . . . . . . 2924

Daniel Lafond, Benoit Roberge-Valliere, Marie-Eve Saint-Louis, Sebastien Tremblay

The Eects of Worked Examples on Transfer of Statistical Reasoning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2925

Marianna Lamnina, Daniel Fienup

The colors and textures of musical sounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2926

Thomas Langlois, Joshua Peterson, Stephen Palmer

"No way!": Similar contribution of visual and auditory cues to sarcasm comprehension . . . . . . . . . 2927

Alina Larson, John Collins, Nicolas Davidenko

Semantic Richness Eects in Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2928

Mabel Lau, Winston Goh, Melvin Yap

Long-Term Memory and Working Memory can be Improved by Cognitive Training . . . . . . . . . . . . . . . 2929

David, M. Lim, Michael, D. Patterson

Vocabulary Size is Correlated with Non-Native Tone Sensitivity In English Learning Infants .... 2930

Candise Lin, Toben Mintz

The Cognitive Niches of Knowledge-Based Decision Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2931

Daniela Link, Julian Marewski

Belief in the unbelievable: The relationship between tendencies to believe pseudoscience,


paranormal, and conspiracy theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2932

Emilio Lobato, Corinne Zimmerman

Modern Symbolic Communication Through Non-Word Text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2933

Milagros Florentina Lopez Obeso, Morgan Magnus Fleming

P3 as a neural index of response inhibition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2934

Sara López-Martín, Jacobo Albert, Sandra Hoyos, Alberto Sánchez-Carmona, Luis Carretié

Topological Relations between Objects Are Categorically Coded . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2935

Andrew Lovett, Steve Franconeri

The Role of Embodiment on Children's Understanding and Motivation in Science Learning ..... 2936

Carol M. Lu, John B. Black

Inferring causal structure and hidden causes from event sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2937

Christopher Lucas, Kenneth Holstein, Michael Pacer

Argument Strength Computation Based on Satisability Degree and Agents' Beliefs . . . . . . . . . . . . . . 2938

Jian Luo, Fuan Pu, Guiming Luo

Asymmetry of causal inference in reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2939

Yingyi Luo, Manami Sato, Yunzhu Wang, Satoshi Ito, Hiromu Sakai

xxxiv
Does tactile softness and hardness alter our acceptance of utilitarian judgment? . . . . . . . . . . . . . . . . . 2940

Yoshimasa Majima, Hiroko Nakamura

Neural precursors of decisions that matter  an ERP study of the role of consciousness in
deliberate and random choices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2941

Uri Maoz, Liad Mudrik, Ram Rivlin, Gideon Yae, Ralph Adolphs, Christof Koch

Varying Eects of Subgoal Labeled Procedural Instructions in STEM Learning . . . . . . . . . . . . . . . . . . . 2942

Lauren Margulieux, Richard Catrambone

The role of text in scientic reasoning: Priming misconceptions can facilitate learning . . . . . . . . . . 2943

Amy Masnick, Kristin Weingartner, Marisa Cohen

Speech and Print: Two Dierent Communication Media and Implications for Acquiring Literacy
Naturally . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2944

Dominic Massaro

Acquisition of perceptual knowledge via information search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2945

Miki Matsumuro, Kazuhisa Miwa, Hitoshi Terai, Misaki Kurita

How soon is now? The language of timing in joint activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2946

Gregory Mills

Invertible signals: A challenge for theories of communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2947

Jennifer Misyak, Takao Noguchi, Nick Chater

Harmonization eects between a word's meaning and typography: An investigation using the
visual world paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2948

Kozue Miyashiro, Etsuko T. Harada

Language input from child-directed speech and children's picture books are dierent . . . . . . . . . . . . . . 2949

Jessica Montag, Michael Jones, Linda Smith

Cognitive Modeling of Life Story: Reconstructing Our Memories from a Photo Library . . . . . . . . . . 2950

Junya Morita, Takatsugu Hirayama, Kenji Mase, Kazunori Yamada

How semantic is unconscious semantic integration? A visual masking study ..................... 2951

Liad Mudrik, Nathan Faivre, Sid Kouider, Christof Koch

A Computational Account of Novel Word Generalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2952

Aida Nematzadeh, Erin Grant, Suzanne Stevenson

How bookies make your money . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2953

Philip Newall

Choice Facilitates 4-Year-Olds' Cognitive Flexibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2954

Allison O'Leary, Vladimir Sloustky

Alternating Estimation of Local Objective and Global Purpose by Two-Layer Model of


Emphasizing Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2955

Yoshimasa Ohmoto, Asami Matsumoto, Toyoaki Nishida

The Eect of the Structural Dierences of Concepts on Learning by Drawing versus Reading
Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2956

Kayoko Ohtsu

xxxv
Individual dierences in the use of cues during insight problem solving . . . . . . . . . . . . . . . . . . . . . . . . . . . 2957

Ryo Orita, Masasi Hattori

Individual dierences in older adults' working memory capacity and speed of using touch
interfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2958

Kazunori Otsuka

How did Homo Heuristicus become ecologically rational? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2959

Maria Otworowska, Marieke Sweers, Robin Wellner, Todd Wareham, Iris van Rooij

Describing Causal Events: Evidence from Patients with Focal Brain Injury . . . . . . . . . . . . . . . . . . . . . . 2960

Demet Ozer, Idil Bostan, Anjan Chatterjee, Tilbe Goksun

Sex Dierences in Virtual Navigation Inuenced by Scale, Visual Cues, Spatial Abilities and
Lifetime Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2961

Lace Padilla, Sarah Creem-Regehr, Jeanine Stefanucci, Elizabeth Cashdan

Does Learning Magnitude Knowledge help Students Learn Procedural Knowledge or Vice Versa? 2962

Rony Patel, Ken Koedinger

Finger Gnosis And Symbolic Number Comparison as Robust Predictors of Adult Numeracy . . . . . 2963

Marcie Penner-Wilger, Rylan Waring, Adam Newton, Cindel White

Emotionally mediated crossmodal correspondences aect classication performance . . . . . . . . . . . . . . 2964

Joshua Peterson, Stephen Palmer

Semantic, not positional distances between words aect processing diculty for sentences with
relative clauses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2965

Fenna Poletiek, Jun Lai

A PDP Account of Transitions in Conceptual Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2966

Robert Powers, David Plaut

How is the result of the categorization process represented? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2967

Sandeep Prasada

Phonetic abilities of walking and crawling infants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2968

Gina Pretzer, Anne Warlaumont, Eric Walle

Implicit learning in dynamic decision making: A glass-box approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2969

Sylvain Pronovost, Marie-Ève St-Louis, Daniel Lafond, Jean-François Gagnon, Sébastien

Tremblay

An ERP study of syntactic anomaly processing in Mandarin sentences . . . . . . . . . . . . . . . . . . . . . . . . . . . 2970

Zhiying Qian, Susan Garnsey

Emotion and Morality: The Main Factors In Moral Judgment and Moral Behaviour ............ 2971

Nalini Ramlakhan

Yes, No, Maybe So: The Eect of Ambiguity, Falsication, and Conrmation on
Re-Categorization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2972

Jared Ramsburg, Stellan Ohlsson

The Social Evolution and Communicative Function of Noun Classication . . . . . . . . . . . . . . . . . . . . . . . 2973

Michael Ramscar, Melody Dye, Petar Milin, Richard Futrell

xxxvi
Sensitivity to communicative norms when deceiving without lying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2974

Keith Ransom, Wouter Voorspoels, Amy Perfors, Daniel Navarro

Modeling the Role of Hippocampus in Extinction and Spontaneous Recovery . . . . . . . . . . . . . . . . . . . . . . 2975

Jerey Rodny, David Noelle

Cultural consensus modeling of Tibetan Buddhist concepts in cognitive science: Enhancing


cross-cultural science education through mutual understanding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2976

Michael Romano, Geshe Dadul Namgyal, Tsondue Samphel, Carol Worthman

Inuence of High and Low Groove Music on Postural Sway Dynamics ........................... 2977

Jessica Ross, Anne Warlaumont, Lillian Rigoli, Ramesh Balasubramaniam

Asking useful questions: Active learning with rich queries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2978

Anselm Rothe, Brenden Lake, Todd Gureckis

Brain activities related to target- versus trajectory-based strategies in visually-guided movement


control: A functional MRI study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2979

Je-Kwang Ryu, Hee Sun Eum, Kyoung-Min Lee

Accuracy and awareness of image veracity in human perceptions of manipulated and


unmanipulated images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2980

Caldwell Sabrina, Gedeon Tamás, Jones Richard, Copeland Leana

Individual Dierences in Base-rate Neglect: A Computational Dual Process Model . . . . . . . . . . . . . . . 2981

Carlos Salas, Tim Sparer, Sabrina Velez, Thomas Grin

Inference, Not Dilution in the Dilution Eect ..................................................... 2982

Adam Sanborn, Takao Noguchi, James Tripp, Neil Stewart

Giving dyads the silent treatment: Anticipatory joint action and the need for external action
feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2983

Daniel Schloesser, Jiuyang Bai, Jerome Scott Jordan

Apple, pomme, manzana: Productive vocabulary and cognitive exibility in bilingual preschoolers 2984

Christina Schonberg, Natsuki Atagi, Catherine Sandhofer

The Eect of Spatial Representations on Discounting Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2985

Andrea J. Sell, Terry Spehar-Fahey, Michael Gagliardo

The eect of empathy on comprehension and attitude in text reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2986

Hideaki Shimada

Decreasing Music Familiarity Increases Incorporation of Music Themes in a Generation Task .. 2987

Cynthia Sifonis, Jonathan Saulter, William Fuss

Contextual determinants of category-based expectations during single-word recognition . . . . . . . . . . . 2988

Francis Smith, Danielle Reece, Padraic Monaghan, Morten Christiansen, Thomas Farmer

Distributed Cognition in the Age of Distributed Systems .......................................... 2989

Ethan Soutar-Rau, Brian Fisher

Alien Bacteria Found on Mars! A Model of Conceptual Change using the Re-categorization
Paradigm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2990

Tim Sparer, Jared Ramsburg, Carlos Salas, Stellan Ohlsson

xxxvii
Introducing the Cognitive Systems Institute Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2991

Jim Spohrer

The spiral of anxiety: a cognitive account . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2992

Nisheeth Srivastava

Topological Dependence of Rate Code Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2993

William B. St. Clair, David C. Noelle

Is the listener really listening? Exploring the eect of verbal and gestural speaker cues on
backchanneling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2994

Matthew Stave, Eric Pederson

Learning with Concrete and Virtual Manipulative Models: Are Models Scaolds or Crutches? .. 2995

Andrew Stull, Mary Hegarty

Real-world implementation of Newcomb's thought experiment, using mouse-tracking techniques . 2996

Maryam Tabatabaeian, Shaun Pilkington, Rick Dale

Activation and Rejection of Irrelevant Meaning in Simile Sentences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2997

Tomohiro Taira

Imagine That: The Relationship between Imagery Measures and Imagery Types . . . . . . . . . . . . . . . . . . 2998

Margaret Tarampi, Boris Khanukayev, Rebecca Schaefer

Does prior knowledge reveal cognitive and metacognitive processes during learning with a
hypermedia-learning system based on eye-tracking data? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2999

Michelle Taub, Jesse J. Farnsworth, Roger Azevedo

Perceptual Learning with Adaptively-triggered Comparisons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3000

Khanh-Phuong Thai, Sally Krasne, Philip Kellman

A Foreign Language Eect or a Language Prociency Eect? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3001

Paul Thibodeau, Evelyn Kalafus-Mastenbrook, Matias Berretta, Aliya Tuzhilin, Nupur

Agrawal

Multiple Strategies in Conjunction and Disjunction Judgments: Most People are Normative Part
of the Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3002

James Tripp, Adam Sanborn, Neil Stewart, Takao Noguchi

The dierences of semantic features between Chinese concrete, abstract, and emotional concept . 3003

Yueh-Lin Tsai, Chi-Lin Yu, Yong-Ru Hsiao, Shu-Ling Cho, Hsueh-Chih Chen, Jon-Fan Hu

Which Algorithms Can and Can't Learn Identity Eects in Phonological Grammars . . . . . . . . . . . . . 3004

Paul Tupper

Induction with Familiar and Newly-Learned Categories in Young Children . . . . . . . . . . . . . . . . . . . . . . . 3005

Layla Unger, Anna Fisher

An Empirical Examination of Barrett's Intuitive Expectation Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3006

M. Afzal Upal

"The baking stick thing": Automatization of co-speech gesture during lexical access . . . . . . . . . . . . . . 3007

Prakaiwan Vajrabhaya, Eric Pederson

Social network structure contributes to dierences in language use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3008

David Vinson, Rick Dale

xxxviii
Investigating the Visual/Analytic Shift in Students' Knowledge in Chemistry .................... 3009

Maria Vlacholia, Stella Vosniadou, Katerina Salta, Petros Roussos, Smaragda Kazi,

Michael Sigalas, Chrysa Tsougraki

Gricean maxims inuence inductive inference with negative observations . . . . . . . . . . . . . . . . . . . . . . . . . 3010

Wouter Voorspoels, Daniel Navarro, Amy Perfors, Keith Ransom

Interdependence of Fixations and Saccades . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3011

Sebastian Wallot, Charles Coey, Mike Richardson

Congural and featural face processing are modulated by spatial attention: evidence from
event-related brain potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3012

Hailing Wang, Shimin Fu

Cross-situational Word Learning Results in Explicit Memory Representations . . . . . . . . . . . . . . . . . . . . 3013

Felix Wang, Toben Mintz

Developing an Integrated and Comprehensive Traditional Chinese Corpus Based on


Multi-Character Words for Studying relations between words and lexicons . . . . . . . . . . . . . . . . . . . . . . . . 3014

Chung-Ching Wang, Sau-chin Chen, Yueh Lin Tsai, Yong-Ru Hsiao, Jon-Fan Hu

Getting what you Ordered: Symbolic and Non-Symbolic Ordinality as Predictors of Exact and
Approximate Calculation in Adults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3015

Rylan J. Waring, Marcie Penner-Wilger

Culture, causal attributions, and development: A comparison of Chinese and U.S. 4-and
6-year-olds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3016

Adrienne Wente, Sophie Bridgers, Xin Zhao, Yixin Cui, Elizabeth Seiver, Li Zhanxing, Liqi

Zhu, Alison Gopnik

Context vs. Compositionality: How Do Context-induced Ad-hoc Aordances Interact with


Semantically Stored Telic Information?  An ERP Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3017

Markus Werning, Jarmo Kontinen, Erica Cosentino

Multisensory Integration Induces Body Ownership of an External Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . 3018

Veronica Weser, Gianluca Finotti, Dennis Prott

Semantic Processing in the Context of the PRP Paradigm: Structurally or Strategically


Bottlenecked? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3019

Darcy White, Derek Besner

Historical Cognition: An Investigation of Factors Aecting Reasoning about Historical Causality 3020

Cindel White, Marcie Penner-Wilger, Graham Broad

Perspective Taking in Communicative Pointing: An Optimal Feedback Control Modeling


Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3021

Tobias Winner, Luc Selen, Lennart Verhagen, Pieter Medendorp, Ivan Toni, Iris van Rooij

Gestures in the TV News reect mental number space: "Tiny" and "low" numbers . . . . . . . . . . . . . . 3022

Bodo Winter, Marcus Perlman, Teenie Matlock

Children's ability to infer beliefs and desires from emotional reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3023

Yang Wu, Chris Baker, Josh Tenenbaum, Laura Schulz

A model comparison on perception of arm movements in point-light display . . . . . . . . . . . . . . . . . . . . . . 3024

Reiko Yakushijin, Sachiyo Ueda

xxxix
Adaptation to Unexpected Word-Forms in Highly Predictive Sentential Contexts . . . . . . . . . . . . . . . . . 3025

Shaorong Yan, Thomas Farmer

Neural Basis of Episodic Memory Development: Evidence from Single Nucleotide Polymorphisms 3026

Hyungwook Yim, Simon Dennis, Christopher Bartlett, Vladimir Sloutsky

Do we use L1 probabilistic phonotactics in L2 listening? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3027

Michael C. W. Yip

Using false belief task to explore the eect of empathy situation on Theory of Mind function . . . . 3028

Chi-Lin Yu, Min-Ying Wang, Pei-Wen Chen, Joe-Yi Yap, Jen-Shen Chang, Yong-Ru Hsiao,

Jon-Fan Hu

What senses of agency can infants have? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3029

Lorijn Zaadnoordijk, Sabine Hunnius, Marlene Meyer, Johan Kwisthout, Iris van Rooij

Categorical Perception of Labeled and Unlabeled ASL Facial Expressions in Hearing Non-signers 3030

Hadar Zeigerson, Kevin Holmes

Capturing the relations between metacognition, self-explanation, and analogical comparison: An


exploration of two methodologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3031

Cristina D. Zepeda, Timothy J. Nokes-Malach

Learning multiple kinds of associations during cross-situational word learning ................... 3032

Martin Zettersten, Erica Wojcik, Viridiana L. Benitez, Jenny Saran

Statistical learning of auditory patterns as trajectories through a perceptually dened similarity


space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3033

Jason Zevin, Hao Wang

Author Index ......................................................................................... 3034

Reviewers List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3054

xl
Dear Cognitive Scientists,
Welcome to Pasadena, California for the 37th Annual Conference of the Cogni-
tive Science Society! Our meeting brings together some of the most innovative
and exciting research in cognitive science. The program features cutting-edge
plenary talks by three international figures from diverse areas: Martha Farah, Rich
Ivry, and Rosalind Picard. It also includes three invited symposia aimed at show-
casing the three broad themes for this year: Mind, Technology, and Society.

CogSci 2015 received 882 submissions, including 669 full papers, 186 member ab-
stracts, 3 publication-based presentations, 13 symposium proposals, and 10
workshop and tutorial proposals. After a rigorous review process, we selected
187 papers for oral presentation (28%), 287 papers for poster presentation (43%),
185 member abstracts for poster presentation, 2 publication-based talks, and 6
symposia. All 10 tutorials and workshop submissions were deemed promising and
of wide interest, and were accepted (though one workshop was withdrawn).

We hope that you enjoy the program this year and the city of Pasadena, one of
the most culturally interesting and diverse cities on the West Coast. There are
hundreds of shops and restaurants near the convention center, and it is a con-
venient access point to greater Los Angeles and Southern and Central Califor-
nia. We encourage you to set aside the time to enjoy some of the many activi-
ties the region has to offer.

Your Hosts,

Rick Dale, Carolyn Jennings, Paul P. Maglio, Teenie Matlock, David C. Noelle,
Anne Warlaumont, and Jeff Yoshimi

xli
Acknowledgments
We are grateful to everyone who contributed to the planning and organization
of this year’s meeting and to all reviewers who generously donated time to eval-
uate submissions. We thank the 172 Program Committee members and over
1,000 reviewers who were essential to the review process. We thank the awards
committee who helped assess the prize-winning papers. All committee members
are listed on Pages 4-5 and all reviewers are listed in the proceedings.

We are especially grateful for the assistance of a number of individuals and


groups for many organizational aspects of the meeting. We thank Jessica Wong,
the Cognitive Science Society’s Conference Officer, for managing pretty much
everything; Deborah Gruber, the Cognitive Science Society’s Business Manager,
for handling the business details; Richard Catrambone, Nora Newcombe, and
Susan Goldin-Meadow, and the Cognitive Science Society’s Executive Commit-
tee and Governing Board for their constant support and advice; the Scarritt
Group, Inc., for logistics and support; James Stewart of Precision Conference So-
lutions for maintaining and improving the conference reviewing system; Bodo
Winter, UC Merced, for helping review member abstracts; Greg Bryant, UCLA, for
sharing local knowledge of the Pasadena and Los Angeles areas; Martin Butz
and Anna Belardinelli, University of Tübingen, who wrangled the student volun-
teers; the student volunteers, listed on Page 5, who keep the conference running
smoothly on site; and importantly our conference sponsors, listed on Page 9,
whose generous support enabled us to focus more on the program and less on
the finances.

We also are also thankful for the support of the Cognitive Science community.
We are sure that the suggestions and contributions we received from this com-
munity helped us to make it a more lively, engaging and fun meeting for all.
Enjoy!

Rick Dale, Carolyn Jennings, Paul P. Maglio, Teenie Matlock, David C. Noelle,
Anne Warlaumont, and Jeff Yoshimi

Organizing Committee, Cognitive Science 2015

xlii
Committees

Organizing Committee: Rick Dale


Carolyn Jennings
Paul P. Maglio
Teenie Matlock
David C. Noelle
Anne Warlaumont
Jeff Yoshimi
University of California, Merced

Organizing Committee Chair: Paul P. Maglio

Program Chairs: David C. Noelle and Rick Dale

Member Abstracts Chairs: Teenie Matlock and Jeff Yoshimi

Tutorials & Workshops Chair: Anne Warlaumont


Awards Committee: Rick Dale, chair
Chen Yu, Indiana University
Florencia Reali, Los Andes University
Paul Bello, Office of Naval Research
Max Louwerse, Tilburg University
Student Volunteer Chairs: Martin Butz and Anna Belardinelli
University of Tübingen

Website Coordinators: Jeff Yoshimi and Rick Dale

Artwork: Teenie Matlock

Conference Officer: Jessica Wong, University of Chicago

xliii
Program Committee Members
Dor Abrahamson Morteza Dehghani Gary Jones
Afra Alishahi James Dixon Michael Jones
Erik Altmann Alex Doumas Ion Juvina
Michael Anderson Susan Epstein Michael Kalish
Mark Andrews Thomas Farmer Vsevolod Kapatsinski
Blair Armstrong Caitlin Fausey Irvin Katz
Inbal Arnon Michele Feist Mark Keane
Sudha Arunachalam Jozsef Fiser Christopher Kello
Richard Ashley Anna Fisher William Kennedy
Thomas Barkowsky Ken Forbus Sangeet Khemlani
Mike Barley Mike Frank Celeste Kidd
Daniel Bartels Stefan Frank Agnieszka Konopka
Sieghard Beller Christian Freksa Stefan Kopp
Andrea Bender Bob French Nate Kornell
Sven Bertel Daniel Freudenthal Kenneth Kurtz
Brad Best Wai-Tat Fu Chigusa Kurumada
Perrin Bignoli Danilo Fum David Landy
Dorrit Billman Riccardo Fusaroli Michael Lee
Stephen Blessing Alexia Galati George Luger
Jean-Francois Bonnefon Timothy Gann Gary Lupyan
Anna M. Borghi Dedre Gentner Dermot Lynott
Will Bridewell Ashok Goel Lorenzo Magnani
Henry Brighton Laura Gonnerman Hanspeter Mallot
Monica Bucciarelli Cleotilde Gonzalez Art Markman
Bruce Burns Noah Goodman Ralf Mayrhofer
Daniel Casasanto Lisa Grimm Russ McBride
Sanjay Chandrasekharan Maurice Grinberg Marek McGann
Franklin Chang Prahlad Gupta Craig R. M. McKenzie
Ron Chrisley Todd Gureckis Padraic Monaghan
Morten Christiansen Thomas Hannagan Fermin Moscoso del Pra-
Catherine Clement Stephen José Hanson do Martin
Jonathan Cohen Mary Hegarty Gregory Murphy
Christopher Conway Sebastien Helie Christopher Myers
Fintan Costello Shohei Hidaka Yuki Nagai
Seana Coulson Thomas Hills Hansjoerg Neth
Scotty Craig Eva Hudlicka Sergei Nirenburg
David Danks Stephanie Huette Timothy Nokes-Malach
Judith Danovitch Alistair Isaac Nader Noori
Eddy Davelaar Robert Jacobs Howard Nusbaum
Felipe De Brigard T. Florian Jaeger Marta Olivetti
Gedeon Deák Christian Janssen Andrew Olney

xliv
Luca Onnis Kevin Shockley Mikkel Wallentin
Daniel Osherson Patrick Simen Hongbin Wang
John Pani Chris Sims Anne Warlaumont
David Pautler Michael Spivey Takeo Watanabe
David Peebles Terrence Stewart Markus Werning
Mark Pitt David Stracuzzi Jon Willits
Marie Postma-Nilsenovia Ron Sun Sharon Wood
Marco Ragni Naveen Sundar Eldad Yechiam
Stephen Reed Jakub Szymanik Jeffrey Yoshimi
Marjorie Rhodes Thora Tenbrink Robert Youmans
Daniel Richardson Josh Tenenbaum Jason Zevin
Mike Richardson Paul Thagard Corinne Zimmerman
Kai-Florian Richter Jim Thompson
Etienne Roesch Joe Toscano
Timothy Rogers Jennifer Trueblood
Yasuaki Sakamoto Barbara Tversky
Franz Schmalhofer Kristian Tylén
Ute Schmid M. Afzal Upal
Mike Schoelles Frederic Vallee-
Christian Schunn Tourangeau
John Schwenkler Sashank Varma
Colleen Seifert Paul Vogt

Student Volunteers
Nicole Beckage, University of Colorado, Boulder
Davie Floyd, Columbia University
Sarah Green Goodman, University of California, Santa Cruz
Theodros Haile, University of Massachusetts
Katherine Kimura, University of California, Berkeley
Daniel Lenzen, University of California, San Diego
Rachel A. Lerch, Drexel University
Butovens Médé, University of California, Merced
Alexandra Paxton, University of California, Merced
Jessica Ross, University of California, Merced
Emily Sumner, University of Rochester
Kejkaew Thanasuan, Michigan Technological University
Brandon Thomas, University of Cincinnati
Alfredo Vernazzani, Rheinische Friedrich-Wilhelms-Universität Bonn
Yunan "Charles" Wu, Wabash College

xlv
Registration and Information Desk
Pre-registration badge pickup is in the main level lobby, on the left-hand side of
the main entrance outside the Ballrooms. On-site registration is located in Room
CC 204. Certificates of Attendance can be picked up in Room CC 204.

Registration Badge Pick Up


Wednesday 08:00 – 12:00, 13:00 – 14:00 08:00 – 17:00
Thursday 08:00 – 12:30, 13:30 – 17:00 08:00 – 17:00
Friday 08:00 – 12:30, 14:00 – 16:00 08:00 – 17:00
Saturday 08:30 – 10:30 08:00 – 12:30

Coffee Breaks
Coffee and tea breaks are held in the foyer outside Ballroom C-F. Coffee and
tea are complimentary. Juice, soda, and other beverages can be purchased.

Convention Center Map

Rumelhart&&
Recep;on&
!

Poster&Sessions&

Conference&
Center&(CC)&
Badge&Pickup&
Lobby,&Street&Level&
Coffee&Breaks&
Onsite&Registra;on&
Room&204,&Street&Level&

Conference& Paper&Sessions&will&be&held&on&
Center&(CC)& the&Lower&Level&(Rooms&101H107)&

xlvi
Internet Access
1. Connect to the PCOC_visitor wireless network.
2. Once connected, open your web browser.
3. You will be redirect to the splash page, where you will enter your passcode.
4. Enter passcode cogsci2015 (case sensitive) and select “agree to terms”.
5. You will be connected and redirected back to the web.

Cognitive Science Society Business Meeting


The Cognitive Science Society Business Meeting is scheduled for Thursday, 23 Ju-
ly at 18:00 in Ballroom F.

In Case of Emergency
Call 911

xlvii
Proceedings
The proceedings will be available at:

http://mindmodeling.org/cogsci2015/

The proceedings for past year’s conferences are available at:

http://cognitivesciencesociety.org/conference_past.html

Presentation Instructions
Oral presentations: Each standard oral presentation is allocated 20 minutes,
which is to include a closing period for addressing questions from attendees. A
recommended partitioning of this time is 15 minutes of presentation followed by
5 minutes of questions. Each room is equipped with a video projector with
standard VGA input, as well as a microphone. The projectors support 4:3 aspect
ratio – for best results, please ensure your presentations use 4:3 aspect ratio (ra-
ther than any wide-screen format). All speakers must bring their own laptops.
Mac/Apple users must also bring an Apple-VGA connector cable.

Note: Presentations that are part of workshops, tutorials, or symposia may use
different schedules. Please contact the organizer of the given workshop, tutorial,
or symposium for information about the correct amount of presentation time.

Session chairs: This year, the last speaker of each session will serve as the session
chair. That individual will quickly announce each speaker and strictly monitor
the time of each talk. Each room will have signs that the chair can use to signal
to the speaker how much time is left.

Poster presentations: Posters will be presented on poster boards during the post-
er sessions. Poster boards and push-pins will be provided. Maximum poster di-
mensions are 8 feet wide and 4 feet high (244 cm x 122 cm), though of course
they can be smaller. Presenters must remove their posters at the end of the
poster session.

xlviii
How to Cite Your Paper
APA formatted citation for a 6-page paper:
Author A. & Author B.. (2015). This is the title of the paper. In D. C. Noelle, R. Dale,
A. S. Warlaumont, J. Yoshimi, T. Matlock, C. D. Jennings, & P. P. Maglio
(Eds.), Proceedings of the 37th Annual Conference of the Cognitive Science So-
ciety (pp. PAGES). Austin, TX: Cognitive Science Society.

APA formatted citation for a published abstract:


Author A. & Author B. (2015). This is the title of the abstract
[Abstract]. In D. C. Noelle, R. Dale, A. S. Warlaumont, J. Yoshimi, T. Matlock, C. D.
Jennings, & P. P. Maglio (Eds.), Proceedings of the 37th Annual Conference of
the Cognitive Science Society (p. NUMBER). Austin TX: Cognitive Science Society.

APA formatted citation for a talk (or poster) presentation:


Author A. & Author B. (2015, July). This is the title of the talk or
poster. Paper (or Poster) presented at the 37th Annual Conference of the
Cognitive Science Society. Pasadena, California USA.

Sponsors
The Robert J. Glushko and Pamela Samuelson Foundation
IBM
The Mind Science Foundation
University of California, Berkeley
University of California, Merced
University of Southern California

Thank you again for your support!

xlix
The MIT Press

CONSCIOUSNESS, A NATURAL HISTORY OF SPACE IN MIND THE MORAL BRAIN UNIFYING THE MIND
ATTENTION, AND NATURAL THEOLOGY Concepts for Spatial Learning A Multidisciplinary Perspective Cognitive Representations
CONSCIOUS ATTENTION The Cognitive Science and Education edited by Jean Decety and as Graphical Models
Carlos Montemayor and of Theology and Philosophy edited by Daniel R. Montello, Thalia Wheatley David Danks
Harry Haroutioun Haladjian of Religion Karl Grossner, An overview of the latest inter- “Few philosophers address ques-
A rigorous analysis of current Helen De Cruz and Donald G. Janelle disciplinary research on human tions of interest to working sci-
empirical and theoretical work and Johan De Smedt Leading researchers offer a range morality, capturing moral sensibil- entists. David Danks is one of the
supporting the argument that “Why is religion culturally of disciplinary perspectives on ity as a sophisticated integration few. His ideas about conceptual-
consciousness and attention universal? Why do our senses of the implications of spatial think- of cognitive, emotional, and izing cognitive representations as
are largely dissociated. order, design, and beauty lead us ing and reasoning for education motivational mechanisms. graphical models have profound
292 pp., 10 illus., $40 cloth to infer a Designer? De Cruz and and learning. 328 pp., 8 illus., $35 cloth implications for all mind-brain
De Smedt lucidly and seamlessly 352 pp., 36 illus., $45 cloth investigators…”
join philosophy with cognitive —John T. Bruer, President,
PROCESSING INACCURATE science to provide accessible, THE CONCEPTUAL MIND
INFORMATION James S. McDonnell Foundation
empirically based answers. ARTIFICIAL COGNITIVE New Directions in the Study
Theoretical and Applied SYSTEMS 304 pp., 24 illus., $40 cloth
Their huge achievement greatly of Concepts
Perspectives from Cognitive advances religious studies.” A Primer edited by Eric Margolis and
Science and the Educational —Stewart Elliott Guthrie, Pro- David Vernon Stephen Laurence COMPUTER GAMES FOR
Sciences fessor Emeritus of Anthropology, LEARNING
A concise introduction to a New essays by leading philoso-
edited by David N. Rapp Fordham University; author of complex field, bringing together phers and cognitive scientists An Evidence-Based Approach
and Jason L. G. Braasch Faces in the Clouds: A New Theory recent work in cognitive science that present recent findings and Richard E. Mayer
Interdisciplinary approaches to of Religion and cognitive robotics to offer a theoretical developments in the A comprehensive and up-to-
identifying, understanding, and 264 pp., 1 illus., $40 cloth solid grounding on key issues. study of concepts. date investigation of what
remediating people’s reliance on research shows about the
288 pp., 27 illus., $40 cloth 640 pp., 11 color illus., 35 b&w illus., $58 cloth
inaccurate information that they educational value of computer
should know to be wrong. AFTER PHRENOLOGY
games for learning.
480 pp., 23 illus., $50 cloth How to Study the Brain MINDS WITHOUT MEANING CATEGORIZING COGNITION
296 pp., 29 illus., $35 cloth
Michael L. Anderson An Essay on the Content Toward Conceptual Coherence
A proposal for a fully post- of Concepts in the Foundations of Psychology
phrenological neuroscience Jerry A. Fodor Graeme S. Halford,
that details the evolutionary and Zenon W. Pylyshyn William H. Wilson,
roots of functional diversity in
brain regions and networks.
Two prominent thinkers argue
for the possibility of a theory of
Glenda Andrews, and
Steven Phillips Visit our
A Bradford Book
432 pp., 7 color illus., 26 b&w illus., $45 cloth
concepts that takes reference
to be concepts’ sole semantic
A proposal for a categorization of
cognition based on core proper- BOOTH
property. ties of the constituent processes
that integrates theory and empiri-
for a 30%
216 pp., 13 illus., $30 cloth
cal findings across domains. DISCOUNT
The MIT Press mitpress.mit.edu 352 pp., 46 illus., $45 cloth

l
Exhibitors
Oxford University Press
Featuring two new Oxford Handbooks in the Oxford Library of Psychology Series:
Busemeyer et al., Oxford Handbook of Computational and Mathematical Psy-
chology and Calvo et al., Oxford Handbook of Affective Computing. Also new
in Paperback: Eliasmith's How To Build a Brain in the Oxford Series in Cognitive
Models and Architectures and Kolko's Exposing the Magic of Design in the HTI
Series. And much more from Oxford University Press.

The MIT Press


The MIT Press publishes extensively in the field. Please visit our booth to browse
our new and classic titles. 30% discount

Advertisers
Elsevier
We are proud to play an integral part within the Cognitive Science community
by delivering first class information and innovative tools, to meet the needs of
educators, researchers and students worldwide.

The MIT Press

The MIT Press publishes extensively in the field. Please visit our booth to browse
our new and classic titles. 30% discount

Wiley
Wiley is a global provider of knowledge and knowledge-enabled services that
improve outcomes in areas of research, professional practice, and education.
Visit www.wiley.com for more information.

li
MITCogNet
From the MIT Press

Authoritative and unrivaled, MIT CogNet is the essential research tool


for scholars in the Brain and Cognitive Sciences. Since its inception in
2000, MIT CogNet has become an essential resource for those interested
in cutting-edge primary research across the range of fields that study the
nature of the human mind.

2015 MARKS THE LAUNCH OF A NEW, ENHANCED MIT COGNET.

FEATURES INCLUDE:

r.PSFUIBO.*51SFTTCPPLT .*51SFTTSFGFSFODFXPSLT 
BOE.*51SFTT+PVSOBMT XJUIOFXDPOUFOUBEEFESFHVMBSMZ

r%3.GSFFCPPL SFGFSFODF BOEKPVSOBMDPOUFOU

r"OFX DMFBO NPEFSOEFTJHO

r*OUFHSBUFECPPL SFGFSFODF BOEKPVSOBMTFBSDI

r/FX.*5&$4 The MIT Encyclopedia of the Cognitive


Sciences
CBTFEUBYPOPNZGPSCFUUFSCSPXTJOH

A FULLY-FUNCTIONAL r/FXTUBUFPGUIFBSUTFSWFSBOE%SVQBMCBTFEQMBUGPSN
VERSION OF
MIT COGNET IS AVAILABLE GPSHSFBUMZJNQSPWFEQFSGPSNBODFBOETUBCJMJUZ
FOR A FREE, 60-DAY TRIAL
FOR ANY INSTITUTION.

Contact us at
cognet@mit.edu
“MIT CogNet is an invaluable resource for all cognitive science research.
It provides one-stop access to the best in both reference material and
cutting-edge research, and is my starting point for every new project.”

— David Danks, Professor of Philosophy & Psychology,


Carnegie Mellon University

cognet.mit.edu

lii
Robert J. Glushko Dissertation Prizes
The Cognitive Science Society and the Glushko-Samuelson Foundation awards
up to five outstanding dissertation prizes in cognitive science each year. The
goals of these prizes are to increase the prominence of cognitive science and
encourage students to engage in interdisciplinary efforts to understand minds
and intelligent systems. The hope is that the prizes will recognize and honor
young researchers conducting ground-breaking research in cognitive science.
The eventual goal is to aid in efforts to bridge between the areas of cognitive
science and create theories of general interest to the multiple fields concerned
with scientifically understanding the nature of minds and intelligent systems.
Promoting a unified cognitive science is consistent with the belief that under-
standing how minds work will require the synthesis of many different empirical
methods, formal tools, and analytic theories. 2011 was the inaugural year of this
prize, and a new competition is held annually.

The 2015 recipients of the Robert J. Glushko Prizes for Outstanding Doctoral Dis-
sertations / Theses in Cognitive Science are listed below.

Dr. Harm Brouwer, The Electrophysiology of Language Comprehension: A Neu-


rocomputational Model, PhD 2014, University of Groningen

Dr. Da Cheong (Jena) Hwang, Identification and Representation of Caused Mo-


tion Constructions, PhD 2014, University of Colorado

Dr. Brenden Lake, Towards more human-like concept learning in machines:


Compositionality, causality, and learning-to-learn, PhD 2014, Massachusetts Insti-
tute of Technology

Dr. Jessica Sullivan, The Roles of Inference and Associative Learning in the Con-
struction of Mappings Between Number Words and Numerical Magnitudes, PhD
2014, University of California, San Diego

For more information about awardees and their dissertations, see


http://www.cognitivesciencesociety.org/about_awards_glushko_recipients.html

The Glushko Dissertation Prize Symposium showcases the award winning PhD re-
search projects, moderated by Linda B. Smith (Indiana University).
Thursday, 23 July, 15:00

liii
Paper Awards
Marr Prize

The Marr Prize, named in honor of the late David Marr, is awarded to the best
student paper at the conference. All student first authors were eligible for the
Marr Prize for the best student paper. The Marr Prize includes an honorarium of
$1000 and is sponsored by The Cognitive Science Society. The winner of the 2015
Marr Prize for the Best Student Paper is

Tiffany Doan, Stephanie Denison, Christopher Lucas, and Alison Gopnik: Learning
to reason about desires: An infant training study
Friday, 24 July, 10:30

Computational Modeling Prizes

Four prizes worth $1000 each are awarded for the best full paper submissions to
CogSci 2015 that involve computational cognitive modeling. The four prizes rep-
resent the best modeling work in the areas of perception/action language,
higher-level cognition, and applied cognition. These prizes are sponsored by The
Cognitive Science Society. The winners of the 2015 Computational Modeling
Prizes are listed below.

Perception and Action: Judith E. Fan, Daniel L. K. Yamins and Nicholas B. Turk-
Browne: Common object representations for visual recognition and production
Thursday, 23 July, 10:30

Language: Yang Xu, Terry Regier and Barbara Malt: Semantic chaining and effi-
cient communication: The case of container names
Saturday, 25 July, 10:30

Higher-Level Cognition: Nisheeth Srivastava and Ed Vul: Attention dynamics in


multiple object tracking
Friday, 24 July, 15:00

Applied Cognition: Peter Krafft, Robert X.D. Hawkins, Alex Pentland, Noah
Goodman and Josh Tenenbaum: Emergent Collective Sensing in Human Groups
Thursday, 23 July, 10;30

liv
Student Travel Awards
The Robert J. Glushko and Pamela Samuelson Foundation generously sponsored
$10,000 for student travel awards. Travel awards have been provided to students
whose submissions were accepted as full papers, received high rankings, and
who indicated a need for travel funding. This year’s travel awards went to the
individuals shown below.

Nicole Beckage, University of Colorado, Boulder


Neil Bramley, University College London
Gregory Cox, Syracuse University
Tiffany Doan, University of Waterloo
Richard Fereday, Cardiff University
Jessica Hamrick, University of California, Berkeley
Rose Hendricks, University of California, San Diego
Julian Jara-Ettinger, Massachusetts Institute of Technology
Max Kleiman-Weiner, Massachusetts Institute of Technology
Peter Krafft, Massachusetts Institute of Technology
Saebyul Lee, The Ohio State University
Timothy Lew, University of California, San Diego
Tze Kwan Li, University of Hong Kong
Stewart McCauley, Cornell University
Francis Mollica, University of Rochester
Amanda Pogue, University of Rochester
Alexander Rich, New York University
Joshua Rule, Massachusetts Institute of Technology
Eric Schulz, University College London
Janelle Szary, University of California, Merced

lv
Invited Presentations

Rumelhart Prize Lecture and Symposium

Explorations in Representation
Michael Jordan, UC Berkeley
Friday, 24 July, 17:00

The Symposium in Honor of Michael Jordan will be moderated by Tom Griffiths


(UC Berkeley) and include panelists David Blei (Columbia University), Emily Fox
(University of Washington), Geoff Hinton (Google/University of Toronto), Robbie
Jacobs (Rochester), and Josh Tenenbaum (MIT).
Friday, 24 July, 15:00

Heineken Prize Lecture

Toward a Parallel Distributed Processing Approach to Mathematical Cognition


James McClelland, Stanford University
Thursday, 23 July, 17:00

Keynotes

Embodied Decision Making: System Interactions in Movement Execution and Ac-


tion Selection
Richard Ivry, UC Berkeley
Thursday, 23 July, 09:00

Surprising Findings from Measuring Emotion in Real Life


Rosalind Picard, MIT Media Lab
Friday, 24 July, 09:00

Neuroscience and Society: Past, Present and Future


Martha Farah, University of Pennsylvania
Saturday, 25 July, 09:00

lvi
Invited Symposia

Cognitive Science in Society


Organizer: Teenie Matlock, UC Merced
Panelists:
Michael Ranney, University of California, Berkeley
Daniel M. Russell, Google
Daniel C. Richardson, University College London
Thursday, 23 July, 10:30

Human-Machine Symbiosis 50 Years Later


Organizers: Paul Maglio and David C. Noelle, UC Merced
Panelists:
Mike Van Lent, SoarTech
Leila Takayama, GoogleX
David Woods, The Ohio State University
Friday, 24 July, 10:30

Contemporary Perspectives on Sensory Perception and Multimodality


Organizers: Carolyn Jennings and Jeff Yoshimi, UC Merced
Panelists:
Clare Batty, University of Kentucky
Casey O'Callaghan, Washington University in St. Louis
Ladan Shams, UCLA
Saturday, 25 July, 10:30

lvii
Applying for National Science Foundation Funding in Cognitive Science:
Cognition, Computation, Development, Education, and Neuroscience
Anne Cleary1(acleary@nsf.gov), Hector Avila-Munoz2(hmunoz@nsf.gov), Evan
Heit (ekheit@nsf.gov), Chris Hoadley2,3(choadley@nsf.gov), Laura Namy1(lnamy@nsf.gov), Alumit
3

Ishai1(aishai@nsf.gov), Betty Tuller1*(btuller@nsf.gov)


1
Division of Behavioral and Cognitive Sciences, Directorate for Social, Behavioral & Economic Sciences
2
Division of Information and Intelligent Systems, Directorate for Computer & Information Science & Engineering
3
Division of Research on Learning, Directorate for Education & Human Resources
National Science Foundation
4201 Wilson Boulevard
Arlington, Virginia 22230 USA
*Corresponding author

Keywords: cognitive science; research funding; grants; those most relevant to cognitive science, as well career
workshop; National Science Foundation. development programs and international opportunities.
3. Attendees will participate in simulated grant review
Objectives and Background panels, where they will read actual cognitive science grant
This half-day workshop will provide information and hands- proposals and evaluate them in terms of NSF review
on experience related to applying for National Science criteria.
Foundation (NSF) funding in cognitive science. Program 4. Finally, there will be an Ask a Program Officer
officers will discuss the NSF review process and NSF merit session, for questions.
criteria. Details regarding a range of cognitive science
research programs will be covered, including cognition, Table 1. Individual Programs
computation, development, education, and neuroscience. In Cognition
addition, as an interactive activity, attendees will participate Perception, Action & Cognition (PAC)
in simulated review panels using actual cognitive science
grant proposals. Computation
It is expected that attendees will increase their knowledge Information & Intelligent Systems (IIS)
of opportunities at NSF in support of cognitive science, for
example a psychology researcher may learn more about Development
education research opportunities and a computational Developmental and Learning Sciences (DLS)
researcher may learn more about neuroscience research
opportunities. In addition, attendees will increase their Education
understanding of how grant proposals are reviewed and (see Common Guidelines for Education Research)
funding decisions are made at NSF. EHR Core Research (ECR)
The target audience of this workshop is anyone who Discovery Research K-12 (DRK-12)
intends to seek funding for cognitive science research, Improving Undergraduate STEM Education (IUSE)
including graduate students, postdocs, new faculty, and Educational and Human Resources Directorate
experienced researchers. This workshop is an outreach
event for NSF, which seeks broad participation in research
Neuroscience
and strives to fund excellent research in cognitive science
Cognitive Neuroscience
and related fields.
Table 2. Cross-directorate and NSF-wide initiatives
Outline of the Workshop
CAREER (Faculty early career development)
The workshop will have four parts. Computational Cognition
1. General information about applying for NSF funding Cyberlearning
will be covered, such as eligibility, parts of a proposal,
         Research Experiences for Undergraduates (REU)
budget issues, and the merit criteria for review.
2. Program-specific information will be provided, Understanding the Brain
including areas of emphasis and program-specific  
considerations. See Table 1 for a list of individual programs
and Table 2 for a list of cross-directorate or NSF-wide
initiatives. The focus will be on the newest initiatives and

  45
1
Presenters
NSF program officers involved in cognitive science-related
programs, a subset of the workshop co-authors, will lead the
workshop. Follow-up opportunities to meet program
officers may occur during the conference itself.

  46
2
The Workshop of “Physical and Social Scene Understanding”
Tao Gao (taogao@mit.eud)
Department of Brain and Cognitive Sciences, MIT
Cambridge MA, 02139 USA

Yibiao Zhao (ybzhao@ucla.edu)


The Center for Vision, Cognition, Learning, and Art, UCLA
Los Angeles CA, 90095 USA

Lap-Fai (Craig) Yu (craigyu@cs.umb.edu)


Computer Graphics Lab, University of Massachusetts
Boston MA, 02125 USA

The introduction of FPIC will advance cognitive models


Keywords: Causality; Physics; Functionality; Intentionality
in three aspects: (a) transfer learning. As higher-level
representation, FPIC tends to be globally invariant across
the entire human living space. Therefore, learning in one
Theme type of scenes can be transferred to novel situations; (b)
Computer vision has made significant progress in locating small sample learning. Leaning of FPIC, which is consistent
and recognizing objects in real images. However, beyond and noise-free, is possible even without a wealth of previous
the scope of this “what is where” challenge, it lacks the experience or “big data”; and (c) bidirectional inference.
abilities to understand scenes characterizing human visual Inference with FPIC requires the combination of top-down
experience. The mission of this workshop is to (a) identify abstract knowledge and bottom-up visual patterns. The
the key domains in which human visual perception and bidirectional processes can boost the performance of each
cognition outperform computer vision; (b) formalize the other as a result.
computational challenges in these domains; and (c) provide Several key themes are:
promising frameworks for solving these challenges by Physically grounded scene interpretation
conducting cognitive science and computer vision studies. Causal model of vision and cognition
We propose FPIC as four key domains beyond “what is Reasoning about goals and intents of agents in scenes
where”: Human-object-scene interaction
Functionality (e.g., what can be done with this slotted Top-down and Bottom-up inference algorithms
spoon?) In conjunction with CogSci 2015, our “Physical and Social
Physics (e.g., will the spoon be able to pick up the Scene Understanding” workshop will bring together
meatball?) researchers from cognitive science, computer vision and
Intentionality (e.g., is the person trying to scoop up the robotics, to illuminate cognitively-motivated vision systems
cheese or point toward it?) going beyond labeling “what is where” in an image. These
Causality (e.g., why does the gravy pass through the systems work closely together to achieve a sophisticated and
spoon?) coherent understanding of scenes with respect to
The combination of these largely orthogonal dimensions can Functionality, Physics, Intentionality and Causality (FPIC).
span a large space for scene understanding. Despite their In effect, these systems are expected to answer an almost
apparent differences, these domains do connect with each limitless range of questions about an image using a finite
other in ways that are theoretically important: (a) they and general-purpose model. In the meanwhile, we also want
usually don’t project onto explicit visual features; (b) to highlight that FPIC is never meant to be an exclusive set
existing computer vision algorithms are neither competent of scene understanding problems. We welcome the insights
in these domains, nor (in most cases) applicable at all; and of scholars who share the same perspective but are working
(c) human cognition is nevertheless highly efficient at these on different problems.
domains. Therefore, studying FPIC should significantly fill
the gap between computer vision and human vision. On the Speakers
one hand, human studies on FPIC-related topics can inspire We will invite speakers working in cognitive science,
the invention of novel, cognitively-motivated computer computer vision, computer graphics and robotics , who have
vision systems. On the other hand, state-of-the-art computer fundamental insights of visual understanding. We plan to
vision systems can expand the scope of cognitive sciences choose eight speakers from the list below, but are not
to address challenges in real scenes. limited to.

3
Confirmed Speakers: workshop was highly successful and very well-received,
Phillip Wolff (Professor, Cognitive Science, Emory) attracting around 300 audience. This indicates the
Intuitive Physics and Causality enthusiasm towards recent cognitive science studies in the
computer vision community. All talks were video recorded
Adam Sanborn (Professor, Cognitive Science, Warwick)
Intuitive Physics and Causality and posted online at:
http://www.visionmeetscognition.org/fpic2014/
Joshua Tenenbaum (Professor, Cognitive Science, MIT) Here is a brief summary of our previous workshop. We are
Intuitive Physics and Theory of Mind
dedicated to continue the success at CogSci 2015.
Jason Fischer (Post-Doc, Cognitive Neuroscience, MIT) Speakers: We invited 8 keynote speakers from the
Neural circuits of physical and social perception computer vision, cognitive science and computer graphics
Song-Chun Zhu (Professor, Computer Vision, UCLA) fields: Song-Chun Zhu, Katsushi Ikeuchi, William T.
Causal Parsing with Commonsense Reasoning Freeman, Demetri Terzopoulos, Josh Tenenbaum, Ashutosh
Saxena, Benjamin Kuipers, Larry Zitnick. Their talks
Brian Scassellati (Professor, Robotics, Yale)
Social Robotics provided diverse insights from different perspectives which
are all highly relevant to the theme of our workshop: Vision
Brian Ziebart (Professor, Robotics, UIC) meets Cognition.
Purposeful Prediction
Audiences:There were 367 people who signed up for our
Demetri Terzopoulos (Professor, Computer Graphics, UCLA) workshop during the registration. Our keynote talks were
Artificial Life very popular and most of the time our room was fully
seated. At peak time, our workshop attained attendance of
Workshop Program over 200 workshop participants.
Our workshop will be a full day workshop hosting talks of Accepted papers: There were 38 carefully peer-reviewed
eight invited speakers who are leading researchers in their papers accepted by our workshop, which involved more
research fields. than 200 paper authors. Each paper was reviewed by 2-4
 Each speaker will have 35 minutes to experts in the field, chosen among our 30 program
present. committee. We broadcasted a trailer video composed of
 One-hour panel discussion at the end. spotlight slides to promote all of our accepted papers.
 All talks and discussions will be video recorded Sponsors: Our workshop has also aroused significant
and posted on the workshop website. industry interest, and was generously supported by 4
Tentative Schedule: sponsors: Microsoft Research, A9, Vicarious and the Office
9:00am - 9:10am Welcome speech for Naval Research.
9:15am - 9:45am Invited talk 1 Souvenirs: All of our invited speakers received our custom-
9:55am - 10:25am Invited talk 2 designed souvenirs in recognition of their contribution. We
10:30pm - 11:00 am Invited talk 3 also designed and manufactured 200 magic mugs for all our
11:05pm - 11:35 am Invited talk 4 workshop attendants and guests.
11:40am - 1:00pm Lunch Break Based on our organizing experience, we are highly
1:00pm - 1:30pm Invited talk 5 confident that the workshop of “Physical and Social Scene
1:35pm - 2:05pm Invited talk 6 Understanding” at CogSci 2015 can achieve an even bigger
2:10pm - 2:40pm Invited talk 7 success.
2:45pm - 3:15pm Invited talk 8
3:20pm - 4:20pm Panel Discussion

Potential Financial Support


We are currently looking for sponsorship from the Center of
Brian, Mind and Machine (CBMM) at MIT. We are
planning to get support from our sponsors (e.g. Microsoft
Research, A9) of our previous workshop again. The
sponsorship will be used for covering the travel fees of
some of our invited speakers; making souvenirs for our
contributors and workshop participants; recording a video
for all of our talks. With this support, we aim at organizing a
workshop that every participant enjoys, and further boosting
the impact of our workshop and the CogSci conference.

Success of Our Previous Workshop


We held a related workshop on “Vision meets Cognition” at
CVPR 2014 (a premiere computer vision conference). The

4
Language & Common Sense
Integrating across psychology, linguistics, and computer science
!
Joshua K. Hartshorne (jharts@mit.edu) & Joshua B. Tenenbaum (jbt@mit.edu)
Department of Brain & Cognitive Sciences, 77 Massachusetts Avenue
Cambridge, MA 02139 USA
!!
Keywords: language; common sense; world knowledge; online, domain-general reasoning, or do we have language-
pragmatics; intuitive theories; cognitive architectures; natural specific strategies and representations? To the degree that
language processing online reasoning is involved, how can it be characterized
(e.g., as rote heuristics, as inference over intuitive theories,
Introduction as distributional probabilities, etc.)?
The language understanding that underlies state-of-the-art Related issues arise in relating language to perception. We
Internet search, machine translation, and dictation software talk about what we see, but again the information is highly
is undeniably impressive. Equally undeniable is that these compressed: The elephant was on the truck is consistent
systems do not really understand language. What is with a staggering range of visual input. How do the
missing? One candidate is common sense. Language is a semantic representations of language relate to the mid- or
mechanism for moving ideas from one mind to another – high-level representations used in perception? What role
ideas that are meant to be understood in the context of pre- does knowledge of the world play (elephants are more likely
existing, interlocking beliefs. Thus, fully exploiting its to be on certain parts of certain trucks)?
power may require sophisticated, explicit representations of The different cognitive sciences have approached
world knowledge – that is, common sense. different aspects of these questions in different ways. This
That understanding language requires deploying workshop brings together ten researchers from across the
knowledge about the world is not a new observation (cf. cognitive sciences to disseminate and discuss successes (and
Winograd, 1972). However, new opportunities for challenges), and also to help formulate an agenda for the
significant progress have been suddenly opened up by field: What phenomena and challenge problems should be
recent, rapid advances in the science of common sense, explored? How can progress in formalizing pragmatics and
along with related advances in machine vision, natural semantics be used to formalize theories of common sense
language processing, and computational tools for (and vice versa)? Which questions raised in linguistics may
developing more precise cognitive theories (Liang & Potts, have answers in psychology or computer science (and vice
2015; Sonka, Hlavac, & Boyle, 2014; Tenenbaum, Kemp, versa)?
Griffiths, & Goodman, 2011).
This workshop brings together researchers from across Winograd Schema
the cognitive sciences – including developmental and In addition to presentations covering a range of topics at the
cognitive psychology, linguistics, natural language intersection of language and common sense, the workshop
processing, artificial intelligence, and robotics – to will contain a special session on Winograd Schema, so
disseminate recent findings, discuss approaches to future named for Winograd’s (1972) classic demonstration of
progress, and set an agenda for the field. Given the commonsense reasoning’s influence on pronoun
interdisciplinary nature of the participants and of the interpretation:
research challenges faced, the Annual Meeting of the (1) The city council denied the protesters a permit
Cognitive Science Society is an ideal venue for these because they advocated violence.
conversations. (2) The city council denied the protesters a permit
because they feared violence.
Goals and Scope Most readers agree that the pronoun they refers to the
protesters in (1) but the city council in (2). This seems to
Language, as a vehicle for conveying thoughts and beliefs, derive from our understanding of city councils, protesters,
is highly compressed. In many cases, resolving the resulting and the permitting process: City councils rarely advocate
ambiguity seems to require knowledge of the world. Our violence, and even if they did, that would be poor reason for
knowledge of summertime activities suggests that the bank them to deny protesters permits.
in Sally dove off the bank into the river is probably not a Winograd Schema like (1-2) are of growing interest in
financial institution. Our knowledge of current market artificial intelligence, where it has been suggested that they
prices implies that these pencils cost $100 is probably an provide a sophisticated, alternative Turing Test, given that
exaggeration for emotional effect (Kao, Wu, Bergen, & they involve both language and common sense (Levesque,
Goodman, 2014). That the elephant on the wall probably Davis, & Morgenstern, 2012). They also provide an enticing
describes a hanging picture whereas the elephant on the
opportunity for cross-disciplinary dialog: There is a deep,
truck describes a translocating pachyderm may be inferred robust literature on such phenomena in both psychology and
from our knowledge of physics and zoology. linguistics, including (recently) computational models. In
These observations leave open how world knowledge is keeping with the overall theme of the workshop, this session
incorporated in language understanding. Do we make use of

5
will consist of presentations by a computer scientist Andrew Kehler
(Charles Ortiz), a linguist (Andrew Kehler), and a Department of Linguistics
psychologist (Joshua Hartshorne). University of California-San Diego
!
Percy Liang
Workshop Organization
Department of Computer Science
The workshop will be organized around a set of thirty- Stanford University
minute presentations (including Q&A) and panel
discussions. The presentations will range from theoretical
!
Charles L. Ortiz, Jr.
overviews to detailed discussion of specific phenomena. Natural Language and Artificial Intelligence Laboratory
The panel discussions and coffee breaks will help spur Nuance Communications
discussion about promising avenues for future research and
help build a common vocabulary and agenda.
!
Jeffrey M. Siskind
School of Electrical and Computer Engineering
Workshop Organizers Purdue University
Joshua K. Hartshorne is a Ruth L. Kirschstein NRSA
post-doctoral fellow in the Department of Brain and
!
Stefanie Tellex
Cognitive Sciences at MIT and an incoming assistant Computer Science Department
professor at in the Department of Psychology at Boston Brown University
College. His research focuses on how cognitive
representations constrain and inform language, both in
!
Joshua B. Tenenbaum
online processing (Hartshorne, O’Donnell, & Tenenbaum, in Department of Brain & Cognitive Sciences
press) and during development (Hartshorne, Pogue, & Massachusetts Institute of Technology
Snedeker, in press). Joshua B. Tenenbaum is Professor of
Cognitive Science and Computation at MIT. His recent
work focuses on computational models of commonsense
References
reasoning and intuitive theories (Tenenbaum et al., 2011; Battaglia, P. W., Hamrick, J. B., & Tenenbaum, J. B. (2013).
Battaglia, Hamrick, & Tenenbaum, 2013). Simulation as an engine of physical scene understanding.
PNAS, 110, 18327-332.
Target Audience Hartshorne, J. K., O’Donnell, T. J., & Tenenbaum, J. B. (in
press). The causes and consequences explicit in verbs.
The target audience for this workshop overlaps significantly Language, Cognition, and Neuroscience.
with the target audience of CogSci. The workshop Hartshorne, J. K., Pogue, A., & Snedeker, J. (in press). Love
incorporates two themes (language and commonsense is hard to understand: The relationship between
reasoning) that are central across many of the cognitive transitivity and caused events in the acquisition of
science disciplines (artificial intelligence, linguistics, emotion verbs. Journal of Child Language.
psychology, etc.). Moreover, the workshop approaches these Kao, J. T., Wu, J.Y., Bergen, L., & Goodman, N. D. (2015).
themes from a multidisciplinary perspective, as seen in the Nonliteral understanding of number words. PNAS, 111,
disciplinary diversity of the participants. Because the 12002-07.
presentations will be geared towards an interdisciplinary Liang, P., & Potts, C. (2015). Bringing machine learning
audience, they should be approachable by a broad cognitive and compositional semantics together. Annual Review of
science audience. Linguistics, 1, 355-376.
Levesque, H. J., Davis, E., & Morgenstern, L. (2012). The
Confirmed Speakers Wi n o g r a d S c h e m a c h a l l e n g e . P ro c e e d i n g s o f
David Barner Commonsense Reasoning 2012.
Department of Psychology Sonka, M., Hlavac, V., & Boyle, R. (2014). Image
University of California-San Diego processing, analysis, and machine vision, 4th Edition,
!
Nancy Chang
Stamford, CT: Cengage Learning.
Tenenbaum, J. B., Kemp, C., Griffiths, T. L., & Goodman,
Google N. D. (2011). How to grown a mind: Statistics, structure,
!
Noah D. Goodman
and abstraction. Science, 331, 1279-1285.
Winograd, T. (1972). Understanding natural language.
Department of Psychology Cognitive Psychology, 1-191.
Stanford University
!
Joshua K. Hartshorne
Department of Brain & Cognitive Sciences
Massachusetts Institute of Technology
!!

6
Evidence Accumulation Modeling:
Bayesian Estimation using Differential Evolution
Andrew Heathcote (andrew.heathcote@utas.edu.au)
School of Medicine, The University of Tasmania
Social Sciences Building, Sandy Bay, 7005, Tasmania, Australia
Brandon Turner (turner.826@osu.edu)
Department of Psychology, The Ohio State University, Columbus, OH 43227 USA
Scott D. Brown (scott.brown@newcastle.edu.au)
School of Psychology, The University of Newcastle,
Psychology Building, University Avenue, Callaghan, 2308, NSW, Australia

Keywords: Bayesian modeling; Mathematical modeling;


by the difference between the observed response time (RT)
Decision making; Response times and decision time (i.e., the time that evidence in the winning
accumulator first equals the threshold). Brown and
Evidence Accumulation Modeling Heathcote (2008) showed that the LBA is able to account
for a comprehensive set of benchmark phenomena in simple
Understanding decision making requires a dynamic choice paradigms, such as speed-accuracy tradeoff (e.g.,
approach that accounts for the time taken to make choices as being more accurate at the cost of longer decision times by
well as the choices that are made. The success of the raising b). The LBA is easily extended to more than two
dynamic approach is underpinned by cognitive models, such choices, whereas the DDM, which consists of a single unit
as the drift-diffusion model (DDM: Ratcliff & McKoon, with two thresholds and extra sources of noise (in the
2008) and linear ballistic accumulator (LBA: Brown & evidence within trials and uniform trial-to-trial variability in
Heathcote, 2005, 2008), that attribute decisions to an non-decision time), only applies to binary choice.
evidence accumulation processes. The ability of these However, applying models like the LBA and DDM to real
models, and elaborations of them, to account for the speed data sets can be challenging for a range of reasons: 1)
and accuracy with which people make decisions across a suitable experimental designs are required with sufficient
broad range of tasks has led to an increasing number of number of observations, experimental control, and
applications in Cognitive Science and Neuroscience (for manipulations that help to identify model parameters, 2)
recent examples see Cassey, Heathcote & Brown, 2014; nonlinear interactions within the models, and only partial
Heathcote, Loft & Remington, in press; Mittner et al., 2014; observation of the accumulation process (i.e., its end point),
Turner, Van Maanen, Forstmann, in press). cause strong correlations among parameters that make them
difficult to estimate, and 3) fitting can be computationally
Threshold (b)
demanding and suffer from problems of numerical
left
instability.
Although excellent estimation packages based on both
right
maximization (Vandekerckhove & Tuerlinckx, 2008; Voss
Start Point
from [0,A]

v) & Voss, 2007) and Bayesian (Vandekerckhove, Tuerlinckx


te(
Ra
Dr
ift & Lee, 2011; Wiecki, Sofer & Frank, 2013) methods are
now available that ease the computational problems for the
Decision Time DDM, they can impose assumptions that do not make them
flexible enough for some applications, particularly in the
Figure 1. LBA Model for a decision between “left” and Bayesian setting. This tutorial provides training in a more
“right” responses. Accumulators (arrows) race and the flexible approach that can be used with the LBA and DDM.
first one to reach threshold determines the choice.
Tutorial Overview
Figure 1 schematically illustrates an LBA model, for a
The tutorial is presented by the developers of the LBA
binary (“left” vs. “right”) choice. Each choice has its own
model, Scott Brown and Andrew Heathcote, along with
accumulator that linearly accrues corresponding evidence
Brandon Turner, who with Brown and colleagues proposed
(illustrated by the arrows in Figure 1), starting from starting
using the Differential Evolution algorithm as a way of
points (uniformly distributed over the interval 0-A) that
dealing with the problem of correlated parameters in the
represent random biases from trial to trial. The rate of
Bayesian context (Turner, Sederberg, Brown, & Steyvers,
accumulation (v), which also varies normally from trial to
2013). It teaches attendees to apply Bayesian estimation
trial, corresponds to the strength of evidence for a choice.
using DE, with a focus on the LBA model, although the
The first accumulator to reach its threshold (b) determines
techniques taught can be applied to any evidence
the choice. The time for non-decision processes (e.g.,
accumulation model for which a likelihood can be
stimulus encoding and response selection, ter) is estimated
computed.

7
Flexibility is obtained by providing attendees with easily distributions. R package version 0.2-6.
modifiable source code in the R language (R Core Team, http://CRAN.R-project.org/package=rtdists
2014). This means that some familiarity with the R language Brown, S. D., & Heathcote, A. (2008). The simplest
is necessary for attendees to best benefit from the tutorial. complete model of choice response time: Linear ballistic
However, much of the tedious bookkeeping necessary to fit accumulation. Cognitive Psychology, 57(3), 153–178.
models, as well as graphing to check the results of sampling, Cassey, P., Heathcote, A., & Brown, S. D. (2014) Brain and
is taken care of by convenience functions that users should Behavior in Decision-Making. PLoS Computational
not have to modify in most standard contexts. Hence, Biology, 10(7), e1003700.
knowledge of R is mainly required to read in data and Heathcote, A. (2014). Flexibility and selective influence in
analyze the outputs of sampling. One the other hand more evidence accumulation models. Keynote paper presented
advanced R users can use these functions as a basis to at the European Mathematical Psychology Group
implement model variants and/or to address non-standard Meeting, July, Tubingen, Germany.
applications. Heathcote, A., Brown, S.D. & Mewhort, D.J.K. (2002).
Likelihood computation uses the newly developed rtdists Quantile Maximum Likelihood Estimation of Response
R package (Brown, Gretton, Heathcote & Singmann, 2014), Time Distributions. Psychonomic Bulletin and Review, 9,
so users do not have to be concerned with the associated 394-401.
mathematical details. This package robustly and efficiently Heathcote, A., Loft, S. & Remington, R. (accepted
implements likelihood computations for the LBA, including 31/12/14). Slow down and remember to remember! A
uniform trial-to-trial variability in non-decision time, and delay theory of prospective memory costs, Psychological
the full DDM model with trial-to-trial variability in starting Review
points, mean rate and non-decision time1. Mittner, M., Boekel, W., Tucker, A., Turner, b., Heathcote,
The tutorial will begin with 30-minute overviews of A. & Forstmann, B. U. (2014). When the brain takes a
Bayesian estimation, the LBA model, and DE sampling break: A model-based analysis of mind wandering,
given, respectively, by Turner, Brown and Heathcote. The Journal of Neuroscience, 34(49),16286-16295.
remainder of the tutorial will consist of hands on exercises R Core Team (2014). R: A language and environment for
using the suite of R functions provided to simulate data and statistical computing. R Foundation for Statistical
run sampling for the LBA model, and to examine and Computing, Vienna, Austria. URL http://www.R-
interpret outputs. The first session up to lunch will focus on project.org/.
estimation for an individual participant in a design with two Ratcliff, R., & McKoon, G. (2008). The diffusion decision
conditions. After lunch the idea of hierarchical modeling model: Theory and data for two-choice decision tasks.
will be introduced (i.e., estimation of group as well as Neural Computation, 20(4), 873–922.
individual level parameters in a design with a within- Turner, B., Sederberg, P., Brown, S.D., & Steyvers, M.
subjects factor. If time permits students will be introduced (2013) A method for efficiently sampling from
to code provided for more complex designs (e.g., both with distributions with correlated dimensions. Psychological
and between-subjects factors). Methods, 18(3), 368-384.
Finally, time will be set aside to discuss with the Turner, B., Van Maanen, L., & Forstmann, B. (in press).
presenters how to approach the applications of the LBA and Combining cognitive abstractions with neurophysiology:
DDM model they are interested in making. Advice will be The Neural Drift Diffusion Model. Psychological Review.
given on the design of appropriate experiments, analyses of Vandekerckhove, J., & Tuerlinckx, F. (2008). Diffusion
existing data and approaches to more advanced designs and Model Analysis with MATLAB: A DMAT Primer.
models. Students are encouraged to bring their own data sets Behavior Research Methods, 40, 61-72.
or proposed designs in order to facilitate discussion. Voss, A., Nagler, M., & Lerche, V. (2013). Diffusion
Models in Experimental Psychology. Experimental
References Psychology; Experimental Psychology, 60(6), 385–402.
Scott Brown, Matthew Gretton, Andrew Heathcote and Voss, A., & Voss, J. (2007). Fast-dm: A Free Program for
Henrik Singmann (2014). rtdists: Response time Efficient Diffusion Model Analysis. Behavioral Research
Methods, 39, 767-775.Voss, A., & Voss, J. (2007). Fast-
dm: A Free Program for Efficient Diffusion Model
1
The likelihood for the DDM with only within-trial variability can Analysis. Behavioral Research Methods, 39, 767-775.
be computed efficiently (Wabersich & Vandekerckhove, 2014), Wiecki, T.V., Sofer, I. & Frank, M.J. (2013). HDDM:
but adding trial-to-trial variability requires numerical integration, Hierarchical Bayesian estimation of the Drift-Diffusion
which can be unstable and slow. Hence, most most previous full- Model in Python. Frontiers in Neuroinformatics, 7:14.
DDM applications have been used quantile-based methods (e.g.,
Wabersich, D., & Vandekerckhove, J. (2014). The RWiener
Heathcote, Brown & Mewhort, 2002) that are not ideal in the
Bayesian context, or have implemented trail-to-trial variability package: an R package providing distribution functions
hierarchically (Vandekerckhove et al., 2013), which can result in for the Wiener diffusion model. The R Journal, 6, 49-56.
an overly flexible model (Heathcote, 2014). The rtdists package is
based on Voss, Nagler and Lerche’s (2013) efficient C code,
making likelihood-based estimation of the full DDM practical.

8
Workshop on Optimizing Experimental Designs:
Theory, Practice, and Applications
Jay I. Myung (myung.1@osu.edu)
Mark A. Pitt (pitt.2@osu.edu)
Department of Psychology,
Ohio State University, Columbus OH 43210 USA

Maarten Speekenbrink (m.speekenbrink@ucl.ac.uk)


Department of Experimental Psychology
University College London, London WC1H 0AP UK

Keywords: cognitive modeling; experimental design;


The purpose of the workshop is to introduce the principles
active learning; adaptive experimentation. underlying OED, illustrate how to apply OED in practice
using widely and freely available software tools (e.g., R) to
Background and Purpose showcase applications of OED in areas such as cognitive
psychology, education and assessment, and machine
The accurate and efficient measurement of observations is learning, and provide a platform to share work on OED.
at the core of empirical scientific research. To ensure
measurement is optimal, and thereby maximize inference,
there has been a recent surge of interest among researchers
Workshop Format
in the design of experiments that lead to rapid accumulation This full-day workshop will be organized around two
of information about the phenomenon under study with the specific goals: (1) to educate the cognitive science
fewest possible measurements. community about optimal experimental design (OED) and
Statisticians have contributed to this area by introducing (2) to bring practitioners together who use it to share and
methods of optimizing experimental design (OED: e.g., showcase their latest work with the community. The first
Atkinson & Donev, 1992; Lindley, 1956), which is related goal will be met in the morning session, which will include
to active learning in machine learning (Cohn, Ghahramani a 75-minute tutorial on the theoretical and computational
& Jordan, 1996) and to computerized adaptive testing in foundations of OED given by Jay Myung and then another
psychometrics (van der Linden & Glass, 2000). The 75-minute hands-on session on the practical and
methodology involves adapting the experimental design in implementation aspects of OED given by Maarten
real time as the experiment progresses. Specifically, in Speekenbrink. The second goal will be met in the afternoon,
OED, an experiment is run as a sequence of stages, or mini- which will consist of six 30-minute invited presentations
experiments, in which the values of design variables (e.g., featuring example applications demonstrating the use of
stimulus properties, task parameters, testing schedule) for OED in various disciplines.
the next stage are chosen based on the information (e.g., There will be a website with a workshop program, the
responses) gathered at earlier stages, so as to be maximally titles and abstracts of all presentations, and recommended
informative about the question of interest (i.e., the goal of readings.
the experiment).
OED has become increasing popular in recent years, Target Audience
largely due to the advent of fast computing, which has made Graduate students, postdoctoral researchers, and scientists,
it possible to solve more complex optimization problems, who are new to OED and have workable knowledge of
and as such is starting to reach everyday experimental statistics on a graduate level. We anticipate that around 40-
scientists. A growing number of labs are applying OED 50 participants would attend the workshop.
across scientific fields, including cognitive psychology
(Myung & Pitt, 2009; Cavagnaro, Myung, Pitt & Kujala, Workshop Organizers
2010), neuroscience (Lewi, Butera & Paninski, 2009), Jay Myung is Professor of Psychology at the Ohio State
psychophysics (Lesmes, Jeon, Lu, & Dosher, 2006), University. He received his PhD in 1990 in psychology at
systems biology (Kreutz & Timmer, 2009), astrophysics Purdue University and spent a year as a postdoc at the
(Loredo, 2004), systems engineering (Allen, Yu & Schmitz, University of Virginia. His research interests in the fields of
2003), and clinical drug trials (Wathen & Thall, 2008). OED cognitive and mathematical psychology include optimal
is not only a useful framework to enhance scientific experimental design, Bayesian inference, model
research, but the underlying principles are also useful as a comparison, and neural networks. Homepage:
framework to understand how intelligent agents actively http://faculty.psy.ohio-state.edu/myung/personal/.
sample information to enhance their learning (e.g., Bramley,
Lagnado & Speekenbrink, 2014; Nelson, 2005).

9
Mark Pitt is Professor of Psychology at the Ohio State Acknowledgments
University. He received his PhD in 1989 in psychology at
This research is supported in part by National Institute of
Yale University, twiddled his thumbs for a while, and then
Health Grant R01-MH093838 to JIM and MAP.
joined the faculty at OSU. His research interests are in
model evaluation, design optimization, and in the field of
language and spoken word recognition. Homepage:
References
http://lpl.psy.ohio-state.edu/. Allen, T., Yu, L., & Schmitz, J. (2003). An experimental
design criterion for minimizing meta-model prediction
Maarten Speekenbrink is Lecturer in Mathematical errors applied to die casting process design. Journal of the
Psychology at University College London. He received his Royal Statistical Society Series C (Applied Statistics), 52,
PhD in 2005 in psychology at the University of Amsterdam, 103–117.
focusing on psychological methodology. Afterwards, he Atkinson, A. and Donev, A. (1992). Optimum Experimental
moved to UCL for a postdoctoral research position, where Designs. Oxford University Press.
he was later appointed as lecturer. His research interests Bramley, N. R., Lagnado, D. A. & Speekenbrink, M.
include optimal experimental design, statistics, (2014). Conservative forgetful scholars: How people learn
computational modeling, learning, and decision making. causal structure through interventions. Journal of
Homepage: http://www.ucl.ac.uk/speekenbrink-lab/. Experimental Psychology: Learning, Memory &
Cognition (published ahead of print, 20 October 2014)
Presenters Cavagnaro, D. R., Myung, J. I, Pitt, M. A., and Kujala, J.
(2010). Adaptive design optimization: A mutual
The following invited speakers have confirmed their
information-based approach to model discrimination in
participation:
cognitive science. Neural Computation, 22(4), 887–905.
Cohn, D., Ghahramani, Z., and Jordan, M. (1996). Active
Daniel Cavagnaro
learning with statistical models. Journal of Artificial
Dept. of Information Systems and Decision Sciences
Intelligence Research, 4,129-145.
California State University Fullerton, USA
Kreutz, C., & Timmer, J. (2009). Systems biology:
experimental design. FEBS Journal, 276, 923–942.
Christopher DiMattina
Lesmes, L., Jeon, S.-T., Lu, Z.-L., & Dosher, B. (2006).
Department of Psychology
Bayesian adaptive estimation of threshold versus contrast
Florida Gulf Coast University, USA
external noise functions: the quick TvC method. Vision
Research, 46, 3160–3176.
Woojae Kim
Lewi, J., Butera, R., & Paninski, L. (2009). Sequential
Department of Psychology
optimal design of neurophysiology experiments. Neural
Ohio State University, USA
Computation, 21, 619–687.
Lindley, D. V. (1956). On a measure of the information
Jay Myung
provided by an experiment. The Annals of Mathematical
Department of Psychology
Statistics, 27, 986-1005.
Ohio State University, USA
Loredo, T. J. (2004). Bayesian adaptive exploration. In G. J.
Erickson, & Y. Zhai (Eds.), Bayesian Inference and
Jonathan Nelson
Maximum Entropy Methods in Science and Engineering:
Center for Adaptive Behavior and Cognition
23rd International Workshop on Bayesian Inference and
Max Planck Inst. for Human Development, GERMANY
Maximum Entropy Methods in Science and Engineering:
Vol. 707 (pp. 330–346). American Institute of Physics.
Anna Rafferty
Myung, J. I. & Pitt, M. A. (2009). Optimal experimental
Department of Computer Science
design for model discrimination. Psychological Review,
Carleton College, USA
116, 499-518.
Nelson, J. D. (2005). Finding useful questions: on Bayesian
Eric Schulz
diagnosticity, probability, impact and information gain.
Department of Experimental Psychology
Psychological Review, 112, 979-999.
University College London, UK
van der Linden, W. J., & Glas, C. A. W. (2000).
Computerized Adaptive Testing. Boston, MA: Kluwer
Maarten Speekenbrink
Academic Publishers.
Department of Experimental Psychology
Wathen, J. K., & Thall, P. F. (2008). Bayesian adaptive
University College London, UK
model selection for optimizing group sequential clinical
trials. Statistics in Medicine, 27, 5586–5604.
Byoung-Tak Zhang
Department of Computer Science & Engineering
Seoul National University, KOREA

10
Quantifying the Dynamics of Interpersonal Interaction:
A Primer on Cross-Recurrence Quantification Analysis using R
Moreno I. Coco (micoco@psicologia.ulisboa.pt)
Faculdade de Psicologia, Universidade de Lisboa
Alameda da Universidade, Lisboa, 1649-013, Portugal

Rick Dale (rdale@ucmerced.edu)


Cognitive and Information Sciences, University of California
Merced, 5200 North Lake Rd., Merced, CA 95343

Keywords: cognitive dynamics; cross-recurrence quantifi- and at different stages of their career (from graduates students
cation analysis; R-tutorial; interpersonal interaction; sta- to senior scientists) will learn a dynamical systems frame-
tistical methods
work to interpret and understand interpersonal interaction,
and acquire the analytical principles of C/RQA, which helps
Objectives and Scope framing this approach.
Humans live in a very interactive context, requiring very fre-
quent exchange of information with con-specifics, which it- Tutorial format and pre-requisites
self requires subtle temporal calibration of linguistic and non-
linguistic activities. Over the last decade, the dynamics of In this half-day tutorial, we aim to achieve two main goals:
interpersonal interaction has become a growing topic in cog- (1) a basic understanding of C/RQA and its applicability
nitive science, precisely because of the important implica- to research in cognitive science, (2) a primer, hands-on, of
tions that interpersonal dynamics carry on shaping our ‘so- C/RQA to behavioral data using the package M. I. Coco and
cial cognitive system.’ Research on this topic has helped us Dale (2014) in R.
understand, for example, how overt behavior such as body We begin the tutorial by illustrating how C/RQA developed
sway and eye movements of interacting individuals converge within the literature in cognitive science, and explain how
or diverge in various ways (Shockley, Santana, & Fowler, this methodology can offer more than traditional approaches
2003; Richardson & Dale, 2005), whether temporal ‘calibra- based on aggregation (M. Coco & Dale, 2014).
tion’ occurs over multiple behavioral scales (Louwerse, Dale, Then, we present the theoretical backbone of C/RQA in
Bard, & Jeuniaux, 2012); as well as, developmental child- more details, and show-cast its applicability to various kinds
caregiver dynamics (Yu & Smith, 2013). of data, mostly relevant to language, from non-verbal behav-
Many important advances on this research topic have been ior such as eye-movements, to transcripts drawn from speech.
possible through the application of concepts and statistical We focus on applications of C/RQA to categorical data, such
methods, which provide quantification for the dynamic struc- as sequences of eye-movement fixation, or lyrics from songs,
ture of cognitive responses observed when individuals inter- and explain how a Recurrence Plot (RP), the core component
act. Recurrence Quantification Analysis (RQA) is one of of C/RQA, is built from them. From the Recurrence Plot,
such framework, and has received growing attention for re- we subsequently describe the different measures that can be
search on interpersonal dynamics. Conceptually, RQA makes computed on it, such as recurrence rate, determinism, etc., as
it possible to quantify how, and the extent to which, a signal well as guiding the participants to interpret the implications
is revisiting a similar state in time (Marwan, Carmen Ro- of such measurement on the understanding of cognitive pro-
mano, Thiel, & Kurths, 2007). When RQA is applied on cesses. Just to give a flavor of what RQA could do, in Figure 1
two different streams of the same information, such as the we show a RP built on an extract of the lyrics from the song
eye-movement trajectories of two interlocutors, it takes the ’Call me maybe’ by Carly Jepsen using the crqa package.
name of Cross-Recurrence Quantification Analysis (C/RQA). The points along the diagonals are the sequence of words in
C/RQA can be used, for example, to examine the temporal the extract, that are repeated in time. In practice, along the
organization of eye-movement trajectories of dyads of inter- diagonals we quantify how much, and for how long, is the
locutors as they complete a communicative task, and estab- system (in this case a sequence of words) synchronizing with
lish their attentional correspondence, their feedback dynam- itself. In C/RQA, we follow the same logic, but instead of
ics (e.g., leader-follower lag), as well as examine how experi- looking at the synchronism of a time-series with itself, we
mental variables might foster or disrupt such synchronism. In want to discover the temporal dynamics of coupling between
a sense, this makes C/RQA a very comprehensive time series two different time-series.
technique for obtaining new descriptive statistics, and some In the last section of the tutorial, we provide the partici-
have referred to it as a sort of generalized nonlinear cross- pants with a practical hands-on, coding session, using R. In
correlation function (Marwan et al., 2007). particular, we will provide participants with worked-out in-
In this tutorial, cognitive scientists from different fields, struction sheets written in Markdown, a marked-up language

11
Table 1: Tutorial Schedule
Time Content
9am - 10am Theory: C/RQA in cognitive science
research.
10am - 11am Application: C/RQA for categorical
time-series.
11am - 12am Hands-on: C/RQA on eye-movement
and speech data.

search aims at quantifying cognitive dynamics during human


communication, focusing on language use and evolution, as
well as its intimate relation with action. In the context of
C/RQA, he has pioneered the use of such method to quantify
cognitive mechanisms of alignment, initially eye-movements,
then extended to a wider range of cognitive responses. He
will mostly cover the theoretical part of this tutorial.
Figure 1: Recurrence Plot for speech extract: before you came Both instructors have given together a week-long postgrad-
into my life I missed you so bad, I missed you so bad I missed uate workshop on C/RQA (29th April - 2nd May, 2014) at the
you so so bad before you came into my life I missed you so School of Philosophy, Psychology and Language Sciences
bad from the song ’Call me maybe’ by Carly Jepsen. (University of Edinburgh).
RR: Percentage of points in the plot = 8.81%; L: Average
length of diagonal lines = 4.87; maxL: Longest diagonal Tutorial Structure
= 11; DET: Percentage of the points on diagonal lines = The participants of this tutorial will be guided through practi-
81.25%; ENTR: Entropy distribution of diagonal lines = 0.92 cal research examples and hands-on activities (including data
analysis) to understand the theoretical principles of CRQA. A
detailed schedule is provided in Table 1.
which makes possible to integrate R-code, its output, and ex-
planation within the same document (using knitR). References
It would be preferable if tutorial participants have al- Coco, M., & Dale, R. (2014). Cross-recurrence quantification
ready basic familiarity with R. However, no specific analysis of behavioral streams: An r package. Quantitative
knowledge of the crqa package is required. Partici- Psychology and Measurement, 5, 510.
pants are also expected to bring their laptops with R in- Coco, M. I., & Dale, R. (2014). crqa: Cross-recurrence
stalled (http://cran.r-project.org/), along with the quantification analysis for categorical and continuous time-
crqa package, so that they can actively participate to the series [Computer software manual]. Retrieved from
hands-on session. We are agnostic as to the editor you should http://CRAN.R-project.org/package=crqa (R pack-
use to compile R. R-Studio (http://www.rstudio.com/) age version 1.0.5)
is becoming a popular choice because it provides a cross- Louwerse, M. M., Dale, R., Bard, E. G., & Jeuniaux, P.
platform environment, and several other features. However, (2012). Behavior matching in multimodal communication
also simple editing softwares, such as EMACS, can work. is synchronized. Cognitive Science, 36(8), 1404–1426.
Marwan, N., Carmen Romano, M., Thiel, M., & Kurths, J.
Instructors
(2007). Recurrence plots for the analysis of complex sys-
Moreno I. Coco is an Independently Funded Post-Doctoral tems. Physics Reports, 438(5), 237–329.
Researcher at the University of Lisbon, Department of Psy- Richardson, D. C., & Dale, R. (2005). Looking to under-
chology. His research interests span the cognitive science on stand: The coupling between speakers’ and listeners’ eye
a variety of topics, such as psycholinguistics, visual cogni- movements and its relationship to discourse comprehen-
tion and more recently, dialogue. His work combines ex- sion. Cognitive Science, 29, 39–54.
perimental data collection with experimental modeling. He Shockley, K., Santana, M., & Fowler, C. (2003). Mutual in-
is an avid user, and active developer of R (e.g., crqa), and terpersonal postural constraints are involved in cooperative
has been invited to organize hands-on workshops on analyz- conversation. Journal of Experimental Psychology Human
ing eye-tracking data (his primary expertise) using R. He will Perception and Performance, 29, 326-332.
mostly cover the hands-on part of the tutorial. Yu, C., & Smith, L. B. (2013). Joint attention without gaze
Rick Dale is an Associate Professor at the University of following: human infants and their parents coordinate vi-
California, Merced. He is a full rounded cognitive scien- sual attention to objects through eye-hand coordination.
tist with interests and expertise in a wide range of areas of PloS one, 8(11), e79659.
research, from interpersonal dynamics to deception. His re-

12
Programming online experiments with jsPsych
Joshua R. de Leeuw (jodeleeu@indiana.edu)
Department of Psychological and Brain Science & Program in Cognitive Science
Indiana University, 1101 E. 10th St.
Bloomington, IN 47405 USA

Keywords: online experiments; methodology; jsPsych;


feature of jsPsych is that each task is defined in its own code
JavaScript file, known as a plugin. Plugins have a standardized, yet
extremely flexible, structure. This makes it possible to
Overview create custom plugins for tasks that are not possible with the
set of plugins included in jsPsych. It is also easy to share
This tutorial is an introduction to jsPsych, which is a free plugins, to make replications and further manipulations of a
and open-source software package for creating experiments particular task relatively easy to implement for other
that run in a web browser. Participants in the tutorial will researchers.
learn how to build an experiment using jsPsych, and how to Information about jsPsych was presented at the 2014
extend and customize jsPsych for novel experimental Cognitive Science Society meeting as part of a larger
paradigms. tutorial about creating online experiments (de Leeuw et al.,
Running experiments online is a popular method among 2014). This tutorial will go into significantly more depth,
cognitive scientists. Data collection is (extremely) fast and covering more features of the library, how to develop new
cheap, and the quality of the data is generally quite high tasks/plugins, and demonstrating a set of new features that
(Buhrmester, Kwang, & Gosling, 2011; Crump, McDonnell, were added in a major update in October 2014. This update
& Gureckis, 2013; Simcox & Fiez, 2014; Zwaan & Pecher, made it possible to implement a variety of different
2012). There are methodological benefits as well, such as a experimental designs that were previously impossible with
more diverse subject pool (Arnett, 2008; Ross, Irani, jsPsych, including conditional branching and looping
Silberman, Zaldivar, & Tomlinson, 2010) and a reduction in structures. Other new features from the update include the
possible experimenter-induced biases. ability to easily randomize trial order, repeat sets of trials,
Building an experiment that can be run online requires and automatically display a progress bar. For a complete list
proficiency in web development techniques that many of features, see the online documentation at
researchers lack. As a result, there is a demand for tools that http://docs.jspsych.org.
make online experiments easier to develop and run. A few The tutorial is targeted at researchers who have some
such tools are now available and used within the cognitive familiarity and comfort with programming and an interest in
science community, including PsiTurk (McDonnell et al., developing experiments for the web. Researchers who have
2012), QRTEngine (Barnhoorn, Haasnoot, Bocanegra, & no programming background may find it difficult to follow
van Steenbergen, 2014), and jsPsych (de Leeuw, 2014). along, as the basics of programming won’t be covered. The
The main benefit of using jsPsych is that it reduces the tutorial should be of interest to researchers with all levels of
complexity of programming experiments for the web. web-development expertise. Those who are less familiar
Researchers using jsPsych will still need to know how to with web-development techniques will find it easier to learn
program (in JavaScript), but the programming tasks will to use jsPsych than learning to create experiments from
map more naturally on to the design (rather than the scratch, while researchers with a web-development
implementation) of the experiment. For example, it’s not background may find that jsPsych offers a streamlined way
necessary to write code that will determine what key was of building experiments that is more efficient than
pressed and what the response time is. jsPsych handles this, programming experiments on a case-by-case basis.
and other functionality that is common across most Participants are strongly encouraged to bring a laptop with a
experiments, such as figuring out which task/trial to run programming-friendly text editor, such as Atom
next, controlling the flow of the participant through the (http://www.atom.io), to follow along, but may also find it
study, storing data, and so on. However, it is necessary to informative to just observe and learn about what is possible
describe, in code, the design of the experiment, including with jsPsych.
what kinds of tasks the subject will complete, what stimuli
they will see, how long displays will last, and so on.
Summary of Tutorial Content
Experiments in jsPsych are composed of individual tasks,
such as showing the subject instructions, displaying a The morning session of the tutorial will be focused on
stimulus and getting a response, or filling out a survey describing the capabilities of jsPsych and how to use the
question. These tasks are assembled, by the researcher, into software. In the afternoon, tutorial participants will have the
a timeline. A timeline describes the tasks, the parameters for opportunity to work hands-on building a jsPsych
the tasks, and what order the tasks will occur. A main design experiment. Participants are encouraged to work on
developing experiments for their own research, and should

13
bring materials such as stimuli needed to assemble the References
experiment.
The morning presentation will be divided in three parts. Arnett, J. J. (2008). The neglected 95%: Why American
The first part will be a lecture-style tour of jsPsych. This psychology needs to become less American. The
will include describing the conceptual design goals behind American Psychologist, 63(7), 602–14. doi:10.1037/0003-
jsPsych, what problems it aims to solve, and situations in 066X.63.7.602
which it is and isn’t a useful tool. jsPsych will be compared Barnhoorn, J. S., Haasnoot, E., Bocanegra, B. R., & van
to other tools that are available for online research, to Steenbergen, H. (2014). QRTEngine: An easy solution for
highlight the relative strengths and weaknesses of each. A running online reaction time experiments using Qualtrics.
variety of jsPsych experiments will be demonstrated, to give Behavior Research Methods, Advance Online
participants an idea of what’s possible with the library. Data Publication.
from a recent experiment validating the response time Buhrmester, M., Kwang, T., & Gosling, S. D. (2011).
accuracy of jsPsych, and JavaScript response time Amazon’s Mechanical Turk: A new source of
measurement in general, will also be covered (de Leeuw & inexpensive, yet high-quality, data? Perspectives on
Motz, in press). Psychological Science, 6(1), 3–5.
The second part will be a hands-on activity in which doi:10.1177/1745691610393980
participants assemble a jsPsych experiment from start to Crump, M. J. C., McDonnell, J. V, & Gureckis, T. M.
finish. This part of the tutorial will cover the basics of (2013). Evaluating Amazon’s Mechanical Turk as a tool
working with jsPsych all the way through advanced features for experimental behavioral research. PloS ONE, 8(3),
of the library such as conditional looping structures. By the e51382. doi:10.1371/journal.pone.0057410
end of this part, participants will have had the opportunity to de Leeuw, J. R. (2014). jsPsych: A JavaScript library for
build a simple experiment that uses a number of different creating behavioral experiments in a Web browser.
features of the library. Behavior Research Methods, Advance Online
The final part will discuss how to extend and customize Publication. doi:10.3758/s13428-014-0458-y
jsPsych. The main focus of this part will be explaining how de Leeuw, J. R., Coenen, A., Markant, D., Martin, J. B.,
to create a new jsPsych plugin, which enables researchers to McDonnell, J. V, Rich, A. S., & Gureckis, T. M. (2014).
program virtually any computer-based task and include it Online Experiments using jsPsych, psiTurk, and Amazon
within the framework of jsPsych. Customizing jsPsych Mechanical Turk. In P. Bello, M. Guarini, M. McShane,
allows researchers to take advantage of the numerous & B. Scassellati (Eds.), Proceedings of the 36th Annual
features of jsPsych, while still programming their own tasks Conference of the Cognitive Science Society (pp. 41–42).
that may not be possible with the set of tasks that jsPsych Austin, TX: Cognitive Science Society.
includes. de Leeuw, J. R., & Motz, B. A. (in press). Psychophysics in
Materials from the tutorial, including code files and a web browser? Comparing response times collected with
presentation slides, will be made available online. JavaScript and Psychophysics Toolbox in a visual search
task. Behavior Research Methods.
Presenter McDonnell, J., Martin, J. B., Markant, D. B., Coenen, A.,
Rich, A. S., & Gureckis, T. M. (2012). psiTurk. New
Josh de Leeuw is a graduate student at Indiana University.
York, NY: New York University. Retrieved from
He is the creator of jsPsych. In addition to using jsPsych for
https://github.com/NYUCCL/psiTurk
nearly all of his own research, he regularly provides
Ross, J., Irani, L., Silberman, M. S., Zaldivar, A., &
assistance and advice to researchers who are using the
Tomlinson, B. (2010). Who are the crowdworkers?
platform. He has presented several talks and tutorials about
Shifting demographics in Mechanical Turk. In CHI EA
jsPsych, including at the 2014 Cognitive Science Society
’10 (pp. 2863–2872). New York, NY: ACM.
meeting in Quebec City (de Leeuw et al., 2014).
Simcox, T., & Fiez, J. A. (2014). Collecting response times
using Amazon Mechanical Turk and Adobe Flash.
Behavior Research Methods, 46(1), 95–111.
doi:10.3758/s13428-013-0345-y
Zwaan, R. A., & Pecher, D. (2012). Revisiting mental
simulation in language comprehension: six replication
attempts. PloS ONE, 7(12), e51382.
doi:10.1371/journal.pone.0051382

14
Tutorial: Bayesian data analysis

John K. Kruschke (johnkruschke@gmail.com)


Department of Psychological and Brain Sciences, Indiana University
1101 E. 10th St., Bloomington, IN 47405 USA
Tutorial web site: http://tiny.cc/BayesAtCogSci2015

Keywords: Bayesian; data analysis; regression; hierarchical Audience


model; statistics; methods; Markov chain Monte Carlo; R pro-
gramming language; p value; confidence interval; puppies The intended audience is graduate students, faculty, and other
researchers, from all disciplines, who want a ground-floor in-
troduction to doing Bayesian data analysis.
AYESIAN DATA ANALYSIS is superseding traditional

B methods in sciences from anthropology to zoology.


Bayesian methods solve many problems inherent
in p values and confidence intervals. More importantly,
Content and Schedule
9:00–10:20. The day begins with a genuine beginner’s in-
troduction to foundational concepts. An introductory chapter
Bayesian methods are more richly and intuitively informa- that covers this material is available online at the tutorial’s
tive. Bayesian analysis applies flexibly and seamlessly to web site (shown under the title / byline of this document).
simple situations or complex hierarchical models and real- The session continues with a complete example of Bayesian
istic data structures, including small samples, large samples, comparison of two groups. A video summarizing this mate-
unbalanced designs, missing data, censored data, outliers, etc. rial is also available at the tutorial’s web site.
Bayesian analysis software is flexible and can be used for a 10:20–10:40. Break
wide variety of data-analytic and psychometric models. 10:40–12:00. The second morning session covers the ideas
This full-day tutorial presents a ground-level, hands-on in- behind the essential algorithms that make modern Bayesian
troduction to doing Bayesian data analysis. The presenter is analysis possible: Markov chain Monte Carlo (MCMC).
an award-winning teacher who has honed new materials from While it is important to understand the ideas of MCMC, for-
many previous courses. He has written an acclaimed text- tunately we don’t have to deal with MCMC directly because
book on the topic, now greatly expanded in its second edi- the programming language JAGS makes it easy to implement
tion. The tutorial materials include free software and numer- a huge variety of models. By lunch time you will have a
ous programs that can be used for real data analysis. chance to make JAGS do Bayesian analyses for you.
12:00–1:30. Lunch (on your own).
Objectives
1:30–2:50. The first afternoon session considers frequently
Attendees will learn: used models, including various regression models. You will
see how easy it is to implement hierarchical models in JAGS.
• the rich and intuitive information provided by Bayesian
2:50–3:10. Break
analysis and how it differs from traditional (frequentist)
3:10–4:30. The second afternoon explores how null val-
methods.
ues are assessed in Bayesian and frequentist analyses. After
• the concepts and hands-on use of modern algorithms
briefly reviewing the perils of p values and the con game of
(”Markov chain Monte Carlo”) that achieve Bayesian anal-
confidence intervals, two Bayesian approaches to null value
ysis for realistic applications.
assessment will be explored.
• how to use the free software, called R and JAGS, for
4:30–5:00. The tutorial concludes with an open question-
Bayesian analysis, along with many programs created by
and-answer period.
the instructor that are readily useable and adaptable for
your research. Presenter
• many useful applications, including comparison of two
The presenter is eight-time winner of Teaching Excellence
groups, regression models, and hierarchical models.
Recognition Awards from Indiana University. He has given
numerous well-received workshops on Bayesian data analy-
Prerequisites
sis, and is the author of the acclaimed book, Doing Bayesian
No specific mathematical expertise is presumed. In particu- Data Analysis (Second Edition, Kruschke, 2015), along with
lar, no matrix algebra and no calculus is used in the tutorial. numerous articles on Bayesian data analysis (e.g., Kruschke,
Some previous familiarity with statistical methods such as a 2010, 2013). The presenter is Professor of Psychological and
t-test or linear regression can be helpful, as would be some Brain Sciences, and Adjunct Professor of Statistics, at Indiana
previous experience with computer programming, but these University in Bloomington. He was awarded the Troland Re-
are not crucial. search Award from the National Academy of Sciences, and

15
the Remak Distinguished Scholar Award from Indiana Uni- (A) Prior (C) Prior
versity. He has been on the editorial boards of various scien-

0.8
Credibility
tific journals, including Psychological Review, the Journal of
Experimental Psychology: General, and the Journal of Math-

0.4
95% HDI
ematical Psychology. (On the other hand, he put pictures of 0.05 0.997

0.0
puppies on the cover of the book he wrote.) A B C D 0.0 0.2 0.4 0.6 0.8 1.0
Possibile Suspects θ
Computer software and background material
(B) Posterior (D) Posterior for z=36, N=50
Attendees are strongly encouraged to bring a notebook com- Data: A, B, & C mode = 0.72

0.8
Credibility
puter to the tutorial. Install software on your notebook com- are impossible
puter before arriving. See instructions at the tutorial’s web

0.4
site. Be sure to install R, RStudio, JAGS, and the programs 95% HDI
0.587 0.83

0.0
from the book. All software is free. A B C D 0.0 0.2 0.4 0.6 0.8 1.0

The book, Doing Bayesian Data Analysis, Second Edition, Possibile Suspects θ
is highly recommended as background and follow-up to the
tutorial. Extensive information about the book, and a link to Figure 1: Bayesian analysis re-allocates credibility across
a publisher’s discount, can be found at the tutorial’s web site. possibilities. Left column: Reasoning of Sherlock Holmes.
Slides from the presentations at the tutorial will also be made (A) Prior distribution of credibility (i.e., culpability) across
available to attendees. four suspects for a crime. (B) After data indicate that sus-
pects A, B, and C could not have committed the crime, the
What is Bayesian data analysis and why learn it? posterior distribution loads all credibility on suspect D. Right
Bayesian reasoning is simply the re-allocation of credibility column: Bayesian estimation of the cure probability, θ, of a
across possibilities. For a given domain of data, we begin drug. (C) Prior distribution is broad, meaning a wide range
with a set of possible explanations and the prior credibility of of cure probabilities is possible. (D) After observing 36 cures
each explanation. Then we observe some data, and re-allocate in 50 patients, the posterior distribution is narrower, and pre-
credibility toward the explanations that are more consistent cisely displays the most credible probability and the uncer-
with the data. This sort of re-allocation is intuitive in every- tainty of the estimate. (HDI = highest density interval.)
day reasoning, as when Sherlock Holmes argued that when
you have eliminated the impossible, whatever remains must
be the truth — as illustrated in the left side of Figure 1. The
mathematically exact way to re-allocate credibility is done by
Bayesian analysis. A realistic illustration appears in the right
column of Figure 1.
Anything we want to know from the analysis is directly
“read off” the posterior distribution; e.g., the most credible
value and the exact uncertainty of the estimate. There is no
need to derive p values (and their associated confidence in-
tervals) from auxiliary assumptions about sampling distribu-
tions and null hypotheses. The posterior distribution provides
exactly the information that we intuitively already think that Figure 2: Two historical trends in data analysis converge on
frequentist analysis provides but does not. Bayesian software Bayesian estimation with uncertainty, as taught in this tuto-
applies seamlessly to simple and complex models. rial. (NHST = null hypothesis significance testing. MLE =
The analysis methods are at the convergence of two histor- maximum likelihood estimate.)
ical trends in the practice of data analysis, shown in Figure 2.
The trend from frequentist to Bayesian methods is shown
across columns. A second trend, from a focus on null hypoth- Kruschke, J. K. (2013). Bayesian estimation supersedes
esis testing to a focus on estimation with uncertainty, is shown the t test. Journal of Experimental Psychology: General,
across rows. Each cell of Figure 2 indicates combinations of 142(2), 573–603. (DOI: 10.1037/a0029146)
method and focus. This tutorial aims at the convergence of Kruschke, J. K. (2015). Doing Bayesian data analysis, Sec-
the two trends in the lower-right cell. ond Edition: A tutorial with R, JAGS, and Stan. Burlington,
MA: Academic Press / Elsevier.
References
Kruschke, J. K. (2010). What to believe: Bayesian methods
for data analysis. Trends in Cognitive Sciences, 14(7), 293–
300. doi: 10.1016/j.tics.2010.05.001

16
Full Day Tutorial on Quantum Models of Cognition and Decision
Jennifer S. Trueblood (jstruebl@uci.edu)
Department of Cognitive Sciences, University of California, Irvine, CA 92697 USA

James M. Yearsley (james.yearsley.1@city.ac.uk)


Department of Psychology, City University London, Northampton Square, London, EC1V 0HB, UK

Zheng (Joyce) Wang (wang.1243@osu.edu)


School of Communications, Center for Cognitive and Brain Sciences
The Ohio State University, Columbus, OH 43210 USA

Jerome R. Busemeyer (jbusemey@indiana.edu)


Cognitive Science, Indiana University, 1101 E. 10th Street, Bloomington, IN 47405 USA

Keywords: classical information processing; quantum introduces and trains cognitive scientists on this promising
information processing; logic and mathematical new theoretical and modeling approach.
foundation; Bayesian probability; quantum probability;
Markov and quantum processes; quantum Presenters
entanglement; quantum game theory; conceptual
combinations; decision making; memory. Jennifer Trueblood is an assistant professor at the
University of California, Irvine. She has published many
General Purpose articles on the topic of quantum cognition, and her work
This full day tutorial is an exposition of a rapidly has been funded by NSF. James M. Yearsley is a research
growing new alternative approach to building assistant at City University, London. He has a PhD in the
computational models of cognition and decision based on foundations of quantum theory from Imperial College,
quantum theory. The cognitive revolution that occurred in London and worked in the Centre for Quantum
the 1960’s was based on classical computational logic, and Information and Foundations at the University of
the connectionist/neural network movements of the 1970’s Cambridge. Zheng (Joyce) Wang is an associate professor
were based on classical dynamical systems. These classical at The Ohio State University. She was Co-Editor for a
assumptions remain at the heart of both cognitive special issue on quantum cognition that appeared in Topics
architecture and neural network theories, and they are so in Cognitive Science (2013), Vol. 5 (4)). Her work on
commonly and widely applied that we take them for quantum cognition has been funded by NSF and AFOSR.
granted and presume them to be true. What are these Jerome Busemeyer is Provost Professor of Psychological
critical but hidden assumptions upon which all traditional and Brain Sciences at Indiana University. He is Editor of
theories rely? Quantum theory provides a fundamentally Decision and Associate Editor of Psychological Review,
different approach to logic, reasoning, probabilistic and was Editor of Journal of Mathematical Psychology. He
inference, and dynamical systems. For example, quantum is also author with Peter Bruza of the book Quantum
logic does not follow the distributive axiom of Boolean models of Cognition and Decision.
logic; quantum probabilities do not obey the disjunctive
axiom of Kolmogorov probability; quantum reasoning does Previous Tutorials and Symposia
not obey the principle of monotonic reasoning. It turns out The tutorial has been presented at the Cognitive Science
that humans do not obey these restrictions either, which is meetings in Nashville (2007), Washington DC (2008),
why we consider a quantum approach. Amsterdam (2009), Sopporo (2012), Berlin (2013), and
This tutorial will provide an exposition of the basic Quebec City (2014) with about 30 to 50 participants each
assumptions of classical versus quantum theories. These time. The ratings from participants after the tutorial were
basic assumptions will be examined, side-by-side, in a all very positive. Also, this tutorial follows a symposium
parallel and elementary manner. The logic and on quantum cognition at the Cognitive Science meeting
mathematical foundation of classical and quantum theory 2011 whose papers appeared as a special issue in Topics in
will be laid out in an accessible manner that uncovers the Cognitive Science (2013). A similar tutorial was presented
mysteries of both theories. We will show that quantum at the 3rd and 4th Annual Meetings on Quantum
theory provides a unified and powerful explanation for a Interaction in Saarbruecken, Germany (2009) and
wide variety of paradoxes found in human cognition and Aberdeen, Scotland (2010), and at the Society for
decision ranging from attitude, inference, causal reasoning, Mathematical Psychology (2012) and BRiMS (2013), with
judgment and decision, conceptual combinations, memory about 40 participants in each.
recognition, and associative memory. This tutorial

17
Participants Background dynamics of an open quantum system and show how the
This tutorial will introduce participants to an entirely ‘quantum-ness’ of a cognitive system may be lost through
new area and no previous experience or background with interaction with its environment. The implications for
quantum theory will be assumed. No background in cognitive models will be discussed. (1.5 hours)
physics is required. In fact, except for a few simple 5. Finally, we will review progress in quantum cognition
examples to motivate the idea, little or no reference to research and propose future directions. (30 minutes)
physics will be made during main part of the tutorial. What See the references and the website below for some of the
is required is an elementary background in classical logic material to be covered and relevant background material:
and probability. During the tutorial, we will review basic http://mypage.iu.edu/~jbusemey/quantum/Quantum
concepts of linear algebra needed for quantum theory (e.g., Cognition Notes.htm
vectors, projectors, unitary transformations).
References
Material to be Covered Busemeyer, J. R., & Bruza, P. D. (2012). Quantum
1. First, we will examine major differences between models of cognition and decision. Cambridge, UK:
classical versus quantum theories of probability. The Cambridge University Press.
concept of superposition is introduced and distinguished Busemeyer, J. R., Pothos, E., Franco, R., & Trueblood, J.
from classical probability mixtures. The important issue of S. (2011). A quantum theoretical explanation for
measurement in classical and quantum systems will be probability judgment ‘errors.’ Psychological Review,
compared and examined. We will include several dramatic 118, 193-218.
empirical examples illustrating empirical violations of the Busemeyer, J., Wang, Z., & Lampert-Mogiliansky, A. L.
classical laws of probability (e.g., conjunction, disjunction, (2009). Empirical comparison of Markov and quantum
and total probability) and the parsimonious explanation of models of decision making. Journal of Mathematical
all these violations by quantum theory. (1.5 hours) Psychology, 53, 423-433.
2. Then we will examine the differences between Busemeyer, J. R., Wang, Z., & Townsend, J. T. (2006).
classical and quantum dynamical systems. The basic idea Quantum dynamics of human decision-making. Journal
of a Markov processes will be introduced and compared of Mathematical Psychology, 50, 220-241.
with quantum processes. (Cognitive architectures and Pothos, E. M., & Busemeyer, J. R. (2009). A quantum
many neural networks can be represented as Markov probability explanation for violations of “rational”
processes). A parallel development of Markov and decision theory. Proceedings of the Royal Society B. 276
quantum processes will be shown. The concept of a state (1665), 2171-2178.
will be distinguished for Markov and quantum systems. Pothos, E. M., & Busemeyer, J. R. (2013). Can quantum
The effects of measurement on the state of the system are probability provide a new direction for cognitive
compared for Markov and quantum systems. A key goal is modeling? Behavioral & Brain Sciences, 36, 255-274.
to show when and how quantum processes depart from Trueblood, J. S. & Busemeyer, J. R. (2011). A Quantum
Markov processes, and how we can empirically test probability account of order effects in inference.
whether a system is Markov or quantum. (1.5 hours) Cognitive Science, 35, 1518-1552.
3. Next, we will use a concrete example to show how to Wang, Z., Busemeyer, J. R., Atmanspacher, H., & b
build computational models based upon quantum theory. Pothos, E. M. (2013). The potential of using quantum
We will present the details of MATLAB and R programs theory to build models of cognition. Topics in Cognitive
used to compute the choice probability and response time Science, 5, 672-688.
predictions of a dynamic quantum model that has been Wang, Z., & Busemeyer, J. R. (2013). A quantum question
developed to explain three ongoing research programs in order model supported by empirical tests of an a priori
cognitive and decision making: violations of the “sure and precise prediction. Topics in Cognitive Science, 5,
thing principle” of rational decision theory, violations of 689-710.
dynamic consistency in decisions, and interference of Wang, Z., Solloway, T., Shiffrin, R. M., & Busemeyer, J.
categorization on decisions. (1 hour) (2014). Context effects produced by question orders
4. In the fourth part, we will introduce advanced tools reveal quantum nature of human judgments. PNAS,
and concepts needed for building quantum models of 111(26), 9431-9436.
realistic cognitive systems. We will show how the Yearsley, J.M. and Pothos E.M. (2014). Challenging the
description of a quantum state may be extended to include classical notion of time in cognition: a quantum
both quantum and classical uncertainty, and we will perspective. Proc. R. Soc. B 281, 1781, 20133056.
explain how to compute the entropy of a quantum state.
We will introduce the notion of a POVM and explain how Acknowledgments
these may be used to model realistic, noisy, measurements. This tutorial and related research is supported by U.S.
We will discuss the concept of an open quantum system National Science Foundation (SES-1153726, SES-
and the difference between unitary and non-unitary 1153846, SES-1326275) and AFOSR (FA9550-12-1-
dynamics. Finally we will introduce a simple model for the 0397).

18
Connecting learning, memory, and representation in math education
Martha Alibali1 (mwalibali@wisc.edu)
Chuck Kalish2 (kalish@wisc.edu) & Christine Massey1 (massey@seas.upenn.edu) &
Timothy T. Rogers1 (ttrogers@wisc.edu) Phil Kellman2 (Kellman@cognet.ucla.edu)
1 1
Department of Psychology, UW Madison Institute for Research in Cognitive Science, University of
1025 W. Johnson Street, Madison, WI 53706 USA Pennsylvania, 3401 Walnut Street, Philadelphia PA 19004

2
2
Department of Educational Psychology, UW Madison Department of Psychology, UCLA, 1285 Franz Hall, Los
1202 W. Johnson Street, Madison, WI 53706 USA Angeles CA 90095 USA

Vladimir Sloutsky (sloutsky.1@osu.edu) James L. McClelland (mcclelland@stanford.edu)


Department of Psychology, Ohio State University & Kevin W. Mickey (kmickey@stanford.edu)
1835 Neil Avenue, Columbus, OH 43210 USA Department of Psychology, Stanford University, 450
Serra Mall, Stanford, CA 94305 USA

patterns of transfer shown by children and adults in


Keywords: numeracy; math; perception; learning; memory;
cognition; models; education arithmetic. Phil Kellman and Christine Massey will show
that mathematical competency can improve when children
Introduction learn to efficiently encode the perceptual structure of
equations. Vladimir Sloutsky will consider
In math education the goal is for children not only to interrelationships between learning of mathematical and
master the materials and problems presented, but to object concepts in development. Jay McClelland and Kevin
understand underlying principles and properties that can be Mickey will discuss new research investigating the
applied broadly to new problems and situations. Teachers in representational prerequisites that might underlie conceptual
the classroom and policy-makers in Washington thus are both understanding of trigonometric functions. A short group
faced with what is essentially a cognitive question: What question period will follow the four talks.
instructional regimes and practices will produce rapid
learning, deep understanding, and broad transfer? Alibali, Kalish & Rogers: Connecting learning in
This question has often been approached without
mathematical and non-mathematical domains.
connection to cognitive theories of learning, memory, and
representation, but the gap has begun to narrow. On one hand, Different learning tasks can elicit qualitatively different
it is now known that domain-general learning mechanisms patterns of memory and generalization. In paired-associates,
can acquire quite abstract and structured representations that participants who learn to produce "dishtowel" to the probe
go beyond the perceptual structure of the environment—a "locomotive" can correctly generate the reverse pairing
critical requirement for any theory of mathematical (producing "locomotive" given "dishtowel"), but, because
knowledge. Conversely studies in math cognition have pairs are arbitrary, cannot generalize to new probes (e.g.
revealed counter-intuitive behaviors that find ready "caboose"). In categorization, participants remember features
explanations in cognitive models of learning in other that aid in predicting the category label and use these to
domains. For instance, children, adults, and even math generalize, but fail to learn or exploit other item properties.
teachers reliably judge some three-sided figures to be better In property induction, participants learn slowly but remember
triangles than others, sometimes denying that irregular three- and generalize all manner of properties.3 In experiments with
sided figures are in fact triangles.1 Children transitioning adults and children we show analogous phenomena in
from arithmetic to algebra often generate incorrect solutions arithmetic learning. When the graphical elements of an
to equations because they have learned to ignore the equal equation are viewed as arbitrary symbols, participants learn
sign2. Such examples suggest that math learning can be individual problems without transfer, as in paired associates.
subject to the same factors that govern learning other When the quantitative "meaning" of the problem symbols is
domains. Yet it remains unclear whether such effects are highlighted, participants acquire a transferable mapping from
epiphenomenal, or whether they hint at important common problem quantities to response quantities, similar to
principles underlying concept acquisition across multiple categorization. The extent of transfer in this setting depends,
domains. however, on the task: practice retrieving a missing sum
Our symposium investigates this question by bringing transfers to new missing-sum problems, but not to related
together scientists whose research spans the gap between missing-addend problems. The broadest transfer occurs when
cognitive and educational science in the domain of participants practice with a mix of problem types, in a setting
mathematical knowledge. Martha Alibali, Chuck that emphasizes quantitative relationships among elements—
Kalish and Tim Rogers consider how cognitive memory the same properties that produce broad transfer in object
models from non-mathematical domains can shed light on the concepts. These results suggest a tighter coupling between

19
learning in mathematical and non-mathematical domains continued to use and remember all features. Thus, though
than has previously been appreciated. their response strategy changed, their representation did not.
From these findings we argue that perceptually-rich problem
Kellman & Massey: Perceptual structure and instantiations may hinder generalization in math because, like
adaptive learning in math education. stimuli in our research, they possess one relevant
deterministic feature among many irrelevant features. If
While learning of complex structure is often attributed to children naturally acquire dense probabilistic category
higher-order processes, we argue that perceptual learning structures, they may fail to generalize practice problems with
(PL)—experience-driven changes in the process and content sparse structure. We then demonstrate such an impaired
of information extraction—plays a much greater role than has transfer in learning of mathematical concepts in young
previously been appreciated. We consider PL as a crucial children.
component of learning and expertise in mathematics and
other complex cognitive domains. Whereas most formal McClelland & Mickey: Building a core
instruction emphasizes declarative and procedural conceptual structure for trigonometry.
components of learning, learning to extract relevant structure
in mathematical problems and representations provides the How can we help students gain a grasp of the basic ideas
pattern recognition required for effective use of facts and underlying trigonometric functions? Our approach links to
procedures. We will briefly review research on PL the ideas of Robbie Case, who understood the mental number
interventions in the form of perceptual/adaptive learning line as a core conceptual structure for two-digit addition and
modules (PALMs) that facilitate discovery of structure and subtraction upon which one could build an understanding of
recognition of patterns in mathematical domains, including decimal numbers and fractions.4 We extend this approach to
preliminary results from a large efficacy study currently in the 2D coordinate plane, taking the 'unit circle' as a core
progress. These efforts illustrate the promise of PL structure for grounding the extended definitions of the
interventions, as shown on tests of mathematical competence. trigonometric functions outside the range of right triangles.
We also examine direct effects of PL interventions on In empirical studies with Stanford undergraduates, we have
psychophysical endpoints, such as efficient encoding of found that (a) students who report using the unit circle do
equations. Results indicate that even relatively brief PALM better on an assessment of their understanding of
interventions aimed at improving students' seeing of structure trigonometric identities than those who report using rules or
and transformation in algebraic equations leads to reliable other visualizations; (b) a brief presentation of the core unit
changes in basic information extraction. Encoding circle ideas produces better generalization to identities not
improvements were shown most strongly by participants who explicitly covered in the presentation, relative to a rule-based
were initially less proficient at algebra. These changes, which presentation; but (c) only those performing above chance on
were detectable 24 hours after training, provide direct a pre-test showed the benefit from the presentation. A second
evidence for durable changes in information encoding study assessed the unit-circle intervention on a group of high-
produced by a PALM targeting a complex mathematical skill. school seniors, none of whom benefitted. This has led us to
construct a structured series of didactic presentations and
Sloutsky: What can we learn about mathematical interleaved activities designed to ensure students have a well-
cognition from object category learning? grounded understanding of all of the elements on which the
unit circle definitions build. We will report the results of new
The primary difference between mathematical and object studies with this interleaved intervention, and will consider
concepts lies in category structure: the former are rule-based
the implications of our studies for both education and for
and statistically sparse (i.e., few category-relevant and many
understanding the cognitive underpinnings of math
irrelevant features) while the latter are statistically dense (i.e.,
knowledge.
many category-relevant features). Research in object
category learning may then elucidate acquisition of math
concepts. We review evidence that children distribute References
1
attention among multiple stimulus dimensions, making it Lupyan, G. (2013). The difficulties of executing simple algorithms:
difficult to learn statistically sparse concepts like those Why brains make mistakes computers don’t. Cognition, 129(3),
615-636.
central to mathematics. Consequently children and adults 2
McNeil & Alibali (2005). Knowledge change as a function of
may extract different structures from the same learning experience: All contexts are not created equal. Journal of
experiences. Participants learned a category possessing both Cognition and Development, 6(2), 285-306.
(a) a single deterministic rule-like feature and (b) multiple 3Chin-Parker & Ross (2004). Diagnosticity and prototypicality in

inter-correlated probabilistic features. Whereas 4-5-year-olds category learning: A comparison of inference learning and
used multiple probabilistic features to generalize and were classification learning. JEP: LMC, 30(1), 216-226.
more likely to remember these, adults used the deterministic 4 Case et al. (1996). The role of central conceptual structures in the
feature to generalize and were less likely to remember other development of children’s thought. Monographs of the SRCD,
features. When the deterministic feature was made salient, 61(1-2), v-265.
children were more likely to use it in generalization, but they

20
Causality and Agency Across Cultures and Languages

Organizers
Sieghard Beller (Sieghard.Beller@uib.no) & Andrea Bender (Andrea.Bender@uib.no)
Department of Psychosocial Science, University of Bergen, Norway

Presenters
Jürgen Bohnemeyer (jb77@buffalo.edu) York Hagmayer (yhagmay@uni-goettingen.de)
Department of Linguistics & Center for Cognitive Science, Department of Psychology,
University at Buffalo SUNY, USA University of Göttingen, Germany
Annelie Rothe-Wulf Rita Astuti (R.Astuti@lse.ac.uk)
(Annelie.Rothe@psychologie.uni-freiburg.de) Department of Anthropology,
Department of Psychology, Freiburg University, Germany London School of Economics, London, UK

Why does wood float on water? What made Jim shout at me, • Last, but not least, how can we make sure that the meth-
and why had he to be so offending? Who is responsible for ods we use to investigate potential differences across cul-
my son’s sickness? And why should incest be wrong? All tures and languages do really capture the relevant issues
these questions share one important feature: They ask for in an unbiased manner?
causal explanations. Causality is a core concept in our at- Our symposium attempts to advance this field of research at
tempts to make sense of the physical world and of social in- the heart of cognitive science. It brings together researchers
teractions; and this makes causal cognition a topic of prime from various of its sub-fields, who will present theoretical
interest for cognitive science. Yet, in spite of an increasing analyses and empirical findings on those factors that may
body of high-quality and high-profile research, most previ- constrain, trigger, or shape the way in which humans think
ous studies paid only incidental attention to the potential of and talk about causal relationships.
cognitive and linguistic diversity in causal cognition. • Jürgen Bohnemeyer has designed a large-scale survey on
The cross-cultural evidence available so far (reviewed in the linguistic representation of causality that combines
Bender, Beller, & Medin, subm.) indicates that culture plays approaches from psychology, linguistics, and anthropolo-
a crucial role in causal cognition on various levels and in all gy (e.g., Bohnemeyer et al., 2010; Moore et al., in press).
domains. It affects not only how, but even whether people
• Annelie Rothe-Wulf and colleagues combine psychologi-
engage in causal explanations, by defining the settings in
cal and anthropological expertise to investigate the effect
which causal cognition occurs, the manner in which poten-
of cultural concepts and linguistic cues on causal cogni-
tial factors are pondered on, and the choices for highlighting
tion (Beller et al., 2009; Bender & Beller, 2011).
one of several potential causes or for expressing them lin-
guistically in one way or another (e.g., Astuti & Harris, • York Hagmayer has for many years specialized in psy-
2008; Bender & Beller, 2011; Bohnemeyer et al., 2010; No- chological and philosophical aspects of causal reasoning
renzayan & Nisbett, 2000; and see the contributions in Bel- (Hagmayer & Sloman, 2009; Waldmann & Hagmayer,
ler, Bender, & Waldmann, 2014). 2013); here he examines, in collaboration with an anthro-
These findings justify the call for a more thorough investi- pologist, cross-cultural data on explanations for illnesses.
gation of the possibly constitutive role that culture and lan- • Rita Astuti, one of Europe’s leading (cognitive) anthro-
guage may play for causal cognition (Widlok, 2014). While pologists (Astuti & Harris, 2008; Astuti, Solomon, &
it is plausible that most causal learning, and even a consider- Carey, 2004), investigates causal reasoning related to bio-
able proportion of causal explanations, will be invariant logical concepts and moral processes.
across culture, without thoroughly scrutinizing each of the By integrating insights from their various disciplinary back-
candidates for invariance we are not in a position to draw grounds, this symposium will span a broad range of the sub-
any generalizations. Important questions have thus remained fields of cognitive science in an exemplary manner.
unanswered:
• Along which dimensions do socio-linguistic groups dif- Studying the representation of causality
fer in how they speak about causality, and to what extent across languages
do these differences affect how people represent causal
relations? Jürgen Bohnemeyer
• Is causal reasoning always based on the same cognitive This presentation surveys the semantic and conceptual prop-
mechanisms and principles, or do our cultural back- erties of linguistic representations of causal relations across
ground and our native language shape how we process re- languages. The principal aim is to explore the challenges in-
spective information? volved in constructing an ‘etic grid’ for a semantic typology
• How are multiple explanatory frameworks organized and of causative constructions. Etic grids are non-language-spe-
activated for accounts of illnesses or moral reasoning? cific sets of notional variables that jointly define conceptual

21
domains carved up by the meanings of language-specific ex- proposition, we look at how Malagasy people respond to in-
pressions (Moore et al., in press). The presentation will draw cest. While they do not seem to take intentionality into ac-
on the results of a pilot study (Bohnemeyer et al., 2010) in- count in the specific case of incest, when they reason about
volving four unrelated languages: Ewe (Gbe; Ghana, Togo), other types of wrong doing the role of intentionality is well
Japanese, Lao (Tai-Kadai; Laos), and Yucatec (Mayan; Mex- understood. We therefore argue that, when people contem-
ico). The findings will be compared to proposals in the re- plate incest and its consequences, they simultaneously con-
cent typological literature (e.g., Song, 1996). A set of 10 sider two quite different issues. Using the insights we derive
variables will be proposed, all of which have been shown to from this Malagasy case study, we re-examine the results of
potentially influence the perceived simplicity or ‘directness’ Haidt’s (2001) psychological experiment on moral dumb-
of causal chains, and through it the complexity of linguistic foundedness. We suggest that the dumbfoundedness that was
representations. documented among North American students may be ex-
plained by the same kind of complexity that we found in
What makes the difference? Content effects as Madagascar. In light of this, we also note the limitations of
moderators of cross-cultural variability anthropological methods and the benefits of closer cross-dis-
ciplinary collaboration.
Annelie Rothe-Wulf, Gregory Kuhnmünch,
Andrea Bender, & Sieghard Beller
Although causal cognition in the physical domain is regard- References
ed as invariable to culture, recent research yielded complex Astuti, R., & Harris, P. L. (2008). Understanding mortality
patterns of causal attribution within and across cultures for and the life of the ancestors in rural Madagascar. Cognitive
various physical events (Beller et al., 2009; Bender & Beller, Science, 32, 713-740.
2011). One candidate moderator for this variability is the Astuti, R., Solomon, G. E. A., & Carey, S. (2004). Constraints
way in which people construe the content of the event. We on conceptual development. Boston, Oxford: Blackwell.
asked Tongan and German participants to assign causation to Beller, S., Bender, A., & Song, J. (2009). Weighing up phys-
entities involved in a range of physical events. The entities ical causes: Effects of culture, linguistic cues and content.
varied along several dimensions such as concreteness, con- Journal of Cognition and Culture, 9, 347-365.
sistency, or physical type. Content effects emerged in both Beller, S., Bender, A., & Waldmann, M. (Eds.) (2014). Diver-
cultural groups and partially moderated cultural differences. sity and universality in causal cognition. Frontiers in Psy-
In addition, we observed culture-specific patterns, indicating chology: Cognitive Science.
the importance of culturally relevant concepts. Bender, A., & Beller, S. (2011). Causal asymmetry across
cultures: Assigning causal roles in symmetric physical set-
Causes of illness – tings. Frontiers in Psychology, 2:231.
What do different types of causes explain? Bender, A., Beller, S., & Medin, D. L. (subm.). Causal cogni-
York Hagmayer & Ronja Rutschmann tion and culture. In M. R. Waldmann (Ed.), Oxford hand-
book of causal reasoning. New York: Oxford University
Research on lay theories of illness in anthropology and psy- Press.
chology investigated the types of causes people believe in. Bohnemeyer, J., Enfield, N. J., Essegbey, J., & Kita, S.
These causes explain why illness (rather than health) occurs (2010). The macro-event property: The segmentation of
and which type of illness happens under certain conditions. causal chains. In J. Bohnemeyer & E. Pederson (Eds.),
However, research on the questions patients ask indicates Event representation in language. Cambridge: Cambridge
that patients and their relatives also want to know why the University Press.
particular person (rather than another person) was affected Hagmayer, Y., & Sloman, S. A. (2009). Decision makers con-
and why the illness occurred at this particular point in time ceive of their choice as intervention. Journal of Experimen-
(rather than sooner or later). It is an open question whether tal Psychology: General, 138, 22-38.
and which types of causes provide an explanation to these in- Haidt, J. (2001). The emotional dog and its rational tail. Psy-
quiries. We propose a classification scheme of different chological Review, 108, 814-834.
types of causes with respect to the explanations they provide. Moore, R., Donelson, K. T., Eggleston, A., & Bohnemeyer, J.
We argue that many so-called supernatural causes explain (in press). Semantic typology: New approaches to crosslin-
which person is affected at a particular point in time, while guistic variation in language and cognition. Linguistic Van-
many natural causes like somatic and environmental condi- guard.
tions do not. This may explain why people in many cultural Norenzayan, A., & Nisbett, R. E. (2000). Culture and causal
groups believe in supernatural causes. cognition. Current Directions in Psychological Research,
9, 132-135.
The causal cognition of wrong doing: Song, J. J. (1996). Causatives and causation: A universal-
Incest, intentionality and morality typological perspective. London: Longman.
Waldmann, M. R., & Hagmayer, Y. (2013). Causal reasoning.
Rita Astuti & Maurice Bloch In D. Reisberg (Ed.), Oxford handbook of cognitive psy-
Anthropologists have claimed that, in certain non-western chology. New York: Oxford University Press.
societies, people ignore whether an act of wrong doing is Widlok, T. (2014). Agency, time and causality. Frontiers in
committed intentionally or accidentally. To examine this Psychology: Cognitive Science, 5:1264.

22
Communicating Cognitive Science:
Improving Awareness and Understanding Among People Who are Not Ourselves
Organizers

Kevin A. Gluck (kevin.gluck@us.af.mil)


Air Force Research Laboratory, USA

Wayne Gray (grayw@rpi.edu)


Rensselaer Polytechnic Institute, USA

Keywords: Communicating; Cognitive Science; Public Most of us struggle to convey our objectives and results
Understanding; Awareness; Outreach and relevance in a manner that is understandable by
people without PhDs in the same specialty as our own. By
Introduction contrast, some of us seem very successful at getting the
As cognitive scientists, we invest enormous amounts of word of our good works out. Some of our members author
time in our graduate educations and careers learning to popular books or textbooks (an extremely important way
communicate our findings to others in the form of highly of inspiring people to become members of the next
specialized research papers. Indeed, it is hard to imagine generation of cognitive researchers!), participate in radio
how the nuance and distinctions required to advance our interviews, occasionally appear on TV, and write
science, or any science, could be communicated if such successful blogs. How do they do it? Can their methods
were not the case. However, our work is a public be duplicated by others across the world so as to better
enterprise that is largely sustained by institutions that communicate our aspirations, discoveries, and inventions
promise some return to the public good. In other fields, to the world public?
this return may be primarily in terms of ideas and insight For this symposium, we brought together a group of
into the human condition, as might be the case for people with a history of successfully getting the word out
archaeology and history. The return may be in new about their own and others’ cognitive science research.
fundamental discoveries regarding our physical world, Following an introductory presentation by the organizers,
such as recent progress in nanomaterials that promise this group of distinguished speakers will tell what they do,
eventual translation into new forms of energy, wh y they do it, evaluate its utility, and offer
transportation, and communication. Or it may be focused suggestions for the rest of us for communicating cognitive
on the Pasteur’s Quadrant ( Sto k es, 1 9 9 7 ) of research science in ways that improve awareness and understanding
addressing an immediate practical need, such as an Ebola among people who are not ourselves.
vaccine.
We argue that Cognitive Science is a field where the
return to the public good can take any and all of these Marsha Lovett
three forms. Similar to archaeology or history we can Director, Eberly Center for
promise increased insights into the human condition in Teaching Excellence and Educational Innovation
terms of the nature of the mind, memory, and thought. In
common with Physics, our fundamental research on the
Carnegie Mellon University
nature of cognitive control and the integration of lovett@cmu.edu
perception, cognition, and action promises a long-term
translation into applications and products for reducing One of higher education's current challenges is providing
cognitive workload and increasing human effectiveness. effective instruction to a diverse population of learners.
Likewise, in common with u s e - i n s p i r e d medical Research in both cognitive science and learning science
researchers, we have a long tradition of applying and offers a rich body of theory, results, and methods to help
testing our ideas about learning and decision making by generate and refine strategies to address this challenge. Why
incorporating our research into tutoring systems, is this research on learning not having more of an impact on
guidelines for teachers, and real-time decision aids. educational practice? Beyond the constraints of time,
Although we strive to do the right things for the right resources, and institutional infrastructure, there is an
motivations, many of us would have to admit, if inherent difficulty in appropriately translating research
pressed, that our public profile is slim to non-existent, results – derived from either lab or field studies – to specific
both as individual cognitive scientists and as a global classroom contexts. Learning is a complex process where
scientific discipline. It is unusual to find a person outside multiple factors interact and context matters. Finding a
of academia who has any idea what cognitive science is. balance between acknowledging this complexity while
identifying fundamental principles, features, and

23
mechanisms, is the key to progress. Several research-to- will be needed to create cognitive assistants (“cogs”) for all
practice books and related approaches are discussed in terms occupations, so cognitive scientists will have to work well
of how they achieve this balance. with others studying and producing real-world applications.
These applications are expected to generate enormous
Art Markman quantities of performance data for the field. Understanding
Professor, Department of Psychology the performance of individual experts and novices, as well
University of Texas as teams of people, with their cogs will require new methods
and tools. Second, as more people come to depend on
markman@utexas.edu
assistants to improve their lives, cognitive scientists can
play a bigger role informing the general public about
It should not be hard for cognitive scientists to reach out to a
improvement strategies. This has the potential to unlock the
broader community to enlighten them about the work we do,
citizen scientist in everyone to use data and models to
because of its relevance to most people’s lives. So, why
improve their individual and collective performance.
aren’t more cognitive scientists engaged in outreach? There
However, there are also pitfalls to be avoided. For
are three significant factors that limit outreach activities.
example, the science of flight from the study of birds to
First, the community is not well-trained either in the style of
aviation is both intellectually fascinating and of great
writing and speaking that engages broad audiences or in the
economic significance, but job growth happens primarily in
techniques for promotion of outreach that maximize its
the emerging frontier aerospace or drone-based package
effectiveness. Second, at present, there is no significant
delivery application areas. The pitfall that cognitive
expectation that outreach is part of a successful academic
scientists should avoid is the pitfall of becoming overly
research career. Third, academic institutions do not
narrow in focus. Cognitive science can reach a broader
typically value outreach activities in ways that lead to
audience if the core of the cognitive science community
recognition and promotion. To remedy these problems, we
broadens its scope appropriately and orients towards future
need to increase our attention to training mid-career
opportunities for growth.
scientists in the art of communicating to non-scientific
Cognitive science can learn a great deal and benefit from
audiences in order to make them more comfortable speaking
the study of other academic disciplines and professions that
to groups and talking to reporters. We need to create a
have thrived or dwindled into obscurity (Abbott, 1988,
cultural expectation that mature researchers will tithe to the
2001). As a practical first step, “cognitive scientist” should
field by giving (roughly) ten percent of their time to
to be added to the list of O*NET Online occupations with a
outreach activities. Finally, we need to put pressure on
bright future. For example, nanotechnology is documented
administrations to create awards and recognition for faculty
and has a bright outlook for the future, but cognitive science
and researchers who engage the public as part of their
appears missing. The community should rectify this.
scientific mission.

Jim Spohrer References


Director, University Programs and Abbott, A. (1988). The system of professions: An essay on
the division of expert labor. University of Chicago Press.
Cognitive Systems Institute Abbott, A. (2001). Chaos of disciplines. University of
IBM Chicago Press.
spohrer@us.ibm.com Kelly III, J., & Hamm, S. (2013). Smart Machines: IBM's
Watson and the Era of Cognitive Computing. Columbia
In the coming decade, cognitive science is poised to reach a University Press.
broader community. First, as the era of cognitive computing Stokes, D. E. (1997). Pasteur’s quadrant: Basic science
dawns (Kelly & Hamm, 2013), demand is expected to and technological innovation. Brookings Institution
increase for cognitive scientists with the right skills who can Press, Washington, DC.
lead multidisciplinary teams, thereby creating more and
better jobs for cognitive scientists. Multidisciplinary teams

24
Generative and Discriminative Models in Cognitive Science
organizers and speakers
Bradley C. Love (b.love@ucl.ac.uk)
University College London
Michael Ramscar (michael.ramscar@uni-tuebingen.de)
University of Tübingen

invited speakers
Tom Griffiths (tom_griffiths@berkeley.edu)
University of California, Berkeley
Matt Jones (mcj@colorado.edu)
University of Colorado at Boulder

Generative and Discriminative Approaches The impact of sampling assumptions on learning from
One popular distinction in machine learning is between indirect negative evidence
discriminative and generative models (Ng & Jordan, 2001). Tom Griffiths (with Anne Hsu)
Given the cross fertilization between research in human and
machine learning, the time is ripe to ask whether the mind is A classic debate in cognitive science revolves around
a generative or discriminative learning device. This understanding how children learn complex linguistic
symposium tackles this question from a variety of patterns, such as restrictions on verb alternations and
perspectives. The aim is to explore the explanatory value of contractions, without negative evidence. One factor that has
these two basic views of learning, which cut across existing been suggested as playing an important role in solving this
distinctions in cognitive science (e.g., connectionist vs. problem is indirect negative evidence, in which the absence
Bayesian approaches). of a construction in the input provides evidence against its
In brief, generative and discriminative models grammaticality. We consider two different sets of sampling
characterize the task of the learner differently. Generative assumptions that can operate in language learning,
models attempt to learn an internal model of each class (i.e., corresponding loosely to "generative" and "discriminative"
category). In contrast, discriminative models attempt to find approaches to learning. Only one set of assumptions
a boundary that separates classes. Generative models are licenses use of indirect negative evidence. We demonstrate
typically Bayesian in form, whereas discriminative models in a series of experiments in which adults learn artificial
include decision trees, SVMs, regression approaches, and languages that people can produce behavior consistent with
some (but not all) connectionist models. the predictions of probabilistic models using both sets of
In generative models, the learning task is to estimate the sampling assumptions, depending on how the learning
joint probabilities between all variables. These models problem is presented. These results suggest that people use
assume a hidden or latent variable (e.g., category label) information about the way in which linguistic input is
generates observed features. In contrast, discriminative sampled to guide their learning, and show that adult learners
models perform a conditional estimation. For example, make appropriate use of indirect negative evidence when the
logistic regression only estimates the probability of a class appropriate statistical assumptions are satisfied.
(i.e., category) as a function of the predictive features. In
this sense, discriminative models are more focused by the Language Learning From a Discriminative
task, whereas generative models address a broader Perspective
estimation problem, though models of all types have an Michael Ramscar
inductive bias to make learning tractable.
The development of morphological processing has been
Aims and Relevance the focal topic in a debate over the nature of language,
learning and the mind in cognitive science. Particular
The aim of this symposium is to introduce these powerful
attention has been paid to the systematic nature of children’s
ideas from machine learning to the broader cognitive
morphological errors (for example children tend to go
science community. We will evaluate what these two views
through a phase of saying “mouses” as they learn the
say about cognition and learning, and assess their utility in
morphology of English nouns). Because these errors aren’t
organizing findings in our science. At the broadest level, in
explicitly corrected, it has been argued that the transition to
what sense is the mind a generative or discriminative
adult language cannot be explained by learning, and that the
machine, and how can this understanding direct our future
acquisition of even relatively simple aspects of grammar
empirical and theoretical investigations of the mind?
must involve innate, language specific mechanisms. I'll
describe the background to this debate, the generative
models that have traditionally been proposed to explain

25
these behavioral patterns, and a model of morphological expectancy derived from the base rate affects learning about
development based on discrimination learning that offers a the repetition rate and vice versa. This cue competition
very different perspective on morphological processing. manifests in additional, subtle sequential effects that are
This model also generates clear and surprising predictions, confirmed in the data. These additional sequential effects
in particular that exposure to regular plurals (e.g. rats) can thus appear to be signatures of discriminative learning.
actually result in a decrease in children’s tendency to
overregularize irregular plurals (e.g. say "mouses"). I'll Jones, M., Curran, T., Mozer, M. C., & Wilder, M. H.
review some empirical results showing that testing memory (2013). Sequential effects in response time reveal learning
for items with regular plural labels does result in a decrease mechanisms and event representations. Psychological
in plural overregularization in six-year-olds, but also that it Review, 120, 628-666.
results in increases in four-year-olds. These models and
results indicate that when the learning problem facing Getting Discriminative with a Generative Model
children is characterized discriminatively, Bradley C. Love
overregularization can be seen to both arise and then resolve
itself as a result of the distribution of evidence in the Models, whether generative or discriminative, have an
linguistic environment. I'll discuss the wider implications of inductive bias that makes learning tractable. In this talk, I
these and some similar findings for our understanding of will present a generative model of learning and information
language and human communication. sampling whose inductive bias follows from discriminative
principles. The model, like people, is focused on properly
Ramscar, M., Dye, M. & McCauley, S. (2013d) Error and estimating aspects of the environment that are goal relevant.
expectation in language learning: The curious absence of This focus is consistent with conditional estimation in
‘mouses’ in adult speech. Language, 89(4), 670-793 discriminative models. However, the model also benefits
from the strengths of the generative approach, such as the
Sequential Effects as Signatures of Discriminative ability to support planning and sampling processes critical
Learning in decision making (Giguère & Love, 2013). Like people,
Matt Jones current goals and knowledge determine the information
sampled in the world. Completing the cycle of mutual
An important class of psychological models of influence, the information sampled (i.e., attended) in the
discriminative learning are those that learn incrementally world updates the model's knowledge state. This cycle of
from prediction error. One prediction of this iterated error influence depends on two model components. One model
correction is recency effects. In their simplest form, recency component determines the value of potential sources of
effects are simply a bias toward recent events, such that information. The value of a piece of information depends on
error rates and response times (RTs) are lower when the the decision maker’s goals and assumptions about (i.e.,
current trial matches recent feedback. Generative models knowledge of) the world. The second component of the
can also predict these simple recency effects, by assuming model reflects the decision maker’s knowledge of the world,
nonstationarity in latent environmental parameters (Wilder, which is used by the first component to direct information
Jones, & Mozer, 2009; Yu & Cohen, 2008). This is because gathering. This learning component is updated by the
the nonstationarity assumption leads more recent events to information samples selected by the first component,
be more informative about the current state of the completing the cycle of mutual influence. Human learning
environment. However, discriminative models also predict and eye tracking studies support the model. By introducing
more complex sequential effects that generative models do a notion of attention that focuses on goal-discriminating
not anticipate. This talk will focus on one set of such information, a generative model is imparted with
findings, in binary stimulus identification tasks (Jones, discriminative characteristics and displays human-like
Curran, Mozer, & Wilder, 2013). In this paradigm, behaviors.
sequential effects in RT reveal learning of two statistics of
the trial sequence: the base rate and the repetition rate. That Giguère, G. & Love, B.C. (2013). Limits in decision making
is, RT is faster when the current response matches recent arise from limits in memory retrieval. Proceedings of the
responses (a left response preceded by recent left responses, National Academy of Sciences of the United States of
or a right response preceded by recent right responses), and America (PNAS), 110 (19), 7613-7618.
RT is also faster on a repetition trial preceded by recent
repetition trials or on an alternation trial preceded by recent References
alternation trials. This basic pattern is well fit by both a
generative Bayesian model (Wilder et al., 2009) and a Ng, A.Y., & Jordan, M.I. (2001). On Discriminative vs.
discriminative error-correction model (Jones et al., 2013). Generative classifiers: A comparison of logistic
The two models diverge in their predictions for how the two regression and naive Bayes. Advances in neural
learning mechanisms interact, with the error-correction information processing systems 14, 841-849.
model predicting cue-competition effects whereby the

26
Analogical Processes in Language Learning
Micah Goldwater (micah.goldwater@sydney.edu.au)
School of Psychology, University of Sydney, Sydney, New South Wales, Australia

Bozena Pajak (bpajak@northwestern.edu)


Department of Linguistics, Northwestern University, Evanston, IL, USA

Dedre Gentner (gentner@northwestern.edu) & Ruxue Shao (ruxueshao2018@u.northwestern.edu)


Department of Psychology, Northwestern University, Evanston, IL, USA

Discussant: Adele Goldberg (adele@princeton.edu)


Department of Psychology, Princeton University, Princeton, NJ, USA

depends heavily on overall similarity between primes and


Keywords: analogy; language learning; structure priming;
fast-mapping; phonetic categories; second language target; further, they show that priming with high-similarity
acquisition ‘easy’ primes renders children more likely to show priming
from purely syntactic matches. Both these findings are
Introduction directly parallel to work on analogical learning on
nonlinguistic tasks.
Language acquisition is a complex task, encompassing (at
Adele Goldberg will be a discussant.
least) perception and categorization of phonemes,
segmentation of speech, learning word meanings, and
extracting morphological and syntactic regularities. The Structural Priming as Analogical Mapping
daunting nature of this task might suggest that a specialized Micah Goldwater
module is required for language acquisition. Yet there is
increasing evidence that general learning processes play a We examined the development of syntactic knowledge in
major role (e.g., Marcus et al, 1999; Saffran, Aslin & children using a structural priming paradigm. Structural
Newport, 1996). In this symposium we present the case for priming refers to speakers’ tendency to match their syntax
analogical comparison processes in language learning. to that of a recent input sentence (Bock, 1986). It facilitates
Analogical comparison recruits a structure-mapping dialog in adults (Pickering & Garrod, 1998) and can also
process between two instances that highlights their common serve as a gauge of syntactic development in children.
relational structure—a critical feature in abstracting regular Goldwater and Echols provide evidence that children show
patterns across utterances. A further outcome of structure- structural priming when they are able to construct sentences
mapping is that alignable differences (differences that play via semantic and syntactic analogies from the utterances of
the same role in the matching structure) become salient, and others.
this can help learners notice key contrasts. Constructing analogies entails recognizing commonalities
The goal of this symposium is to show how individuals in the relational structure of two situations (or two
spontaneously use analogical reasoning in language sentences). Early in learning, such recognition typically
learning. We bring together empirical work addressing requires concrete similarity as well as relational similarity.
language acquisition in young children and second language To test whether children’s priming results from analogically
learners, across three different levels of linguistic structure: mapping a previous sentence’s structure, we engaged 4- and
phonology, lexical semantics, and syntax. 5-year-old children in a turn-taking scene-description task
B. Pajak will present work showing that learners infer using the typical measure of structural priming. That is, we
commonalities between observed phonetic contrasts in their asked whether children would describe a new picture with a
native language, and that this leads them to expect sentence matching the structure of a previous sentence,
analogous contrasts along the same dimensions when rather than using an equally correct sentence with a different
learning a new language. D. Gentner and R. Shao will how structure. For example, given the previous sentence
analogical processes help children learn new word meanings “Grandma handed Sally the cake,” when describing a new
with limited exposure. They revisit the classic Carey and picture the speaker would say “The teacher gave the boy a
Bartlett (1978) fast-mapping study and show that structural pencil” rather than “The teacher gave a pencil to the boy.”
alignment processes are critical for success. M. Goldwater We found, first, that young children showed syntactic
and C. Echols address the role of analogical processes in priming only when there were correlated concrete
learning constructions, using a structural priming paradigm. commonalities in the characters and events. This is
They show that structural priming in young children consistent with numerous findings showing that young

27
children typically recognize overall similarity before they used. The children chose between two trays that were
can recognize purely relational similarity. However, after identical except for color. According to structure-mapping
processing overall matches, in which structural alignment is theory, this should have promoted structural alignment,
supported by concrete similarities, children are often able to thereby highlighting color as an alignable difference. To test
process relational commonalities without the “training this, we gave 3- and 4-year-olds the classic fast mapping
wheels” of superficial similarity (Gentner, 2010). Consistent task, but varied the alignability of the materials. Children
with this pattern, young children showed structural priming saw two objects and were asked to “point to the chromium
for semantically dissimilar sentences only they had first one, not the blue one.” The High Alignability group saw
processed pairs of semantically similar sentences. two highly alignable alternatives, differing only in color (as
in Carey & Bartlett’s study). The Low Alignability group
Analogy in Learning Second-Language saw a pair that varied in both color and shape, making them
Phonetic Categories harder to align. Both groups accurately pointed to the
chromium object initially, but there was a large difference in
Bozena Pajak what they learned from this. When asked to identify new
chromium objects later, the high-alignability group far
Phonetic category acquisition is a complex problem of outperformed the low-alignability group. A second study
learning a mapping from variable phonetic tokens onto ruled out a purely informational account. These findings
discrete categories. How is this achieved? Prior suggest that structural alignment processes help children
experimental and computational work has identified two learn the word meanings from indirect linguistic
main sources of information available to and used by information.
learners, both infants and adults: statistical distributions of
sounds and lexical context. I will argue that, in addition to
those two sources of information, phonetic category
References
learning is supported by analogy-based abstraction: learners Bock, J. K. (1986). Syntactic persistence in language
infer commonalities between observed phonetic contrasts production. Cognitive psychology, 18(3), 355-387.
(e.g., /b/-/p/, /d/-/t/), which leads them to expect analogous Carey, S., & Bartlett, E. (1978). Acquiring a single new
contrasts defined along the same phonetic dimensions (e.g., word. Papers and Reports on Child Language
/g/-/k/). I will present a computational model of how such Development, 15. 17-29.
analogical reuse of categories might be achieved during Clements, G.N. (2003). Feature economy in sound
acquisition (Pajak, Bicknell, & Levy, 2013), and I will systems. Phonology, 20. 287–333.
support it with experimental evidence. In particular, I will Gentner, D. (2010). Bootstrapping the mind: Analogical
show that (1) the adult perceptual system is sensitive to non- processes and symbol systems. Cognitive Science,
native phonetic contrasts that are analogous to their native- 34(5), 752-775.
language contrasts (Pajak & Levy, 2014; Pajak, Piccinini, & Marcus, G., Vijayan, S., Rao, S., & Vishton, P.M. (1999).
Levy, in progress), and that (2) a brief exposure to novel Rule learning by seven-month-old infants. Science,
second-language phonetic categories leads adults to form 283, 77-80.
expectations about analogous categories in that language Pajak, B., & Levy, R. (2011). Phonological generalization
(Pajak & Levy, 2011). I will argue that analogical from distributional evidence. In L. Carlson, C.
abstraction can effectively bootstrap the acquisition of a Holscher & T. Shipley (Eds.), Proceedings of the
language's entire phonetic system given the typological 33rd Annual Conference of the Cognitive Science
evidence that languages tend to reuse the same phonetic Society (pp. 2673–2678). Austin, TX: Cognitive
dimensions for multiple contrasts (Clements, 2003). Science Society.
Pajak, B., & Levy, R. (2014). The role of abstraction in non-
native speech perception. Journal of Phonetics, 46,
Interactions between Structural Alignment and 147–160.
Pajak, B., Bicknell, K., & Levy, R. (2013). A model of
Language in Word Learning
generalization in distributional learning of phonetic
Dedre Gentner & Ruxue Shao categories. In Proceedings of the 4th Workshop on
Cognitive Modeling and Computational Linguistics
We propose that analogical processes are important in (pp. 11–20). Sofia, Bulgaria: Association for
allowing children to infer word meanings from indirect Computational Linguistics.
speech. Here, we revisit Carey & Bartlett’s (1978) classic Pickering, M. J., & Garrod, S. (2004). Toward a mechanistic
‘fast mapping’ study, in which 3-year-olds learned the psychology of dialogue. Behavioral and brain
meaning of a new color term (chromium) in a single sciences, 27(02), 169-190.
exposure, without direct reference. Children simply Saffran, J. R., Aslin, R. N., & Newport, E. L. (1996).
responded to the request “Give me the chromium one, not Statistical learning by 8-month-old infants.
the red one.” We suggest that a key component of the Science, 274, 1926-1928.
children’s success was the high alignability of the materials

28
The Relevance of Alternative Possibilities throughout Cognition
Jonathan Phillips (phillips01@g.harvard.edu)
Joshua Knobe (joshua.knobe@yale.edu)
Department of Philosophy, Yale University,
P.O. Box 208306, New Haven, CT 06520-8306 USA

Andrew Shtulman (shtulman@oxy.edu)


Department of Psychology, Occidental College,
1600 Campus Road, Los Angeles, CA 90041 USA

Charles Kalish (cwkalish@wisc.edu)


Anne Riggs (aeriggs@wisc.edu)
Department of Educational Psychology, University of Wisconsin-Madison
1025 W. Johnson Street, Madison, WI 53706 USA

Christopher Hitchcock (cricky@caltech.edu)


Division of Humanities and Social Sciences, California Institute of Technology
101-40 Caltech, Pasadena CA 91125 USA

these four presentations showcase new developments in the


Keywords: modality; counterfactuals; counterfactual
availability; norms; moral judgment; causal reasoning; emerging research on modal cognition and its relation to
developmental psychology norms.

Overview Phillips & Knobe: The Psychological


Research in a number of different fields has independently Representation of Modality
argued for the importance of providing a place for modality— A great deal of research has now demonstrated that our
that is, some way of representing alternative possibilities that understanding of physics, probability, and morality impact
could have happened, but actually did not (e.g., Kratzer, many aspects of cognition. One underappreciated fact about
2012; Lewis, 1973; Pearl, 2000). In each of these cases, the this research is that a judgment that something is statistically
key insight has been that people’s understanding of the things improbable often has the same impact on cognition as a
that occur is shaped in some central way by their judgment it is physically impossible or morally bad. The
understanding of these alternative possibilities. similarity of these effects can be seen in phenomena as
Work throughout these fields has emphasized that people diverse as causal selection, assessments of freedom,
do not treat all alternative possibilities equally. Instead, they counterfactual reasoning, predictions of future actions, and
regard certain possibilities as relevant, while treating others the development of thinking about possibilities.
as irrelevant (Portner, 2009; Roese, 1997). Within this We offer a unified account of this similarity by proposing
research, one consistent theme has been that norms that each of these factors is relevant to how people represent
(statistical, moral, conventional, etc.) influence how these possibilities. We lay out a modified version of a standard
alternative possibilities are represented. linguistic framework for modality (Kratzer, 2012), which
This symposium focuses on new empirical and theoretical allows us to capture the impact of these factors on cognition,
approaches to the role of modality throughout human and go on to report new empirical data that support this
cognition, and highlights the role of different norms in modal general account of the psychological representation of
cognition. Phillips and Knobe present a framework for the modality.
psychological representation of modality designed to capture Jonathan Phillips is a postdoctoral researcher in
the impact of factors such as probability and morality, and Psychology at Harvard University. Joshua Knobe is a
then go on to present new data in support of their proposed Professor in the Program in Cognitive Science and
approach. Shtulman discusses the development of modal Department of Philosophy at Yale University. Their work has
cognition, and reports empirical evidence that statistical and been published in Behavioral and Brain Sciences, Cognition,
moral norms affect beliefs about what is possible, permissible Cognitive Science, Psychological Science, Journal of
and real. Kalish presents new research on the modal Philosophy, and Semantics and Pragmatics.
judgments underlying children’s reasoning about norms.
Hitchcock combines research on the availability of
counterfactual alternatives in developing a framework that
accounts for ordinary judgments of causation. As a group,

29
Shtulman: Developmental and Individual Hitchcock: Counterfactuals, Norms, and
Differences in Modal Cognition Causal Judgment
Modal cognition underlies several facets of everyday There is a tradition in philosophy and legal theory of trying
learning and problem solving. In this talk, I will discuss the to understand causation in terms of counterfactuals. C is said
development of modal cognition, focusing on our changing to be a cause of E if E would not have occurred had C been
intuitions about physical possibility. The first half will absent. (See, e.g. Lewis 1973.) However, this account yields
outline the newly emerging consensus that children are more some verdicts that differ from the causal judgments of most
skeptical about physical possibility than are adults. Children subjects. Drawing on research done in collaboration with
initially deny the possibility of any event that defies Joseph Halpern (Cornell Computer Science), I present a
expectation, improbable or impossible, and not until early framework for explaining these discrepancies. According to
adolescence do they reliably differentiate events that violate a counterfactual account, causal judgment requires us to
physical laws from those that violate mere empirical consider what would happen in various hypothetical
regularities, both in their judgments and their justifications situations. Psychological research has shown that some
(Shtulman & Carey, 2007). The second half will explore the hypothetical possibilities are more readily available than
relation between modal judgments and modally-relevant others. Counterfactual availability is strongly influenced by
beliefs, namely, children’s beliefs about fantastical beings various norms, where the norms can be moral, legal,
(Shtulman & Yoo, 2015) and adults’ beliefs about moral statistical, or functional. (See, e.g. Kahneman and Miller
permissibility (Shtulman & Tong, 2012). Overall, I will argue 1986.) Thus, by drawing our attention toward or away from
that differences in the procedural aspects of modal judgment various possibilities, norms can highlight or obscure the
can lead to drastically different beliefs about what is possible, counterfactuals that underwrite particular causal judgments.
what is permissible, and what is real. Christopher Hitchcock is J.O. and Juliette Koepfli
Andrew Shtulman is an associate professor of psychology Professor of Philosophy at the California Institute of
at Occidental College. His interests include conceptual Technology. He works in the philosophy of science with a
development and conceptual change, and his work has focus on the role of causal concepts in scientific and everyday
appeared in such journals as Cognition, Cognitive reasoning. His articles have appeared in journal such as the
Psychology, and Cognitive Science. Journal of Philosophy, the Philosophical Review, Philosophy
of Science, and Cognitive Science.
Kalish: Why Not? Children’s Normative
Evaluations References
What sort of modality is involved in children’s normative Haidt, J. (2001). The emotional dog and its rational tail: A
evaluations? At times it seems that young children conflate social intuitionist approach to moral judgment.
physical and deontic possibility (e.g., denying that it is Psychological Review. 108, 814-834.
possible to violate social norms). There are many types of Kahneman, D. & Miller, D. (1986), Norm theory: comparing
constraints underlying social norms (e.g., prudence, error- reality to its alternatives. Psychological Review 80: 136 –
avoidance). One hypothesis is that children make normative 153.
evaluations by identifying the specific constraint relevant to Kratzer, A. (2012). Modals and conditionals: New and
an action (e.g., “That’s dangerous, so you can’t do it.”). We revised perspectives (Vol. 36). Oxford University Press.
will present data suggesting that young children treat Lewis, D. (1973). Causation. The Journal of Philosophy 70,
violations of conventional norms as wrong in and of 556-567.
themselves. When pressed, children will cite a constraint Pearl, J. (2000). Causality: Models, reasoning, and inference.
justifying their normative judgment, but the justifications New York: Cambridge University Press.
seem post-hoc (akin to the moral dumbfounding findings of Portner, P. (2009). Modality. New York: Oxford University
Haidt, 2001). For young children, social norms may Press.
determine a set of available and nonavailable actions, Roese, N. J. (1997). Counterfactual thinking. Psychological
without clearly specifying why. Bulletin, 121, 133–148.
Charles Kalish is a professor of Educational Psychology at Shtulman, A., & Carey, S. (2007). Improbable or impossible?
UW-Madison. His interests include the nature of normative How children reason about the possibility of extraordinary
concepts, and the role of norms in social cognition. His work events. Child Development, 78, 1015-1032.
on norms has appeared in such journals as Cognition, Child Shtulman, A., & Tong, L. (2013). Cognitive parallels
Development, and Developmental Psychology. Anne Riggs between moral judgment and modal judgment.
is a graduate student in the Department of Psychology at UW- Psychonomic Bulletin & Review, 20, 1327-1335.
Madison Shtulman, A., & Yoo, R. I. (2015, in press). Children’s
understanding of physical possibility constrains their belief
in Santa Claus. Cognitive Development.

30
Eye-tracking situated language comprehension: Immediate actor gaze versus
recent action events
Dato Abashidze (dabashidze@cit-ec.uni-bielefeld.de)

Pia Knoeferle (knoeferl@cit-ec.uni-bielefeld.de)

Maria Nella Carminati (mcarmina@techfak.uni-bielefeld.de)


Cognitive Interaction Technology Excellence Center, Department of Linguistics
Bielefeld University, Germany

Abstract gaze to objects which typically precedes mention of the ob-


jects. Following a speaker’s gaze is beneficial for the listener
Previous visual world eye-tracking studies have shown that
when a sentential verb can refer (via tense information on the since it permits him to anticipate which object the speaker
verb and on a following time adverb) to either a recent and will mention next (Hanna & Brennan, 2007; Knoeferle &
a future action event performed by an actor, people inspected Kreysa, 2012; Staudte, Crocker, Heloir, & Kipp, 2014).
the target of the recent event more often than the (different)
target of the future event. This ’recent event preference’ repli- How precisely do listeners rely on the many available lin-
cated even when the frequency of future events within the ex- guistic and extralinguistic cues? Existing account of sit-
periment greatly exceeded the frequency of recent events (e.g., uated language comprehension predict a rapid interplay of
75% vs 25%). The recent event preference may arise because
the past action is situation-immediate and thus more relevant at language comprehension, (visual) attenion, and visual con-
the particular point in time when the sentence is processed (at text effects (e.g., Altmann & Kamide, 2009; Knoeferle &
that point participants have seen the past action performed and Crocker, 2007). In the absence of specific evidence to the
will not see the future action until after the sentence). If the
situation-immediate relevance of a cue is responsible for the contrary, it would be tempting to predict that the various cues
recent event preference, then we should be able to “overwrite” are all on a par in contributing toward comprehension (all
the effect of the recent action with another situation-immediate else being equal). Alternatively, some cues and / or world-
cue. Accordingly, two current eye-tracking experiments pitted
the recent event preference against a situation-immediate cue, language relations may be preferred over others. Determining
the shift in the actor’s gaze to the target object. Given that inter- the relative contribution of different cues and world-language
locutors’ gaze has been shown to be a powerful cue in guiding relations is an important step in understanding situated lan-
listeners’ attention to objects in the visual context, we hypoth-
esized that the actor’s gaze to the future target should rapidly guage comprehension (Knoeferle, Urbach, & Kutas, 2014)
guide a listener’s attention to it. Analyses revealed indeed that
listeners’ visual attention was rapidly guided to the target by The recent-event preference vs. frequency biases
the actor’s gaze; crucially the gaze cue was particularly helpful
in guiding looks to the future target. Importantly, however, we Consider, for instance, a series of visual-world studies which
still replicated the overall preference to look at the recent target recorded comprehenders eye gaze in a scene as they lis-
regardless of tense and gaze; and even for future gaze condi-
tions, the preference was not immediately reversed, suggesting tened to spoken utterances (Abashidze, Knoeferle, Carmi-
it is surprisingly robust in competition with a situation-specific nati, & Essig, 2011; Knoeferle, Carminati, Abashidze, & Es-
future-biasing cue. sig, 2011). These studies all pitted two world-language re-
Keywords: Sentence comprehension, recent-event preference, lations agains one another, viz. relating a verb to recently-
actor gaze, eye tracking, visual world inspected action and its target compared with using the verb
to anticipate the target of a plausible future action. In the vi-
Introduction sual world study of Knoeferle et al. (2011) participants saw
Our immediate environment contains many extralinguistic an actor sitting in front of a table with two objects. First
cues that we exploit for understanding language. The infor- the actor performed an action on one object (e.g., sugaring
mation present in the visual context has been shown to pro- strawberries) and then participants heard either (Der Versuch-
vide powerful cues in guiding visual attention while we pro- sleiter zuckerte kürzlich die Erdbeeren, ’The experimenter re-
cess language. This rapid and to some extent predictive inter- cently sugared the strawberries’) or (Der Versuchsleiter zuck-
play between language and visual context is reflected in the ert demnächst die Pfannkuchen, ’The experimenter will soon
fact that people tend to gaze at objects as they are mentioned sugar the pancakes’). After the sentence had ended, partici-
(Spivey, Tanenhaus, Eberhard, & Sedivy, 2002; Tanenhaus, pants saw the actor performing the same action on the other
Spivey-Knowlton, Eberhard, & Sedivy, 1995) and often even object (i.e. sugaring pancakes). Analyses of participants’
anticipate their mention (Altmann & Kamide, 1999). For gaze record during sentence comprehension showed a pre-
instance, case marking and verb meaning can rapidly guide dominant preference of rapidly inspecting the target of the
a listener’s visual attention, permitting her to predict what recent event (i.e. strawberries) during and after the verb. This
will happen next (e.g. Kamide, Altmann, & Haywood, 2003; preference emerged even when the verb was in the present
Knoeferle & Crocker, 2007). In addition, listeners tend to tense and the adverb in the future sentence (denoting a fu-
rapidly rely on extralinguistic cues such as a speaker’s eye ture event), and it lasted well into the object noun phrase.

31
In further experiments Abashidze, Carminati, and Knoeferle recent (vs. future) events are anchored more firmly first in
(2014) used the same critical stimuli and design but increased working and then in short-term memory, participants should
the frequency with which during the experiment participants be better at recalling the target of the recent (vs. future)
saw future events (in association with a future sentence) rela- events. In addition in the gated test in Experiment 2 we might
tive to recent events (88% vs. 12% of trials). Even with this replicate the higher recall of past (vs. future) tense sentences.
strong frequency bias in favour of the future target, the overall Alternatively, the gaze cue has a strong influence on visual
preference to inspect the recent event target replicated. How- attention and the anchoring of events in working and short-
ever the frequency bias did modulate participants’ visual at- term memory. If so, then we should see better recall for the
tention in that the preferential inspection of the future event future than recent event target.
target started much earlier (by 1000 ms) than when recent and
future events were equally frequent in the experiment. Experiments 1 and 2: Methods
Participants
Gaze as a situation-specific cue
Thirty-two native speakers of German (aged 19 to 32, all stu-
Perhaps the short-term frequency bias, while modulating vi- dents of Bielefeld University) with normal or corrected-to-
sual attention, did not override the recent-event preference normal vision received 6 Euros each for their participation.
because the most recent cue in the context takes precedence. All gave informed consent.
Indeed, recency effects are well documented in research on
memory and cognition (Glanzer & Cunitz, 1966) and on vi- Materials and design
sual attention (Zelinsky, Loschky, & Dickinson, 2011).
We used the experimental items (N=24) from Abashidze et
One situation-specific cue which has been shown to rapidly al. (2014, see Table 1). All critical sentences had the struc-
guide a listener’s visual attention to upcoming, future ob- ture NP-V-ADV-NP and two male native German speakers
jects is the gaze of an interactant. For example a study by recorded them. The experimental sentences were always
Hanna and Brennan (2007) showed that listeners followed about two objects and presented in two tense conditions. In
the speaker’s gaze shifts to the target before it had been men- one condition, the verb was in the present tense with a time
tioned (for the robustness of this finding across different set- adverb (demnächst, ’soon’) indicating the future (Table 1a).
tings see also Staudte et al., 2014; Macdonald & Tatler, 2013; In the other condition , the verb was in the simple past, and
Knoeferle & Kreysa, 2012). If comprehenders incremen- the following time adverb (kürzlich, ’recently’) also indicated
tally update their visual anticipation based on recent and / the past (Table 1b). Only German regular verbs appeared in
or situation-immediate cues, then showing the actor shift his the critical sentences to ensure the verb was tense-ambiguous
gaze to the future action target during the verb should replace up to but excluding the word-final phoneme which disam-
the recent action as the most recent, situation-immediate cue. biguated towards the simple past in the past tense condition.
The present experiments Given that an interactant’s gaze As can be seen from Table 1, there were two sentences
has been shown to be a powerful situation-immediate cue in for each tense condition; this counterbalancing ensured that
guiding listener’s attention to soon-to-be-mentioned objects, each object was once the target of a past and once of a future
two eye-tracking studies pitted object-directed gaze against event. In turn, this ensured that visual and other characteris-
the recent-event preference. These studies assessed to what tics of any given post-verbal target object contributed equally
extent an actor’s gaze towards the (recent or future) action to each critical condition. The words in a sentence were
target influences listeners’ visual attention. Can the gaze to- matched for spoken syllables and lemma frequency within an
wards the future object overcome the preferred inspection of item (Baayen, Piepenbrock, & Gulikers, 1995).
the recent event target? Between the experiments, we var- For every item we recorded two videos (Mduration =5015
ied when the actor shifted his gaze (either during the verb or ms) showing a person sitting at a table in front of two ob-
slightly earlier at verb onset) to also examine the effects of jects (e.g. cucumbers and tomatoes, one on the left and one
cue onset. We used the design from Knoeferle et al. (2011, on the right). The first video showed the person performing
Experiment 2), described above, with the factor tense (past an action on one object (e.g., flavoring cucumbers, Fig. 1-1)
vs. future) and we added gaze (to the target object vs. straight and the second showed the person performing the same action
ahead) as a factor. The presence of gaze should trigger ear- on the other object (e.g., flavoring tomatoes, Fig. 1-3). The
lier and more frequent looks to the appropriate target objects position of the target objects (right vs. left) was counterbal-
(recent and future). Crucially, if the gaze cue overrides the anced across items. In addition, we created a short gaze video
recent-event preference, it should have a stronger influence in for each experimental item (for about 3-4 seconds, depending
sentences with future than past tense meaning (i.e. tense x on sentence length). The video showed the actor shifting his
gaze interaction). gaze to an appropriate object (e.g., in the future condition the
After the eye-tracking part participants completed a mem- actor would shift gaze to the tomatoes, see Fig. 1, 2a).
ory test (Expt 1) and a gated memory test (Expt 2). Abashidze The tense factor (see Table 1, 1a vs. 1b) was crossed with
et al. (2014) reported better recall of recent target objects. If actor gaze (target-directed vs. straight ahead, see Figure 1),
in the memory test for Experiment 1 we replicate this, and resulting in four experimental conditions. For the gaze condi-

32
gazed at the target object and in the other half he did not. The
Table 1: Example experimental sentences. The indices (’)
second factor was the tense: in 50% of trials the sentence was
indicate counterbalancing versions.
in the past tense and in other 50% of trials in the future tense
Tense Condition & see Table 1 for counterbalancing. The resulting 8 lists used
counterbalancing Sentences a Latin square design; each list contained every critical item
in only one condition and all fillers. Each participant saw an
1a future tense Der Versuchsleiter würzt demnächst
individually pseudo-randomized version of one of the eight
die Tomaten
experimental lists. A gaze detection pretest (N=20) confirmed
’The experimenter will soon flavor
that participants were able to quickly detect the gaze shift.
the tomatoes’
1a’ future tense Der Versuchsleiter würzt demnächst
die Gurken Figure 2: An example of a display for the memory test, Expt
’The experimenter will soon flavor 1 and Sequence of stages in the gated test, Expt 2
the cucumbers’ Expt 1
1b past tense Der Versuchsleiter würzte kürzlich
die Gurken
’The experimenter recently flavored
the cucumbers’
1b’ past tense Der Versuchsleiter würzte kürzlich
$%&'(#+#
die Tomaten Expt 2
$%&'(#)# $%&'(#*#

’The experimenter recently flavored !"#$%"#&'()&*"+,"#$-.#/$0$


1+,2$34)"$"56"#+7"8,"#$9:;<#$0=$
!"#$%"#&'()&*"+,"#$-.#/,"$>"#:?"$0$
34)"$"56"#+7"8,"#$9:;<#"?$#"("8,*@$0=$$
!"#
the tomatoes’ !"#$%"#&'()&*"+,"#$-.#/$0$
1+,2$34)"$"56"#+7"8,"#$9:;<#$0=$
!"#$%"#&'()&*"+,"#$-.#/,$?"78A()&,$0$
34)"$"56"#+7"8,"#$9:;<#&$&<<8$0=$$

Procedure
tion we used the videos showing the actor shifting gaze to the For the eye-tracking experiment, participants were calibrated,
target object (i.e., see Fig. 1, 2a). For the no-gaze condition and after a successful 9-point calibration of the eye tracker,
we created a snapshot from the last frame of the first video the experiment started. Participants were told to inspect the
showing the actor in a static position performing no action scenes and to listen carefully to the sentences. On a given
and looking straight ahead (i.e., see Fig. 1, 2b). Examples of trial, a participant first saw a video of a person performing one
the videos and the snapshot associated with the experimental action (e.g., flavoring the cucumbers, see Fig. 1, 1) and then a
sentences in Table 1 are shown in Figure 1 (1-3). static picture (the last frame of the video) appeared. After 700
ms, the sentence started, and in the gaze condition the actor
Figure 1: Sequence of events of a typical experimental trial started to shift his eye gaze to the target object 480 ms after
the verb onset in Experiment 1 and at verb onset in Experi-
ment 2. In the no-gaze condition the static picture remained
on the screen until 700 ms after the end of the sentence (see
Fig. 1, 2b). 700 ms after the sentence had ended, participants
saw a video of the actor performing the second action (e.g.,
flavoring the tomatoes, see Fig. 1, 3). Post-experiment, par-
ticipants took part in a simple memory test in Experiment 1
and a gated memory test in Experiment 2. Finally, they were
debriefed. Each experiment lasted approximately 50-55 min-
utes with a break after 25-30 minutes.
Memory tests
For the memory test in Experiment 1, we created two snap-
While we used the same 24 experimental sentences/videos shots of the first and second video of each experimental item,
in Experiments 1 and 2, the onset of the actor’s gaze shift i.e., showing the experimenter performing one of the two ac-
differed: In Experiment 1 this shift occurred on average 480 tions (see Fig. 2, Expt 1). The two snapshots associated with
ms after the onset of the verb and in Experiment 2 it occurred each item were combined into one display and shown to par-
at verb onset. Once he had shifted attention to the target, ticipants. Two versions were created in which the respective
the actor continued to fixate it until the end of the sentence. location of the two pictures was counterbalanced. Above the
In addition to the 24 experimental items we created 36 filler picture, one of two questions appeared:
sentences. These ensured that participants were exposed to a (a) Welche Aktion wurde VOR dem Satz durchgeführt?
range of other sentence structures and actions. ”Which action was performed before the sentence?”
Actor gaze was one factor: in half of the trials the actor (b) Welche Aktion wurde NACH dem Satz durchgeführt?

33
”Which action was performed after the sentence?” tense (past vs. future). We assessed the recent-event pref-
Participants responded with a button press (e.g., if they erence in two ways (see Knoeferle et al., 2011). First, we
thought that flavoring tomatoes was correct they would press tested the significance of the intercept overall (a positive in-
the right side button for the image in Fig. 2, Expt 1). tercept indicates a preference to inspect the recent action tar-
In Experiment 2 participants took a gated memory test get). Second, we assessed significance of the intercept by
which provided more detailed insights into recall of sentence condition (assessing effects of gaze and tense on the inspec-
content on a per-constituent basis. Figure 2 (Expt 2) shows tion preference).
an example sentence as presented in the memory test in a 3-
stage procedure. At the first stage, participants saw only the Figure 3: Mean log gaze probability ratios (ln (P(recent tar-
first noun phrase and the verb stem and had to verbally com- get/P(future target))) by condition from verb onset for Expts
plete the verb tense. The second stage added the temporal 1 and 2
adverb, and they had to recall the second noun phrase. If they 3.5  
Preference  for  recent  event  target   a  
were unable to do so, they received a further prompt at the

ln(P(recent  target)/P(future  target))  


2.5  
Average  gaze  shiC  
third stage and had to select the correct referent out of three 1.5  

objects. Two of these were from that sentence trial and the recent  -­‐  Gaze  
0.5  

third was a distractor from another filler item. 0   200   400   600   800   1000   1200   1400   1600   1800   2000   2200   2400   2600   2800   3000   3200  
recent  -­‐  No  Gaze  
future  -­‐  Gaze  
-­‐0.5  
future  -­‐  No  Gaze  

Experiment 1 and 2: Analyses and results -­‐1.5  

-­‐2.5  

Eye tracking We defined a period of interest from the on- -­‐3.5  

set of the verb until the offset of the post-verbal object noun 3.5  
b  
phrase. The measure of interest was fixations to the recent
ln(P(recent  target)/P(future  target))    
2.5  
Gaze  shiC  at  verb  
and future target objects (the cucumbers and the tomatoes) 1.5  
onset  

in the gaze and no-gaze conditions. We first computed gaze 0.5   recent  -­‐  Gaze  
recent  -­‐  No  Gaze  
probabilities to the two target objects in each successive 20 -­‐0.5  
0   200   400   600   800   1000   1200   1400   1600   1800   2000   2200   2400   2600   2800   3000   3200  
future  -­‐  Gaze  
future  -­‐  No  Gaze  
ms time slots. Because looks to these two entities are not -­‐1.5  

linearly independent (more looks to one object imply fewer -­‐2.5  


Preference  for  future  event  target  

looks to the other, and vice-versa), we computed mean log Verb  onset  0  ms   Adv  onset  1148  ms   NP2  onset  2480  ms   NP2  offset  3190  ms  
-­‐3.5  
Time  in  ms  
gaze probability ratios for the recent relative to the future tar-
get (ln (P (recent target)/P (future target))). A score of zero
indicates that both targets are fixated equally frequently; a Mixed effects ANOVAs (by participants and items) showed
positive score reflects a preference for looking at the recent the grand mean (i.e., the mean of all conditions) was posi-
target over the future target, and a negative ratio indicates the tive in all regions in both experiments, showing an overall
opposite (see Abashidze et al., 2014; Knoeferle et al., 2011). preference for the recent target (significant intercept in all the
We used the log-gaze probability ratio to plot the time ANOVAs by region in both experiments). Thus, the current
course from verb onset (Figure 3). In Figure 3 the blue lines experiments replicated the overall preference to look at the
indicate the recent condition (sentence in the past tense) and recent object independent of gaze cue and tense up to the
the green lines indicate the future condition (sentence in the very last sentential region. By contrast, pairwise comparisons
future tense). The dotted lines indicate the gaze condition revealed that gaze (vs. no gaze) enhanced looks to the fu-
and the solid lines indicate the no-gaze condition. As can ture target (p < .05) in the future conditions in Expt 1 and
been seen in these graphs the gaze cue had an early influence 2 in the Verb and Adverb regions, suggesting a mitigation
on target inspection (dotted lines), but only in the future con- of the recent-event preference in the future tense gaze con-
dition. In Experiment 1 participants started to preferentially dition. With regard to the manipulated factors, a significant
inspect the future target (negative ratio) in the future condi- effect of tense (all ps < .05) emerged in both experiments,
tion approximately 400 ms after the gaze shift. However in suggesting a reduction of the recent-event preference in the
Experiment 2 where the gaze shift occurred at the verb onset, future compared with the past tense condition. In addition,
ratios hover around 0 from a very early stage, suggesting the the gaze effect was fully significant in the Verb region in both
recent-event preference was eliminated. By contrast, in the experiments (in the adverb region by participants in Expt 1).
past-tense condition, the gaze manipulation did not affect the There was a Gaze x Tense interaction in all three regions in
distribution of attention until the end of the verb region (Expt Experiment 1, whereas in Experiment 2 the interaction was
1) and until towards the end of the adverb region (Expt 2, see significant by items at the verb and NP2 and marginal at the
blue lines. Fig 3a-b). adverb (p < .07) regions.
For the inferential analyses, the dependent variable was the Experiment 1: Memory test We calculated the percent-
mean log gaze probability ratio averaged over the word re- age of correct answers by condition for participants and items
gions (Verb, Adverb, NP2) by participants and by items, and separately. Figure 4a shows the average percentages (by par-
the independent variables were gaze (gaze vs. no gaze) and ticipant) with 76% of correctly answered questions. The

34
graph shows that participants were more accurate in recog- While participants overall preferred to inspect the recent
nizing the future (78%) than the recent (74%) target objects. event target (as indexed by reliable positive intercepts), repli-
In logistic linear mixed effect (LME) analyses the effect of cating the recent event preference (Abashidze et al., 2014;
target object was significant, the effect of gaze was marginal Knoeferle et al., 2011), the immediate gaze cue clearly af-
with worse accuracy for gaze than no gaze conditions. There fected participants’ visual inspection in the future tense con-
was a marginal interaction between object and gaze such that dition. The eye-tracking results in Experiment 1 did show a
gaze had a stronger effect for future than past tense sentences. rapid gaze cue effect especially in the future tense conditions
Figure 4: Percentage of correct answers by object and tense during which participants started to inspect the future target
(Expt 1); by tense and gaze (Expt 2) already at the end of the verb, 450 ms after the gaze cue on-
set (see green dotted line, Fig. 3a). This effect continued
100  
a  
Percentage  of  correct  answers  

90  
until the end of the sentence. By contrast, in the no-gaze fu-
80   ture condition participants preferentially inspected the recent
70   event target until the middle of the adverb region. The timing
60  
50  
of the gaze shift in relation to verb onset also affected mainly
40   Past   the future tense condition: in Expt 2 where the shift coincided
30   Future   with verb onset, the future target was looked at more and ear-
20  
10  
lier than when it occurred 450 ms after verb onset (Exp 1). In
0   fact, looks to the past and future target were balanced within
Gaze   No  Gaze   Gaze   No  Gaze   100 ms from the gaze shift in Expt 2 (see also for relevant
Future  object   Recent  object  
related results, Friesen & Kingstone, 1998).
100   In the past tense conditions, the gaze effect was less pro-
Percentage  of  correct  answers    

b  
90   nounced and less immediate than in the future tense sen-
80  
tences. However, a better way to characterize this is that the
70  
60  
inspection of the recent target during the verb was already so
50   robust that the additional gaze cue did not lead to a further en-
40   Past   hancement of looks to the recent target. Unlike in the future
30   Future  
20  
conditions, it was only towards the end of the verb region (in
10   Expt 1) and in the middle of the adverb (in Expt 2) that gaze
0   triggered more looks to the recent target in the gaze condition
gaze   no  gaze   gaze   no  gaze   gaze   no  gaze  
than in the no-gaze condition (see blue lines Fig 3a, b), this
1   2   3   difference lasting until the end of the sentence.
Since gaze strongly cues the mention of upcoming objects,
Experiment 2: Gated memory test We calculated the per- we might have expected a more immediate and full reversal
centage of correct answers by conditions for participants and of the recent-event preference at least in the future sentences.
items separately. The average percentages (by participant) However, even the very early effect of gaze in the future con-
are displayed in Figure 4b. Participants overall correctly an- dition in Expt 2 (Fig. 3b, green dotted line) did not lead to
swered 64% of the questions from all three stages. They were a sudden reversal of fixation preferences towards the future
more accurate at stage one (59%) than two (43%), and accu- target; in fact for the first 1400 ms the log ratio hovers around
racy was highest at stage three (with 90%) (see Fig. 4b, stages zero, suggesting strong competition from the recent target.
1, 2 and 3). The LME analyses for stage 3 showed an effect The post-experiment memory test in Experiment 1 did not
of tense (p < .003, higher accuracy for past than future tense completely agree with the eye-gaze data (recall was better for
conditions) and of gaze (p < .01, higher accuracy without future than past targets, against the recent event preference).
than with gaze), in the absence of an interaction. While gaze (vs no-gaze) was beneficial in enhancing atten-
tion to future targets, it did not enhance target recall (Fig. 4
Discussion a) but was, in fact, detrimental. Perhaps gaze is only used ’on
The current studies tested the recent-event preference the fly’ with short-lived effects on cognitive processes (see
(Abashidze et al., 2011; Knoeferle & Crocker, 2007; Knoe- Knoeferle & Kreysa, 2012). In the gated memory-test (Ex-
ferle et al., 2011; Abashidze et al., 2014) by pitting it against periment 2), past sentences were recalled better than future
another situation-immediate cue (the actor’s gaze). We as- ones (in agreement with the recent event preference, Fig. 4
sessed whether the recent-event preference replicates overall b). This provides some evidence for the view that past sen-
and whether participants would follow the actor’s gaze to the tences anchor an event better in memory than future ones.
future action target, thus effectively eliminating their prefer- We can compare our findings with the results of experi-
ence to inspect the recent action target when the verb was ments for which the design and frequency distribution was the
ambiguous between referring to a recent action (and its target same as for the no-gaze condition of the current experiments:
object) and a future action (and its different target object) In Experiment 2 by Knoeferle et al. (2011), the preference to

35
look at the recent object persisted until approximately 3000 Baayen, R., Piepenbrock, R., & Gulikers, L. (1995). The
ms after verb onset (when the ratio became negative). In the celex lexical database philadelphia: University of pennsyl-
current experiments we instead see a considerably earlier re- vania. Linguistic Data Consortium.
versal of the preference with the log ratio becoming negative Friesen, C. K., & Kingstone, A. (1998). The eyes have it!
at 1800 ms (Expt 1) and 2000 ms (Expt 2). Thus the fact that reflexive orienting is triggered by nonpredictive gaze. Psy-
the actor gazed at the targets in some trials, seems to have led chonomic Bulletin & Review, 5(3), 490–495.
to an earlier shift of attention to the future target even in the Glanzer, M., & Cunitz, A. R. (1966). Two storage mecha-
no-gaze future condition of the current experiments. nisms in free recall. Journal of Verbal Learning and Verbal
Compared with a strong short-term frequency bias to- Behavior, 5(4), 351–360.
wards future events (Abashidze et al., 2014), the situation- Hanna, J. E., & Brennan, S. E. (2007). Speakers eye gaze
immediate gaze cue had an earlier effect on the recent-event disambiguates referring expressions early during face-to-
preference. When the actor shifted gaze towards the future face conversation. JML, 57(4), 596–615.
target object, participants followed his gaze from 450 ms (in Kamide, Y., Altmann, G. T. M., & Haywood, S. (2003). The
Exp 1) and from 100 ms (in Expt 2) after its onset (see Fig 3, time course of prediction in incremental sentence process-
no clear recent-event preference and log-gaze probability ra- ing. JML, 49, 133–156.
tios hover around zero). By contrast, in the absence of a gaze Knoeferle, P., Carminati, M. N., Abashidze, D., & Essig, K.
cue when future events were more frequent than past ones (75 (2011). Preferential inspection of recent real-world events
to 88%), participants’ log-gaze probability ratio approached over future events: evidence from eye tracking during spo-
zero only approximately 1900ms after verb onset. ken sentence comprehension. Frontiers in Psychology, 2.
The present findings clearly shows that a situation- Knoeferle, P., & Crocker, M. W. (2007). The influence of
immediate cue modulated the recent-event preference earlier recent scene events on spoken comprehension: evidence
in the sentence than a short-term frequency bias towards fu- from eye-movements. JML, 75, 519–543.
ture events. However, even gaze did not immediately reverse Knoeferle, P., & Kreysa, H. (2012). Can speaker gaze modu-
the preference, speaking to its robustness. The conflicting late syntactic structuring and thematic role assignment dur-
memory-test results suggest we need further experiments to ing spoken sentence comprehension? Frontiers in Psy, 3.
assess the functional contribution of this attention preference. Knoeferle, P., Urbach, T. P., & Kutas, M. (2014). Differ-
It could reflect an epistemic bias whereby a recent event dom- ent mechanisms for role relations versus verb–action con-
inates attention more than assertions about a future event. gruence effects: Evidence from ERPs in picture–sentence
While a past event can generally be verified, a future one can- verification. Acta Psychologica, 152, 133–148.
not until it has actually occurred, and until then it is uncer- Macdonald, R. G., & Tatler, B. W. (2013). Do as eye say:
tain if it will happen (e.g. Staub & Clifton, 2011; MacFar- Gaze cueing and language in a real-world social interac-
lane, 2003). Another possibility is that it reflects attention tion. Journal of Vision, 13(4), 6.
to whichever object representation is most highly activated in MacFarlane, J. (2003). Future contingents and relative truth.
working memory. The Philosophical Quarterly, 53(212), 321–336.
Spivey, M. J., Tanenhaus, M. K., Eberhard, K. M., & Sedivy,
Acknowledgments J. C. (2002). Eye movements and spoken language compre-
This research was funded by the Cognitive Interaction Tech- hension: Effects of visual context on syntactic ambiguity
nology Excellence Center 277 and the SFB 673 “Alignment resolution. Cognitive Psychology, 45(4), 447–481.
in Communication” (German Research Foundation, DFG). Staub, A., & Clifton, C., Jr. (2011). Processing effects of
an indeterminate future: Evidence from self-paced reading.
References University of Massachusetts Occasional Papers in Linguis-
Abashidze, D., Carminati, M. N., & Knoeferle, P. (2014). tics, 38, 131–140.
How robust is the recent event preference? In Proceed- Staudte, M., Crocker, M. W., Heloir, A., & Kipp, M. (2014).
ings of the 36th Aannual Meating of the Cognitive Science The influence of speaker gaze on listener comprehension:
Society. Contrasting visual versus intentional accounts. Cognition,
Abashidze, D., Knoeferle, P., Carminati, M. N., & Essig, K. 133(1), 317–328.
(2011). The role of recent real-world versus future events Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K., &
in the comprehension of referentially ambiguous sentences: Sedivy, J. C. (1995). Integration of visual and linguistic
Evidence from eye tracking. In European Perspectives on information in spoken language comprehension. Science,
Cognitive Science: Proceedings of the EuroCogSci. 268, 632–634.
Altmann, G., & Kamide, Y. (1999). Incremental interpreta- Zelinsky, G. J., Loschky, L. C., & Dickinson, C. A. (2011).
tion at verbs: Restricting the domain of subsequent refer- Do object refixations during scene viewing indicate re-
ence. Cognition, 73(3), 247–264. hearsal in visual working memory? Memory & Cognition,
Altmann, G., & Kamide, Y. (2009). Discourse-mediation of 39(4), 600–613.
the mapping between language and the visual world: Eye
movements and mental representation. Cogn, 111, 55-71.

36
Effect of heaviness on the cognitive evaluation process
Keiga Abe (keiga.abe@gmail.com)
Department of Education, 1-1 Takakuwa-Nishi Yanaizu-Cho Gifu-Shi
Gifu, Japan

Abstract Heaviness and high-level cognition


This paper focuses on the sense of heaviness, because sense
The aim of this study was to clarify how the sense of heaviness of heaviness is related to body state. For example, when we
changes our cognition. According to recent studies in cogni-
tive science, intelligent human behaviors ranging from percep- hold a heavy object, we feel heaviness, change our posture,
tion to inference are not closed mental processes; rather, they and grow fatigued. These physical changes may alter cogni-
are affected by body and action (Wilson, 2002; Gibbs, 2005; tion.
Proffitt, 2006). In previous studies, the sense of heaviness
activated concepts metaphorically related to heaviness, and Seno, Abe, & Kiyokawa (2013) examined the effects of
changed impressions accordingly. However, previous studies heaviness on visually-induced illusory self-motion percep-
have not distinguished between subjective heaviness and phys- tion, also know as ”vection.” They hypothesized that heavier
ical weight. The purpose of this study was to clarify whether
changes in impressions are due to subjective heaviness or phys- items would inhibit vection because they make locomotion
ical weight. To examine this issue, a psychological experiment difficult. They found that wearing heavy clogs made locomo-
using a tasting task was conducted. The results confirmed that tion difficult and inhibited vection, suggesting that cognition
subjective heaviness influences evaluations of price and value.
can alter vection strength. Bhalla & Proffitt (2008) examined
Keywords: Embodied cognition; Size-weight illusion; Haptic perception under various physical conditions. They suggested
priming. that physical states affect people’s judgments about whether
they will be able to go up a slope or path. They showed that
Previous Studies people estimate uphill distance and steepness as being longer
and steeper when they are holding a heavy object and growing
Embodied cognition literature fatigued.
In haptic priming studies, ”heavy” is used as a metaphor
Research on embodied cognition suggests that mental activ- for ”important” or ”serious.” For example, Jostmann, Lakens,
ity is driven by physical body state, posture, and sensory- & Schubert (2009) showed that our abstract concept of im-
motor coordination. Perception can be modulated by bodily portance is affected by heaviness. They asked participants to
actions. For example, viewing visual stimuli between one’s judge importance in various situations while holding either a
legs changes visual perception compared to when such stim- heavy or light clipboard. Results indicated heaviness makes
uli are viewed normally (Higashiyama & Adachi, 2006; Hi- people invest more cognitive effort when engaging in abstract
gashiyama & Toga, 2011). thinking. Another study showed that curriculum vitae pre-
Bodily feedback from physical action can also change af- sented on heavier clipboards were judged to be more impor-
fective states and thoughts. Cacioppo, Priester, & Berntson tant than those presented on lighter clipboards (Ackerman,
(1993) suggested that arm extension gives rise to bodily feed- Nocera, & Bargh, 2010). These studies suggest that heaviness
back associated with avoiding negative stimuli, and arm flex- is associated with importance and seriousness. For example,
ion gives rise to bodily feedback associated with approaching we usually say ”a heavy penalty,” ”heavy responsibility,” and
positive stimuli. Friedman & Forster (2000, 2002) showed ”put more weight on.” It is thought that the sensory experi-
that arm extension and flexion bias participants toward differ- ence of heaviness activates these metaphorical concepts dur-
ent processing styles, which influences creative thinking. The ing haptic priming.
authors manipulated the extent to which non-affective bodily
feedback was associated with either positive or negative he- Hypothesis of this study
donic states, and then examined the effects of this feedback These previous studies have partially clarified the effects of
on cognitive processes related to creative insight. In the ex- heaviness on high-level cognition. Heaviness leads to longer
perimental social psychology literature, it has been suggested and steeper estimates of distance and slope, respectively.
that tactile sensations influence consumer behavior and social Sense of heaviness can also change subjective impressions
attitudes. For example, Krishna & Morrin (2008) showed that and social attitudes toward other people. Previous haptic
the perception of bottle hardness affected the evaluation of priming research found that sensory input activates metaphor-
natural water. Kay, Wheeler, Bargh, & Ross (2004) showed ical concepts.
that the tactile sensation of hardness made participants appear However, it is not clear whether subjective impressions are
more strict and stable, less emotional, and decreased negoti- influenced by subjective heaviness or physical weight. In this
ation flexibility. Embodied cognition research suggests that study, this issue was examined by addressing an estimation
mental activity is driven by physical body state, posture, and task. If the effects are due to the amount of the physical
sensory-motor coordination. load, it may be considered that physical/implicit processes

37
drink and evaluate them.” Participants were not informed that
in fact all cups contained the same natural water until after
the experiment. The three cups of water differed in the size
of the cup and the quantity of water (Figure 1). The small
cup (3 oz) contained 80 g of water. The medium cup (9 oz)
contained 40 g of water. The large cup (16 oz) contained 80
g of water. Participants were only told about the quantity of
water in the medium cup. The large cup contained the same
amount of water as the small cup, but because of the differ-
ence in cup size, it was expected that participants would think
that the large cup was lighter than the small one.

Procedure
A two-factor within-subjects design was used in this exper-
iment. The experiment was divided into three steps. First,
participants evaluated the cups of water, before they actually
Figure 1: The three cups of water in the experiment
drank, in order to test the effect of visual differences in cup
size on participants’ evaluation. Second, we verified that par-
which are separate from the subjective view of the subject ex- ticipants experienced the size-weight illusion. Third, the taste
ert effects on the inference. Conversely, if they are due to the test was conducted. After that, the taste and value of the water
amount of the subjective load, it may be considered that the were evaluated. In order to avoid the possibility that partici-
subjective view of the subject and explicit processes exert the pants guessed the purpose of this experiment, we told them:
effects on the inference. ”To make these cups easy to distinguish, we prepared three
As a means of examining these amounts of physical and sizes of cups.”
subjective loads separately, the ”size-weight illusion” (Char-
pentier, 1891) was used in this study. This illusion is the phe- Pre-test evaluation Participants first evaluated the cups of
nomenon that, if the weights are the same, the larger object water, before they actually drank from them, to test the effect
is sensed as being lighter. Utilizing this illusion, estimation of visual differences in cup size on evaluation. Participants
tasks under conditions of being subject to different subjec- were asked to evaluate how good tasting each cup of water
tive loads while being subject to the same physical load were looked, and how valuable each was, using a 101-point scale
conducted to examine the effects of the physical and objective (100 = good, 0 = bad). Then, they were asked to estimate the
amount of the physical load. price per 2 L of water.
Experiment
In this experiment, the effect of subjective heaviness on value Quantity estimation task After the evaluation task, partic-
judgment was examined using a water evaluation task. Par- ipants were handed the medium cup. The experimenter told
ticipants drank water and evaluated its taste, value, and price. the participants that the medium cup contained 40 g of water.
To examine the effect of differences in subjective heaviness, Participants were then asked to estimate the quantity of water
the size-weight illusion was used. Participants were asked to in the other cups to determine whether participants experi-
evaluate three cups of water. Two of the cups were the same enced the size-weight illusion for the small and large cups.
physical weight, but their subjective heaviness differed due Smaller quantity estimations for the large, compared to the
to the size-weight illusion. If subjective weight affects par- small, cup, indicate a size-weight illusion.
ticipants’ evaluations, there will be a significant difference in
evaluations between the cups of water. Post-test evaluation After the quantity estimation task,
Method participants were asked to drink from the cups of water and
evaluate the tastes. Participants were told to take a sip from
Participants each cup, and that they were not allowed to re-taste from any
Twenty college students participated in the experiment. The cup. They were not allowed to re-taste the water because this
IRB approval has been obtained. might change the weight of the cup and the quantity of water,
which might affect their evaluations. Participants evaluated
Task the water for taste and value. Finally, they decided the price
All participants were asked to drink three cups of water and per 2 L of water. If subjective heaviness affected evaluation,
evaluate them. The experimenter told participants: ”There are water in the small cup would be evaluated as better than the
three cups of water here. These correspond to any of the fol- water in the large cup. The order that participants drank each
lowing: tap water, natural water, and deep-sea water. Please cup of water was randomized to avoid order effects.

38
 J
100 Pre-test Post-test
*RRG
90
100
4XDQWLW\RIZDWHU

80 90
70 80
60 70

7DVWHVFRUH
50 60
40 50
30 40
20 30
10 20
10
0
0
/DUJH 0HGLXP 6PDOO %DG Large Medium Small
6L]HRIFXSV
6L]HRIFXSV

Figure 2: Means and standard errors the quantity estimation


task Figure 3: Means and standard errors of the taste evaluation
task

Debriefing After the experiment, the researcher asked par-


ticipants to describe the purpose of the experiment in order to *RRG 3UHWHVW 3RVWWHVW
determine if participants detected the aim of the experiment. 100
90
Then, the experimenter explained the aim of the experiment
80
to them.
9DOXH6FRUH

70
60
Results 50
According to the answers in the debriefing session, no partic- 40
ipants determined the aim of the experiment.First, to check 30
20
whether participants experienced the size-weight illusion, 10
quantity estimations were examined. Figure 2 shows the 0
mean quantity estimates. A one-way ANOVA revealed a %DG /DUJH 0HGLXP 6PDOO
main effect of size (F(2, 38) = 31.773, p < .01, partialη2 = 6L]HRI&XSV
.626). Multiple comparisons with the Bonferroni method re-
vealed a significant difference between each cup size (large-
small: p < .001; large-medium:p < .001; small-medium: Figure 4: Means and standard errors of the value evaluation
p < .001). The quantity of water in the small and large cups task
was physically the same, but participants thought that their
quantities differed. This confirms that participants experi-
main effect of cup size (F(2, 38) = 1.261, p = .295, n.s.). A
enced a size-weight illusion.
simple main effect test with Bonferroni method revealed there
were no significant differences in value evaluation during the
Taste ratings Figure 3 shows the mean taste ratings. A two- pre-test (F(2, 18) = .372, p = .695, n.s.). This result suggests
way ANOVA revealed no significant interactions (F(2, 38) = that the visual differences between the cups did not affect par-
.361, p = .695), but did reveal a marginally significant ticipants’ evaluations. In contrast, there was a significant dif-
main effect of pre- vs. post-test (F(1, 19) = 4.016, p = ference in post-test value evaluations (F(2, 18) = 4.141, p =
.060, partialη2 = .174), and a significant main effect of cup .033, η2 = .315). A simple main effect test revealed, with the
size (F(2, 38) = 3.356, p = .045, partialη2 = .150). A sim- Bonferroni method, that the large cup was rated significantly
ple main effect test using the Bonferroni method revealed that lower in value than the medium cup (p = .041). There was
there was a marginally significant difference between pre-test a marginally significant difference between large and small
and post-test taste evaluation in the medium size cup condi- cups (p = .080).
tion (p = .066).
Price decisions Figure 5 shows the mean water price es-
Value evaluation Figure 4 shows the mean value ratings. timates. A two-way ANOVA revealed a marginally sig-
A two-way ANOVA revealed a marginally significant inter- nificant interaction (F(2, 38) = 3.066, p = .058, partialη2 =
action (F(2, 38) = 3.251, p = .050, partialη2 = .146), and .139), and a marginally significant main effect of cup size
a significant main effect of pre- vs. post-test (F(1, 19) = (F(2, 38) = 2.687, p = .08, partialη2 = .124). A simple
4.464, p = .048, partialη2 = .190). There was no significant main effect test revealed that there were no significant dif-

39
tion of each cup, because the price decision is open-ended. In
㸦<HQ㸧 3UHWHVW 3RVWWHVW contrast, there were no significant differences in taste evalua-
400 tions, perhaps because the taste of water does not metaphori-
350 cally relate to heaviness. We do not express the taste of water
300 with abstract concept of heaviness.
250
3ULFH

200 Conclusions and future directions


150 This paper examined whether subjective heaviness affects
100 evaluations of value and price judgments. Differences in sub-
50 jective heaviness affected participants’ evaluation of value,
0 even though there was no difference in physical weight be-
/DUJH 0HGLXP 6PDOO tween the large and small cups. Subjective heaviness likely
6L]HRIFXSV
activated concepts metaphorically associated with value, as in
haptic priming.
Figure 5: Means and standard errors of the price decision task Future work should examine whether how heavy objects
are carried affects cognition. Previous haptic priming studies
differ in how participants carried objects. In Bhalla & Proffitt
ferences in price judgments during the pre-test (F(2, 18) = (1999), participants carried a heavy backpack on their backs,
1.334, p = .288, n.s.). This result suggests that the visual dif- but in Ackerman et al. (2010), participants were handed a
ferences between cups did not affect participants’ price deci- heavy clipboard.
sions. In contrast, there was a significant difference in price
Responsibility, pressure, and expectations are often associ-
judgments in the post-test condition (F(2, 18) = 3.760, p =
ated with heaviness on one’s back. For example, the phrase
.043, partialη2 = .295). A simple main effect test with the
”carry life’s burdens on one’s shoulder” indicates a respon-
Bonferroni method revealed that the water in the small cup
sibility for someone’s life. In contrast, acquirement, chance,
was given a significantly higher price than the water in the
and gain are associated with heaviness in the hands. For ex-
medium (p = .046) and large (p = .036) cups. The price
ample, the phrase ”to grab at the chance” means taking a fa-
of the water in the small cup was judged to be significantly
vorable opportunity. Thus, experiencing weight in different
higher in the post-test than in the pre-test.
body parts may activate different metaphorical concepts.
Discussion
The results of this experiment support the hypothesis that sub-
References
jective heaviness affects evaluations. There was a significant Ackerman, J. M., Nocera, C. C., & Bargh, J. A. (2010). In-
interaction and a main effect of pre- vs. post-test for value cidental haptic sensations influence social judgements and
evaluations. The price and value of the water in the small cup decisions. Science, 25, 1712–1715.
were rated significantly higher than these of the large cup in Bhalla, M., & Proffitt, D. R. (1999). Visual-motor recalibra-
the post-test evaluation. These results indicate that subjective tion in geographical slant perception. Journal of Experi-
heaviness information changed the evaluation of value. mental Psychology: Human Perception and Performance,
In the quantity estimation task, participants estimated the 25, 1076–1096.
weight of the small cup to be heavier than that of the large Cacioppo, J., Priester, J., & Berntson, G. (1993). Rudimen-
cup. Therefore, participants experienced the size-weight illu- tary determinations of attitudes. ii: Arm flexion and exten-
sion. Participants subjectively thought that the water in the sion have differential effect of attitudes. Journal of Person-
large size cup was lighter than that in the small size cup. ality and Social Psychology, 65, 5–17.
There was a significant interaction and a main effect of pre- Charpentier, A. (1891). Analyse experimentale: De quelques
vs. post-test for value evaluations. The price and value of elements de la sensation de poids. Arch Physiol Norm
the water in the small cup was rated significantly higher than Pathol, 3, 122-135.
that of the large cup in the post-test evaluation. These re- Friedman, R. S., & Forster, J. (2000). The effects of approach
sults indicate that subjective heaviness information changed and avoidance motor actions on the elements of creative
the evaluation of value. However, in the value evaluation, the insight. Journal of Personality and Social Psychology, 79,
water in the small cup was not significantly higher than that 477-492.
of the medium cup. It is possible that, because participants Friedman, R. S., & Forster, J. (2002). The influence of ap-
performed the value evaluation using a 101 scale, it was dif- proach and avoidance motor actions on creative cognition.
ficult to differentiate the rating of small the cup from that of Journal of Personality and Social Psychology, 38, 41-55.
medium cup, which had been already highly evaluated. On Gibbs, R. W. J. (2005). Embodiment and cognitive science.
the other hand, in the price decision task, evaluation of the New York: Cambridge University Press.
water in the small cup was evaluated as the most expensive. Higashiyama, A., & Adachi, K. (2006). Perceived size and
It is hypothesized that it is easy to differentiate the evalua- perceived distance of targets viewed from between the legs:

40
Evidence for proprioceptive theory. Vision Research, 46,
3961–3976.
Higashiyama, A., & Toga, M. (2011). Brightness and image
definition of pictures viewed from between the legs. Attent.
Percept. Psychophys, 73, 144–159.
Jostmann, N., Lakens, D., & W., S. T. (2009). Weight as
an embodiment of importance. Psychological Science, 20,
1169-1174.
Kay, A. C., Wheeler, S. C., Bargh, J. A., & Ross, L. (2004).
Material priming: The influence of mundane physical ob-
jects on situational construal and competitive behavioral
choice. Organizational Behavior and Human Decision
Processes, 95, 83-95.
Krishna, A., & Morrin, M. (2008). Does touch affect taste?
the perceptual transfer of product container haptic cues.
Journal of Consumer Research, 34, 807-818.
Proffitt, D. R. (2006). Embodied perception and the economy
of action. Perspectives on Psychological Science, 1, 110–
122.
Seno, T., Abe, K., & Kiyokawa, S. (2013). Wearing heavy
iron clogs can inhibit vecton. Multisensory Research, 26,
569-580.
Wilson, M. (2002). Six views of embodied cognition. Psy-
chonomic Bulletin and Review, 9, 625–636.

41
Harmonics co-occurrences bootstrap pitch and tonality perception in music:
Evidence from a statistical unsupervised learning model
Kat Agres (kathleen.agres@qmul.ac.uk)
Queen Mary, University of London
Department of Electronic Engineering and Computer Science
London E1 4NS, UK

Carlos Cancino, Maarten Grachten, Stefan Lattner (firstname.lastname@ofai.at)


Austrian Research Institute for Artificial Intelligence (OFAI)
Freyung 6/6, A-1010 Vienna, Austria

Abstract clustering or categorization. These abstracted representations


may also model human perception (Hinton, 2007; Bartlett,
The ability to extract meaningful relationships from sequences
is crucial to many aspects of perception and cognition, such 2001; Grachten & Krebs, 2014).
as speech and music. This paper explores how leading In music, unsupervised learning techniques have been used
computational techniques may be used to model how hu- effectively to learn feature representations for the harmonic
mans learn abstract musical relationships, namely, tonality relationships between keys (Leman, 1995) and tonal pitch
and octave equivalence. Rather than hard-coding musical
rules, this model uses an unsupervised learning approach to relationships within a key (Cancino, Lattner, & Grachten,
glean tonal relationships from a musical corpus. We de- 2014). The model proposed by Cancino et al. (2014) suc-
velop and test a novel input representation technique, using a cessfully replicates certain aspects of pitch perception, but
perceptually-inspired harmonics-based representation, to boot-
strap the model’s learning of tonal structure. The results are fails to replicate others, such as the perception of octave sim-
compared with behavioral data from listeners’ performance on ilarity (the perceptual similarity of tones one octave apart)
a standard music perception task: the model effectively en- displayed by musicians. This is likely due to the symbolic
codes tonal relationships from musical data, simulating expert
performance on the listening task. Lastly, the results are con- pitch input representation used, which fails to capture har-
trasted with previous findings from a computational model that monic relationship between tones. The current work uses a
uses a more simple symbolic input representation of pitch. novel harmonics-based input representation inspired by hu-
Keywords: Music perception; tonality; unsupervised learning; man pitch perception, with the hypothesis that the additional
Restricted Boltzman Machines information provided from lower resolved harmonics will
bootstrap both the perception of tonal relationships and oc-
Introduction tave similarity. The present research investigates this topic
Learning the rules and structure of sequential information is through the use of unsupervised statistical learning, and tests
of fundamental importance to human perception and cogni- the extent to which these methods are capable of modelling
tion, yet the process by which this occurs is still debated the perception of tonality, through the use of this more rich
and widely investigated across domains. In language, for ex- input representation.
ample, linguistic nativists posit that innate, domain-specific
mechanisms are responsible for grammar learning (e.g., Pitch Perception in Listeners
Berwick, Pietroski, Yankama, & Chomsky, 2011; Pinker, Arguably, the statistical properties of music (such as pitch
1994), while others argue that more general, statistical learn- occurrences and transitional probabilities between tones or
ing mechanisms underlie the induction of grammatical rules chords) enable its structure to be learned from exposure.
(Chater & Manning, 2006; Saffran & Wilson, 2003; Gomez For example, the transitional probabilities between musical
& Gerken, 1999). This debate has spread to other domains, events, and the frequency of occurrence of pitches in tonal
such as the perception of tonal music, which, like language, music, contribute to listeners perception of the hierarchical
is highly structured, and is governed by a set of grammatical relationship of pitches within a key (Smith & Schmuckler,
rules that can be described in music-theoretic terms (Lerdahl 2004). This is known as the “tonal hierarchy”, a phrase
& Jackendoff, 1983). Indeed, listeners’ ability to implicitly that highlights the relative stability or importance of certain
extract statistical regularities and knowledge of tonal rela- pitches in a musical key. In other words, due to the predom-
tionships has received much attention in recent years (Pearce, inance of some notes over others within a tonality (such as
2005; Saffran, Johnson, Aslin, & Newport, 1999). the tonic and fifth scale degree), certain notes are perceived
In an effort to model mechanisms for learning statisti- as belonging more or less to the key than others, and are con-
cal structure, unsupervised learning methods and Restricted sequently perceived as having different functional roles in the
Boltzman Machines (RBMs) have garnered enthusiastic sup- tonality. In the case of C Major, for example, the notes C and
port for examining questions of learning, feature represen- G (the tonic and fifth scale degree) have greater stability than
tations, and the probabilistic structure of (big) data. Once the leading tone (B, the seventh in the scale), or chromatic
an RBM has learned the properties of the given data, its la- pitches not in the key (e.g., F sharp).
tent (learned) feature spaces may be explored, and used for Discovery of the tonal hierarchy was the result of seminal

42
studies by Krumhansl and colleagues (Krumhansl, 1990) us- of musical-phrase boundaries. Although this data-driven ap-
ing a “probe tone paradigm”. In this task, listeners hear a mu- proach has been fairly successful, many statistical approaches
sical context that clearly establishes a key (such as an ascend- lack robustness (e.g., they do not capture an entire conditional
ing or descending scale), but is left incomplete (e.g., without probability distribution), resistance to noise, and flexibility
the final note of the scale). After this context, a subsequent regarding different prior contexts. To circumvent these is-
“probe tone” is played, and listeners rate how well the tone sues, an unsupervised RBM model is presently used to learn
completes the prior context, usually on a scale from 1 (“very the probabilistic structure of tonal music through repeated ex-
bad”) to 7 (“very good”) (Krumhansl & Shepard, 1979). The posure to a musical corpus.
results of probe tone tasks have repeatedly shown that dif- An advantage of RBMs over the Self-Organizing Maps
ferent pitches have different functions in the key. There is (SOMs) used in prior computational modeling approaches to
historical precedence for using human probe tone results as the perception of tonality (Leman, 1995; Tillmann, Bharucha,
a measure of model performance in music, and our computa- & Bigand, 2000) is that the learned representation space in
tional model follows this tradition. SOMs is typically 2- or 3-dimensional, whereas RBMs can
In addition to the statistical properties of music, the charac- learn spaces of arbitrary dimensionality. Low-dimensional
teristics of the acoustic signal also impact pitch and tonality space is convenient for visualization, but there are few
perception (McDermott & Oxenham, 2008; De Cheveigne, biologically-motivated reasons for enforcing learned repre-
2005). Pitch, the psychological perception of frequency, is sentations to be low-dimensional. Although RBMs are not
perceived in logarithmic relation to frequency. Whereas oc- claimed to be plausible models of neural structures, stacked
taves on the linear frequency spectrum become farther apart RBMs have been shown capable of learning biologically real-
the higher the absolute pitch, octaves are equally-spaced on istic receptive fields in vision (Lee, Ekanadham, & Ng, 2008).
the mel scale (such that doubling a frequency creates the per-
Perception-Based Input Representation Applications of
ception of a pitch one octave higher). There is some evidence
neural networks to music often start from symbolic rep-
that the perceptual similarity of pitches an octave apart is uni-
resentations of music, midi notes, or piano roll notation
versal and innate (Demany & Armand, 1984), and nearly all
(Cancino et al., 2014; Grachten & Krebs, 2014; Boulanger-
cultures base their musical scale on a one-octave range.
Lewandowski, Bengio, & Vincent, 2012). This usually im-
From a developmental perspective, given that most voices plies that pitch (octave-specific note name, e.g., ‘G4’) or
and instruments produce tones in which the fundamental pitch chroma (octave-invariant note name, e.g., ‘G’) are used,
pitch (F0) is much stronger than the partials, listeners may but this approach means losing potentially useful information
gradually build up pitch and tonal perception from weak indi- from harmonics that can ain in the extraction of tonal relation-
vidual harmonics. Empirical studies show that adults tend to ships. For example, human listeners perceive co-occurring
be more sensitive to tonal relationships and less influenced by harmonics for pitches that are an octave or a fifth apart; this
pitch proximity than children (Cuddy & Badertscher, 1987). consonance may help listeners develop abilities such as oc-
If greater perception of individual harmonics is gained over tave similarity perception and relative pitch. Therefore, we
the developmental trajectory, models using F0 as input may developed an input representation that could enable the RBM
better simulate children and novice listeners, while models to use harmonics to bootstrap tonal learning.
using harmonics information may reflect more experienced A harmonics representation provides the model with
listeners. richer input than using only note-names or fundamental
Because both low-level acoustic information and implicit pitches. Other computational approaches have represented
statistical learning mechanisms contribute to tonal perception even lower-level information; for example, autocorrelation
in listeners, the present research sought to model how the hi- temporal models (Licklider, 1951; van Noorden, 1982; Med-
erarchical perception of tonality may be learned through ex- dis & Hewitt, 1991; Cariani, 2001) have shown that neural in-
posure to music, utilizing an input representation inspired by terspike interval representations and their subharmonic repre-
the perception of pitch. sentations may potentially underlie the perception of pitch as
well as some basic aspects of tonality. Complementary to this
Computational approaches tradition, we endeavored to test whether resolved harmon-
Hard-coded, rule-based models can describe various cog- ics within the range of the piano (which covers the range of
nitive phenomena with notable accuracy, possibly captur- musical tonality) were sufficient to simulate listeners’ perfor-
ing some of the innate structure that constrains bottom-up, mance on a music perception task addressing the tonal func-
domain-general cognitive processing. Nevertheless, percep- tion of pitches within a key. While innate properties of the
tion reflects, to a substantial degree, what is learned based auditory system (e.g., neural spiking activity) may subserve
on experience. Accordingly, an emphasis has recently been representations of tonality, tonal perception is likely mediated
placed on investigating how features of data are learned by experience. We were therefore interested in whether dif-
from exposure. The development of such systems allows ferent input representations (harmonics vs F0s) would better
researchers to model perception without requiring user in- simulate listeners with varying degrees of musical expertise.
put or the pre-specification of rules. To this end, statistical When examining the perception of tonal structure, our har-
and probabilistic approaches have elucidated aspects of mu- monics representation has the advantage over audio-based
sic perception, such as tonal relationships and the perception representations (such as acoustic spectra computed from

43
tones) that it allows us to focus solely on the effect of coincid- Chain Monte Carlo technique that is well suited for energy
ing resolved harmonics between tones. When working with based models such as RBMs (Hinton, 2002).
acoustic spectra, this effect is blurred by phenomena like in- For this paper, we train a model with 100 hidden units for
harmonicity, and tone quality (timbre). It is beyond the scope 200 epochs, using a single Gibbs sampling step and a mini-
of this article to account for the effect of these phenomena batch size of 100. Different model parameters were explored,
on the perception of tonal structure. Thus, the following ap- such as the size of hidden layer and the amount of train-
proach employs an abstract representation based on human ing epochs. All hyperparameters (learning rate, momentum,
pitch perception, with the hypothesis that co-occurring har- number of steps of Gibbs sampling) were selected according
monics may scaffold the development of relative pitch and to the guidelines proposed by Hinton in (Hinton, 2012).
octave affinity found in musically-trained listeners.
Harmonics input representation
Method A distributed binary input vector was computed for every
Restricted Boltzman Machine model pitch of the piano keyboard, from A0 to C8, tuned in equal
The present research implemented a Restricted Boltzmann temperament. For each pitch the first four harmonics were
Machine, a generative stochastic neural network (Hinton, represented, comprising the fundamental frequency and three
2002). This model consists of a layer of visible units v ∈ Rn , successive harmonics for each pitch. The harmonic series was
which represent the observed data, and a layer of binary hid- computed by multiplying the pitch’s F0 by integer values (2
den units h ∈ {0, 1}l . Both layers form a bipartite graph, i.e. for the second harmonic, 3 for the third harmonic, etc). The
there are no connections between units from each layer. The four harmonics encoded represent the F0, an octave interval
joint probability distribution of v and h described by the RBM above the F0 (second harmonic), a fifth above the second har-
is given by monic (third harmonic), and two octaves above the F0 (fourth
harmonic). The harmonics for all 88 piano tones formed a
1 total of 112 frequency bins, which served as the 112 visi-
p(v, h) = exp (−E(v, h | θ)) , ble input nodes for the model. The binary input vector (visi-
Z
ble units) for each pitch encoded that pitch’s harmonics, i.e.,
where Z is a normalization term, and E(·) is an energy func- there were four “on” nodes in each input vector.
tion, usually a quadratic function of the visible and hid-
den units. This energy function is proportional to the log- Training corpus
likelihood function of the model parameters θ given the vis- Our training corpus consisted of the entire set of 48 fugues
ible and hidden units, and its name was inspired by the Ising from J.S. Bachs Well-Tempered Clavier, regarded as one of
model from statistical thermodynamics. For the standard the most seminal works of classical music. Previous com-
Bernoulli-Bernoulli RBM1 , the energy function is putational modeling shows that the representations derived
from this corpus reflect the “Circle of Fifths”; in other words,
E(v, h | W, a, b) = −vT a − hT b − vT Wb,
the statistics of this corpus yield meaningful relationships be-
where θ = {W, a, b}, with W ∈ Rn×l a weight matrix, a ∈ Rn tween the musical keys (Cancino et al., 2014). Because the
a bias vector for the visible units, and b ∈ Rl a bias vector for fugues span every key and therefore have different distribu-
the hidden units. tions of pitch occurrences, they were all transposed to the
The free energy (FE), denoted by F (v), is a measure of key of C. Transposing or otherwise accounting for key (e.g.,
the expectancy of an input (visible) configuration, since it is by representing scale degree and pitch interval) is common
proportional to the expected value of the conditional proba- practice for training computational models on tonal corpora
bility of the visible units given all possible configurations of that span different keys. Without transposition, the statistics
the hidden units, i.e. defining tonal relationships from different keys will provide
conflicting information to a model that uses absolute pitch
F (v) ∝ − log (E {p(v | h)}) . representation. Each fugue was decomposed into its con-
stituent voices (two to five per fugue), where “voices” refers
Model training to the number of parts in the musical score. Voices in the bass
register were moved to the C3 to C6 range to enable their
The model parameters θ are optimized to maximize the ex- tonal information to be used and integrated by the model.
pected log-likelihood of the observed data. In the machine This yielded a total of 166 voices used for training, and each
learning literature, this optimization process for neural net- voice was considered as a single monophonic melody in the
works is known as training (Bishop, 1995). The standard corpus.
method for training RBMs is known as Contrastive Diver-
The set of voices were converted from their original MIDI
gence, proposed in (Hinton, 2002). In this gradient-descent-
format into the harmonics representation described above (ev-
like algorithm, the gradient of the log-likelihood of the ob-
ery pitch was replaced by its binary harmonics vector). The
served data is approximated using Gibbs sampling, a Markov
RBM was then trained on n-grams of these harmonics vec-
1 For more details on the derivation of energy functions for sev- tors, where an n-gram is defined as a successive set of n tones
eral RBM architectures see (Cancino, Lattner, & Grachten, 2015) in the corpus. N-grams were each eight notes long, and were

44
created by means of a sliding window (e.g., for a particular rience. The model results were therefore compared to highly
melody, notes 1-8 formed the first n-gram, followed by notes trained musicians (experts) and musically-untrained listeners
2-9 for the second n-gram, etc). This n-gram length was cho- (novices). We refer the reader to this paper for further details
sen to allow for the presentation of a seven tone stimulus plus regarding the study.
a single probe tone to the model, as is necessary for compari-
son with human ratings on a probe tone test (see the next sec- Results and Discussion
tion on Model Evaluation). Moreover, Cancino et al. (2014) The performance of the model, as assessed by the FEs of the
found that a minimum of eight notes in an n-gram was nec- probe test stimuli, was compared with average probe tone rat-
essary for optimal categorization of the n-gram in terms of ings by expert and novice listeners (Krumhansl & Shepard,
tonal key. The 8-grams were presented in randomized order 1979). We were interested in comparing the model with the
to provide more robust training for the model. Note that the pattern (or profile) of human responses across probe tones,
RBM computes the probabilities of the elements in each in- but the original variance data of listeners’ responses is no
put vector (the set of visible nodes that encode the input, in longer available, which precludes statistical significance test-
our case, the set of eight pitches), not the probability of a se- ing. Therefore, to compare the patterns of results, we cal-
quence of n-grams. As such, there is no temporal aspect with culated the Kullback-Leibler (KL) divergence (Kullback &
regard to the order of training instances themselves; rather, Leibler, 1951) between the two sets of data, which measures
each n-gram is treated as another time-invariant training in- the distance between two discrete distributions, p(1) and p(2) .
stance. The RBM extracts meaningful relationships between KL divergence was then used as the kernel for a distance-
the pitches within, and not between, each training n-gram. based Similarity measure (Shepard, 1987) that is used to
quantify the similarity between the two distributions:
Model evaluation
    
After training the RBM, the model’s internalization of the Similarity p(1) | p(2) = exp −DKL p(1) | p(2) .
tonal pitch hierarchy was tested. To this end, we implemented
a Krumhansl-style probe tone test: The model was given ei- This similarity measure has the property of being exactly
ther an ascending scale (the octave from C3 to C4) or a de- one if both distributions are identical, and tends asymptoti-
scending (from C6 to C5), without the final C to complete the cally to zero if the KL divergence between both distributions
octave. This musical context was immediately followed by goes to infinity. Similarity (e−KL ) values and Pearson cor-
a probe tone which was selected from the chromatic pitches relations between model and human ratings are provided in
between C4 and C5 (see Figure 1). Table 1 for an RBM tested on probe stimuli with ascending
scale and descending scale contexts.

Ascending Descending Table 1: Comparison of expert and novice listeners’ probe


tone ratings (for ascending or descending stimuli) with an
RBM model tested on ascending or descending scale con-
Probe tones texts. The highest Similarity values are in bold for both of
the model test conditions.
Figure 1: Ascending and descending C major scale context,
and the set of possible chromatic probe tones. Asc model context Desc model context
Expertise R Similarity R Similarity
Expert (Asc) 0.82 0.88 0.72 0.57
To provide these stimuli to the model for testing, we con- Expert (Desc) 0.83 0.84 0.83 0.88
structed n-grams of length 8, each of which contained the Novice (Asc) 0.59 0.75 0.00 0.42
seven pitches from the ascending or descending scale, fol- Novice (Desc) 0.54 0.52 0.75 0.54
lowed by a probe tone. This yielded a test set of 26 stimuli.
The Free Energy (FE) was calculated for each of these probe Given the model’s results for ascending test stimuli, the KL
test stimuli, and then normalized and scaled for comparison divergence is lowest (i.e., the distributions were least differ-
with human ratings. ent), and the Similarity is greatest, for Expert listeners’ rat-
The model’s performance was compared with that of lis- ings of ascending probe stimuli. In other words, the model
teners for both ascending and descending scale stimuli, as re- reflects expert listeners’ behavioral results for this set of stim-
ported in Krumhansl and Shepard (1979). This classic study uli. The RBM results are most highly correlated with expert
was chosen because 1) the probe tone context featured scales listeners for both ascending and descending stimuli.
rather than chords, 2) tones containing harmonics were used These findings are mirrored by the descending stimuli re-
(as opposed to pure tones, or Shepard tones as in Krumhansl sults. For these test stimuli, the KL divergence is lowest and
and Kessler (1982)), and 3) listeners with different levels of the Similarity is highest for Expert listeners who rated de-
training were tested. This last point enabled us to test the scending probe stimuli. The RBM results are most highly cor-
hypothesis that this richer input representation will allow the related with descending ratings from Expert listeners. Once
model to better simulate listeners with greater musical expe- again, the model best reflects expert listeners’ results when

45
the two are compared on the same set of stimuli, and further on learning the probabilistic structure of sequential informa-
demonstrates the stimulus-specific response of the RBM. tion, contribute to the acquisition of abstract, high-level rela-
The comparisons between RBM Free Energy results and tionships in music.
expert listeners’ ratings are plotted in Figure 2 for visualiza- Our novel representation of musical input was inspired by
tion. These graphs illustrate that the RBM was able to model how human listeners process pitch, and this method takes our
the hierarchical tonal relationships exhibited by listeners: The model one step closer to an embodied approach to modeling
model learned the privileged role of diatonic pitches in the C music cognition. Future work will investigate using the en-
major scale, and exhibits a degree of octave similarity. tire frequency spectrum of every tone (e.g., as sampled from
audio recordings). The full spectrum of pitch information
Ascending
7
may result in even better model performance on pitch-related
RBM free energy Expert listener rating

6
tasks, especially with regard to octave equivalence.
5 As hypothesized, a harmonics-based representation assists
4 the model in learning the tonal hierarchy and octave similarity
3 from pitches that share harmonics. The model best simulated
2 expert listeners, which can be taken as evidence that trained
1
musicians likely take advantage of the harmonic spectrum of
C4 C#4 D4 D#4 E4 F4 F#4 G4 G#4 A4 A#4 B4 C5
musical pitches in order to (implicitly) perceptually organize
Descending
the pitches within a key. Our findings may also support the
7 RBM free energy Expert listener rating

6
claim that novice listeners focus more on fundamental fre-
5
quency and pitch proximity than harmonics. More generally,
4 these findings highlight how the choice of representation can
3 have a notable impact on learned features, and that alternative
2 representations may be used to simulate different populations.
1
An extension of this work will consider using representa-
C4 C#4 D4 D#4 E4 F4 F#4 G4 G#4 A4 A#4 B4 C5
tions based on subharmonic patterns, as these are consistent
with temporal models of pitch perception (Cariani, 2001)).
Figure 2: RBM Free Energy results compared with average In addition, superior model performance may result from us-
probe tone ratings by expert listeners for ascending and de- ing stacked RBMs, a method currently popular in the area of
scending scale contexts. deep learning, as additional layers (model depth) may allow
the model to learn increasingly abstract features of tonal rela-
These results were also compared to a version of the RBM tionships.
model that was instead trained on MIDI pitch with no har-
monics, as reported in Cancino et al. (2014), which was Acknowledgments
configured to have the same parameters and hyperparame- This research was made possible through support from the
ters as the model discussed above. This pitch-only model, European Commission. The Lrn2Cre8 and PROSECCO
trained to the same number of epochs, yielded worse perfor- projects acknowledge the financial support of the Future and
mance when compared with listeners. The highest Similarity Emerging Technologies (FET) programme within the Sev-
value for ascending contexts was 0.58 (for non-expert listen- enth Framework Programme for Research of the European
ers rating descending stimuli). The highest Similarity value Commission, under FET grant number 610859 (Lrn2Cre8)
for descending contexts was 0.76 (with novice listeners rating and FET grant number 600653 (PROSECCO).
descending stimuli). The greater similarity to untrained lis-
teners rating descending contexts may reflect the prevalence References
of C4 over C3 in the corpus. Compared to an RBM model
using only a local binary pitch representation, the harmon- Bartlett, M. S. (2001). Face image analysis by unsupervised
ics representation yields better overall results. Also, whereas learning. Springer Science & Business Media.
the harmonics representation best models expert listeners, the Berwick, R. C., Pietroski, P., Yankama, B., & Chomsky, N.
pitch-only representation reflects less experienced listeners. (2011). Poverty of the stimulus revisited. Cognitive Sci-
ence, 35(7), 1207–1242.
Conclusion Bishop, C. M. (1995). Neural networks for patter recogni-
In this paper, we use unsupervised learning techniques to train tion. Claredon Press, Oxford.
a computational model on pitch relationships from a corpus Boulanger-Lewandowski, N., Bengio, Y., & Vincent, P.
of fugues from Bach’s Well Tempered Clavier. Our approach (2012). Modeling temporal dependencies in high-
allows the RBM to learn musical structure (i.e., tonal rela- dimensional sequences: Application to polyphonic music
tionships) from a training corpus without having to hard-code generation and transcription. In Proceedings of the 29th
tonal rules into the model. In fact, the high correlations be- international conference on machine learning.
tween the model and listeners’ performance lend support to Cancino, C., Lattner, S., & Grachten, M. (2014). Devel-
the claim that domain general processing mechanisms, based oping tonal perception through unsupervised learning. In

46
Proceedings of the 15th international society for music in- Kullback, S., & Leibler, R. A. (1951). On information and
formation retrieval conference. sufficiency. The Annals of Mathematical Statistics, 22(1),
Cancino, C., Lattner, S., & Grachten, M. (2015). Deriva- 79-86.
tions of the free energy of restricted boltzmann machines Lee, H., Ekanadham, C., & Ng, A. Y. (2008). Sparse deep
(Technical Report). Austrian Research Institute for Artifi- belief net model for visual area V2. In Advances in neural
cial Intelligence. information processing systems 20 (pp. 873–880).
Cariani, P. (2001). Temporal codes, timing nets, and music Leman, M. (1995). A model of retroactive tone-center per-
perception. Journal of New Music Research, 30(2), 107– ception. Music Perception.
135. Lerdahl, F., & Jackendoff, R. (1983). A generative theory of
Chalnick, A., & Billman, D. (1988). Unsupervised learning tonal music. MIT Press.
of correlational structure. In Proceedings of the tenth an- Licklider, J. (1951). A duplex theory of pitch perception.
nual conference of the cognitive science society (pp. 510– The Journal of the Acoustical Society of America, 23(1),
516). Hillsdale, NJ: Lawrence Erlbaum Associates. 147–147.
Chater, N., & Manning, C. D. (2006). Probabilistic models Matlock, T. (2001). How real is fictive motion? Doctoral
of language processing and acquisition. Trends in cognitive dissertation, Psychology Department, University of Cali-
sciences, 10(7), 335–344. fornia, Santa Cruz.
Cuddy, L. L., & Badertscher, B. (1987). Recovery of the McDermott, J. H., & Oxenham, A. J. (2008). Music per-
tonal hierarchy: Some comparisons across age and levels ception, pitch, and the auditory system. Current opinion in
of musical experience. Perception & Psychophysics, 41(6), neurobiology, 18(4), 452–463.
609–620. Meddis, R., & Hewitt, M. J. (1991). Virtual pitch and phase
De Cheveigne, A. (2005). Pitch perception models. In Pitch sensitivity of a computer model of the auditory periphery. i:
(pp. 169–233). Springer. Pitch identification. The Journal of the Acoustical Society
Demany, L., & Armand, F. (1984). The perceptual reality of of America, 89(6), 2866–2882.
tone chroma in early infancy. The journal of the Acoustical Newell, A., & Simon, H. A. (1972). Human problem solving.
Society of America, 76(1), 57–66. Englewood Cliffs, NJ: Prentice-Hall.
Feigenbaum, E. A. (1963). The simulation of verbal learn- Ohlsson, S., & Langley, P. (1985). Identifying solution paths
ing behavior. In E. A. Feigenbaum & J. Feldman (Eds.), in cognitive diagnosis (Tech. Rep. No. CMU-RI-TR-85-2).
Computers and thought. New York: McGraw-Hill. Pittsburgh, PA: Carnegie Mellon University.
Pearce, M. T. (2005). The construction and evaluation of
Gomez, R. L., & Gerken, L. (1999). Artificial grammar learn-
statistical models of melodic structure in music perception
ing by 1-year-olds leads to specific and abstract knowledge.
and composition. Unpublished doctoral dissertation, City
Cognition, 70(2), 109–135.
University London.
Grachten, M., & Krebs, F. (2014). An assessment of learned
Pinker, S. (1994). The language instinct: The new science of
score features for modeling expressive dynamics in music.
language and mind (Vol. 7529). Penguin UK.
IEEE.
Saffran, J. R., Johnson, E. K., Aslin, R. N., & Newport, E. L.
Hill, J. A. C. (1983). A computational model of language ac-
(1999). Statistical learning of tone sequences by human
quisition in the two-year old. Cognition and Brain Theory,
infants and adults. Cognition, 70(1), 27–52.
6, 287–317.
Saffran, J. R., & Wilson, D. P. (2003). From syllables to syn-
Hinton, G. E. (2002, July). Training products of experts by tax: Multilevel statistical learning by 12-month-old infants.
minimizing contrastive divergence. Neural Computation, Infancy, 4(2), 273–284.
14(8), 1771–1800. Shepard, R. N. (1987). Toward a universal law of generaliza-
Hinton, G. E. (2007). Learning multiple layers of represen- tion for psychological science. Science, 237, 1317-1323.
tation. Trends in cognitive sciences, 11(10), 428–434. Shrager, J., & Langley, P. (Eds.). (1990). Computational
Hinton, G. E. (2012). A practical guide to training re- models of scientific discovery and theory formation. San
stricted boltzmann machines. Neural Networks: Tricks of Mateo, CA: Morgan Kaufmann.
the Trade. Smith, N. A., & Schmuckler, M. A. (2004). The perception of
Krumhansl, C. L. (1990). Cognitive foundations of musical tonal structure through the differentiation and organization
pitch (Vol. 17). Oxford University Press New York. of pitches. Journal of experimental psychology: human
Krumhansl, C. L., & Kessler, E. J. (1982). Tracing the dy- perception and performance, 30(2), 268.
namic changes in perceived tonal organization in a spa- Tillmann, B., Bharucha, J., & Bigand, E. (2000). Implicit
tial representation of musical keys. Psychological review, learning of tonality: a self-organizing approach. Psychol.
89(4), 334. Rev., 107, 885.
Krumhansl, C. L., & Shepard, R. N. (1979). Quantification van Noorden, L. (1982). Two channel pitch perception. In
of the hierarchy of tonal functions within a diatonic con- Music, mind, and brain (pp. 251–269). Springer.
text. Journal of experimental psychology: Human Percep-
tion and Performance, 5(4), 579.

47
Do potential past and future events activate the Lateral Mental Timeline?
Roberto Aguirre (raguirre@psico.edu.uy)
Centre of Basic Research in Psychology. Av. Tristan Narvaja, 1674, 11200-Montevideo, Uruguay

Julio Santiago (santiago@ugr.es)


Mind, Brain, and Behavior Research Center. University of Granada. Campus de Cartuja s/n, 18071-Granada, Spain

Abstract 2014) or whole sentences (Ulrich & Maienborn, 2010). Still


Current evidence provides support for the idea that time is others have used sequences of events which can be objectively
mentally represented by spatial means, as a lateral mental time placed in temporal succession (Furhman & Boroditsky, 2010;
line. However, available studies have tested only factual events, Santiago et al, 2010).
i.e., those which have occurred in the past or will occur in the The aim of the present research is to test whether potential
future. In the present study we tested whether past and future
potential events are also represented along the lateral mental
events are also able to activate the left-right mental timeline.
timeline. In Experiment 1 participants categorized the temporal To our knowledge, no prior study has tapped onto this
reference (past or future) of either factual or potential events and question. The ability to represent potential events is central to
responded by means of a lateralized (left or right) keypress. human cognition. Representing future potential events
Factual events showed a space-time congruency effect that supports the manipulation of alternative scenarios and the
replicated prior findings: participants were faster to categorize evaluation of their consequences in order to make decisions
past events with the left hand and future events with the right about courses of action (Baumeister & Masicampo, 2010;
hand than when using the opposite mapping. More importantly, Hegarty, 2004; Johnson-Laird, 1983). Past potential events are
this also ocurred for potential events. Experiment 2 replicated a necessary component of counterfactuals (e.g., “If I had been
this finding using blocks comprising only potential events. In
order to assess the degree of automaticity of the activation of the
your father, I hadn’t allowed you to do it”; (see Gilead,
mental timeline in these two kinds of events, Experiment 3 asked Liberman, & Maril, 2012), and they are directly related to
participants to judge whether the expressions referred to factual studies of the processing of negation (as any potential past
or potential events. In this case, there was no space-time event is something that did not happen). The mental
congruency effect, showing that the lateral timeline is active representation of uncertain and negated events has recently
only when relevant to the task. Moreover, participants were received strong interest from the perspective of embodied
faster to categorize potential events with the left hand and factual approaches to language comprehension (De Vega et al, 2014;
events with the right hand than when using the opposite Ferguson, Tresh & Leblond, 2013; Kaup et al, 2007; Orenes,
mapping, suggesting for the first time a link between the mental Beltrán & Santamaría, 2014). If comprehension is mediated by
representations of space and potentiality.
detailed, modal mental simulations of linguistic content,
Keywords: Mental timeline; time; space; potentiality; factuality. uncertain and negated events pose an important theoretical
challenge.
A large number of studies support the suggestion by Lakoff Prior research has shown that the mental simulations of
and Johnson (1980) that space is used to conceptualize time. concrete factual events activate a lateral mental timeline. The
Among other possibilities, time can be represented as flowing present study will shed light on whether potential events are
from left to right in space, at least in languages with a left-to- also mentally arranged along a left-right axis. In order to
right orthography (see Santiago, Lupiáñez, Pérez, & Funes, answer this question, the present study used a standard space-
2007, for Spanish; Tversky, Kugelmass, & Winter, 1991, for time conceptual congruency task along the lines of Santiago et
English; Ulrich & Maienborn, 2010, for German. Santiago et al (2007). In Experiment 1, factual past and future events were
al (2007) presented words referring either to the future or to mixed with potential past and future events. Events were
the past, and participants categorized their temporal reference presented by means of short Spanish sentences containing a
by pressing either a left or right response key. Responses were pronoun and a conjugated verb. The conjugation of the verb
faster when past words were responded to with the left hand indicated whether the event was factual or potential. The
and future words with the right hand in comparison to a factual past condition used verbs in Indicative Past form (“ella
reversed mapping condition. This space-time congruency despertó” - “she woke up”) and the factual future condition
effect has been interpreted as evidence of the use of an used verbs in Indicative Future (“nosotros dormiremos” - “we
underlying left-to-right mental timeline. will sleep”). The potential past condition used verbs in
All available studies of this lateral mental timeline have used Subjunctive Pluperfect Past (“él hubiera trabajado”- roughly
past and future factual events. Some studies have used single corresponding to “he would have worked”) or Indicative
words (temporal adverbials and tensed verbs: Flumini & Conditional (“ella se dormiría” - “she would fall asleep”).
Santiago, 2013; Ouellet, Santiago, Funes, & Lupiañez, 2010; Participants were asked to categorize all sentences as referring
Ouellet, Santiago, Israeli, & Gabay, 2010; Santiago et al, 2007; to past or future by means of lateralized left and right
Torralbo, Santiago, & Lupiáñez, 2006; Weger & Pratt, 2008). keypresses. In one block they used a congruent mapping (left-
Others have used short adverbial phrases (Casasanto & Bottini,

48
past right-future) and in another block the mapping was stimuli was used in each block (thereby each block comprising
reversed. 80 trials). Before each block there was a practice block of eight
trials per condition. Instructions appeared on screen at the
Experiment 1 beginning of each block.
We expected that potential events would activate the lateral
mental timeline as well as factual events do. Therefore, we Design Latency and accuracy were analyzed by means of an
predicted an interaction between temporal reference and ANOVA including the factors Potentiality (factual vs.
response side both for factual and potential events. It is potential) X Time (past vs. future) X Response side (left vs.
important to point out that only the interaction with response right) X Order of conditions (congruent-incongruent vs.
side is informative in this design, because the conditions incongruent-congruent). The design was a mixed factorial
defined by the factors potentiality and time were not matched design, with Potentiality, Time, and Response side
in length, word frequency, verb form complexity, verb form manipulated within participants and Order of conditions
frequency, and so on. In other words, time and potentiality are manipulated between participants. The Order of conditions
between-item factors, and therefore, their main effects or two- factor was introduced to decrease error variance, and its effects
way interaction might arise because of uncontrolled item and interactions will not be reported.
variables. In contrast, response side is a within-item factor, and
therefore, its interaction with either time and/or potentiality Results
cannot be accounted for by differences among items. Due to experimenter error, three verbal stimuli in the factual
condition (‘Nosotros silbamos’, ‘Nosotros dormimos’ and
Methods ‘Nosotros soñamos’) were ambiguous as to their conjugation,
Participants Twenty eight students (32.5 mean age, one left- as they take identical forms in Indicative Past and Present.
handed) of the Autonomous University of Barcelona These represented 3.5% (168 trials). Errors occurred on 6.2%
volunteered to participate. All of them were native Spanish of the remaining trials and were excluded from the latency
speakers. analysis. In order to avoid the influence of outliers, after
inspection of the RT distribution we excluded latencies below
Materials Verbal stimuli were 80 Spanish expressions with 400 ms and above 3,500 ms, what amounted to discarding an
conjugated verb forms. The verb forms were generated by additional 1.5% (62) of correct trials.
using 20 intransitive regular Spanish verbs. They were
conjugated in factual past (Indicative Past); potential past Reaction Time Analysis Centrally for our hypotheses, Time
(Subjunctive Pluperfect Past); factual future (Indicative interacted with Response side (F(1,27)=8.71, p=.006, η2=.24).
Future) and potential future (Indicative Conditional). Sentence Moreover, there was no three-way interaction between
length was between 13 and 20 characters. Potentiality, Time, and Response side (F<1), indicating that
the size of the interaction between Time and Response side
Procedure The experiment was programmed in E-Prime was the same for both factual and potential events. This was
(Schneider, Eschman, & Zucolotto, 2002) and run in a sound supported by independent analyses of the interaction between
attenuated room. Stimuli were presented at the centre of a Time and Response side for factual events (F(1,27)=8.33,
computer screen (1024 x 768 pixels, 24.5 x 41 cm), spanning p=.008, η2=.24) and potential events (F(1,27)=7.56, p=.01,
6.23º of visual angle, in white letters over a black background. η2=.22). Figure 1 illustrates these results.
The distance between screen and participant was 0.59 m. One Additionally, both Potentiality (F(1,27)=6.94, p=.01, η2=.21)
session lasted approximately 20 minutes. Participants pressed a and Response side (F(1,27)=5.05, p=.03, η2=.16) yielded main
left and right response keys on a keyboard. The “a” and “6” significant effects. There was no main effect of Time (F<1).
keys were used, covered by stickers of the same colour. At the There was an interaction between Potentiality and Time
beginning of each trial a fixation cross was presented for 500 (F(1,27)=7.86, p=.009, η2=.23), and no interaction between
ms before a randomly chosen sentence appeared on the centre Potentiality and Response side (F(1,27)=1.34, p=.26, η2=.05).
of the screen. It remained on screen until the participant’s
response or a maximum time of 4,000 ms. Then there was an Accuracy Analysis Time and Response side did not interact
interval of 3,000 ms. Wrong responses were followed by a 440 (F<1). Potentiality had a marginally significant effect
Hz beep that lasted 500 ms. The next trial started 3,000 ms (F(1,27)=3.69, p=.07, η2=.12), but neither Time (F<1) nor
after a correct response or the offset of the auditory feedback. Response side (F(1,27)=2.20, p=.15, η2=.08) did. The
There were two experimental blocks, one for the congruent interaction between Potentiality and Time (F(1,27)=13.21,
time-response mapping and the other for the incongruent p=.001, η2=.33) was significant. There was also an interesting,
mapping. In the congruent condition, participants pressed the and unexpected, interaction between Potentiality and Response
left key in response to both past factual and potential verb side (F(1,27)=8.18, p=.008, η2=.23) on the form: better
forms, and the right key in response to both future factual and accuracy to potential events with the left hand and to factual
potential verb forms. In the incongruent condition, this events with the right hand than when using the opposite
mapping was reversed. The order of blocks was mappings.
counterbalanced over participants. The same set of verbal

49
we will assess whether the potential events can activate the
lateral timeline all by themselves.

Experiment 2
The aim of this experiment was to examine whether the
potential past and future verb forms are able to activate left
and right space when presented in a context that does not
include factual events. As in Experiment 1, the interaction
between Time and Response side was crucial for our
hypothesis: we expected that performance would be better in
the congruent conditions.

Methods
Participants Thirty four students of the Universidad de la
República (mean age 26.8 years, 3 left-handed) volunteered to
participate. They were all native Spanish speakers.

Materials Verbal stimuli were the 40 potential expressions of


Experiment 1.

Procedure Regarding sound attenuation, screen size and


resolution, and visual angle, conditions were similar to
Experiment 1. The procedure was identical to Experiment 1 in
all other details.

Design Latency and accuracy were analyzed by means of an


ANOVA including the factors Time (past vs. future) X
Response side (left vs. right) X Order of conditions
(congruent-incongruent vs. incongruent-congruent). Time and
Response side were manipulated within participants and Order
Figure 1: Mean latencies (ms) for factual (Panel A) and of conditions was manipulated between participants.
potential (Panel B) events in Experiment 1 (error bars show
Standard Error of the Mean). Results
Errors occurred on 5.23% (142) of the trials, and were
Discussion excluded from the latency analysis. After inspection of the RT
Experiment 1 revealed a space-time congruency effect both for distribution we also excluded correct trials with latencies
factual and potential events. Participants responded faster to below 335 ms and above 4,000 ms, what amounted to
both kinds of events when past was mapped to the left hand discarding an additional 1.7% (43 trials).
and future to the right hand than with the opposite mapping.
The size of the effect was the same for either kind of event. Reaction Time Analysis There was a main effect of Response
This suggests the activation of the lateral mental timeline in side (F(1,33)=5.06, p=.03, η2=.13), but not of Time (F<1).
both cases. Centrally for our research, a significant interaction between
However, there is an alternative explanation of the Time and Response side emerged (F(1,33)=6.53, p=.02,
interaction between Time and Response side in the processing η2=.17). Figure 2 illustrates these results.
of potential events. On this account, by intermixing factual and We also analyzed the potential trial data from the two
potential trials and assigning response keys to past and future experiments including Experiment as a factor. In the overall
reference all along the block we may have induced a carry- analysis, the interaction between Time and Response side was
over of the space-time congruency effect from factual to also significant (F(1,60)=13.45, p=.001, η2=.18). Moreover,
potential trials. In other words, it is possible that potential trials the three-way interaction between Time, Response side and
only showed the left-right past-future congruency effect Experiment was not significant (F<1). Thus, the space-time
because they were intermixed with factual trials, which do congruency effect had the same size in Experiments 1 and 2.
show the effect.
One possible way to sort out the carry-over account is to Accuracy Analysis The interaction between Time and
remove the factual trials altogether, keeping only the potential Response side approached significance (F(1,33)=3.15, p=.09,
trials. The carry-over account is based on the possibility that η2=.09). Neither Time (F<1) nor Response side (F(1,33)=1.31,
factuality would play a role on activating the left-right past- p=.26, η2=.04) produced significant main effects.
future mental timeline. Then, by eliminating the factual trials,

50
Materials Verbal stimuli were the same 80 Spanish
expressions of Experiment 1, with four exceptions: Firstly, the
ambiguous items in Experiment 1 (“Nosotros silbamos”,
“Nosotros dormimos” and “Nosotros soñamos”) were fixed
by changing their conjugation from first person plural to third
person singular. Additionally, the verb “permanecer”
(“remain”) was replaced by the verb “sonreír” (“smile”)
because by itself “permanecer” does not express a specific
event.

Procedure The procedure followed closely Experiment 1, with


the following exceptions. At the beginning of the session, we
ensured that participants clearly discriminated factual from
potential expressions using an example. Additionally, the
practice block was extended to sixteen trials per condition.
This was because, on pilot testing, the potentiality task was
shown to be more difficult than the temporality task. In one
Figure 2: Mean latencies (ms) for potential events in mapping condition, participants pressed the left key in
Experiment 2 (error bars show Standard Error of the Mean). response to a factual event and the right key in response to a
potential event. In the other mapping condition, the assignment
Discussion was reversed.
A clear space-time congruency effect was observed when
potential past and future events were presented without factual Design Latency and accuracy were analyzed by means of an
events in the experimental context: participants responded ANOVA including the same factors as in Experiment 1:
faster when past was mapped to the left hand and future to the Potentiality (factual vs. potential) X Time (past vs. future) X
right hand, than with the opposite mapping. The size of the Response side (left vs. right) X Order of mapping conditions.
effect was not different from that observed in Experiment 1.
Therefore, present data rule out the possibility that the Results
congruency effect observed for potential events in Experiment Errors occurred on 5.4% (257) of the trials, and were excluded
1 was induced by the presence of factual events in the from the latency analysis. After inspection of the RT
experimental materials. distribution we excluded correct trials with latencies below
The results of these experiments provide evidence of 450 ms and above 3,200 ms, what amounted to discarding an
genuine space-to-time mappings for potential events additional 1.6% (74 trials).
. Available studies suggest that the activation of these space-
time associations appears to be non-automatic (Ulrich & Reaction Time Analysis Centrally for our concerns, the
Maienborn, 2010). In order to assess whether there is an interaction between Time and Response side and the three-way
automatic activation of the left-right timeline for potential interaction between Potentiality, Time, and Response side
events, in Experiment 3 we asked participants to judge the were not significant (all F<1). Figure 3 illustrates these results.
potentiality of the event, instead of its reference to past or We also observed an unexpected interaction between
future. Potentiality and Response side (F(1,29)=6.99, p=.01, η2=.19):
responses were faster when potential events were mapped onto
Experiment 3 the left hand and factual events onto the right hand than when
The aim of this experiment was to examine whether there is an using the opposite mapping. Additionally, Potentiality
automatic activation of the left-right mental timeline for (F(1,29)=4.51, p=.04, η2=.14) produced a main effect, as in
potential (and factual) events. With this goal, we asked Experiment 1. In contrast to that experiment, the main effect of
participants to judge whether each expression referred to a Time was significant (F(1,29)=18.87, p<.001, η2=.39) whereas
factual or potential event. Thus, the potentiality dimension Response side was not (F<1). The interaction between
became task-relevant and the temporal dimension task- Potentiality and Time was replicated (F(1,29)=12.21, p=.002,
irrelevant. We did not expect a space-time congruency effect η2=.29).
under these conditions, neither for factual nor potential events. With the aim of comparing the effects of the type of task
(time vs. potentiality judgment) on the interactions between
Methods Time and Response side, as well as on the newly found
Participants Thirty new undergraduate students of the interaction between Potentiality and Response side, we
Universidad de la República participated as volunteers (mean analyzed together the data from Experiments 1 and 3. The
age 26 years, no left-handers). They were all native Spanish overall two-way interaction between Time and Response side
speakers. was significant (F(1,56)=8.55, p=.005, η2=.13), and it was
modulated by Experiment (F(1,56)=8.21, p=.006, η2=.13),
supporting a change in the space-time congruency effect, from

51
being present in Experiment 1 to being absent in Experiment 3. Discussion
Additionally, the overall two-way interaction between In a task using a potentiality judgment, the latency measure did
Potentiality and Response side reached significance not detect any space-time congruency effect, neither for factual
(F(1,56)=7.94, p=.007, η2=.12), and it was also qualified by nor potential events. This result supports the non-automaticity
Experiment (F(1,56)=3.52, p=.07, η2=.06): it was absent in of the activation of the lateral mental timeline, as suggested by
Experiment 1 and present in Experiment 3. Therefore, the task- Ulrich and Maienborn (2010).
relevant conceptual dimension in each experiment interacted Instead, there was an unexpected space-potentiality
with the side of response. interaction: participants responded faster when potential events
were mapped to the left hand, and factual events to the right
Accuracy Analysis Neither the interaction between Time and hand, than when using the opposite mapping. This interaction
Response (F<1) nor the interaction between Time, Potentiality, was suggested by the accuracy analysis of Experiment 1 and it
and Response (F(1,29)=2.52, p=.12, η2=.08) were significant. was confirmed by the omnibus ANOVA of latency data in
The interaction between Potentiality and Response side fell Experiments 1 and 3. We discuss this finding in the following
short of significance (F(1,29)=2.65, p=.11, η2=.08). There was section.
also an interaction between Time and Potentiality
(F(1,29)=4.91, p=.04, η2=.15), a main effect of Time General Discussion
(F(1,29)=40.09, p<.001, η2=.58), and the main effect of
Response side approached significance (F(1,29)=3.06, p=.09, Do potential events activate the mental time line? The present
η2=.10). study provided a clear answer to this question: Yes, speakers
map time onto space when processing potential events.
Experiment 1 showed that the space-time congruency effect
for potential events was indistinguishable from the effect
observed for factual events. Experiment 2 showed that the
effect is genuine and arises even when the experimental
materials comprise only potential events. Finally, Experiment
3 showed that the activation of the lateral mental timeline is
non-automatic for both kinds of events. Events occurring at
different moments in both factual and fictive worlds are
mentally represented along a continuum that runs along the
lateral axis.
The present study also revealed an unexpected congruency
effect between lateral space and potentiality, such that the
processing of expressions was facilitated when potential events
were mapped onto the left hand and factual events onto the
right hand (as compared to the opposite mapping). This space-
potentiality mapping is also non-automatic, as it only arised
when participants judged potentiality and not time.
What can be the causes of this effect? One possibility relies
on the inherently potential character of future events. Speakers
of Aymara refer to the future using the word for “back”, and to
the past using the word for “front” (Núñez & Sweetser, 2006).
These authors suggested that the motivation for this conceptual
mapping is the fact that the past can be “seen” clearly, as it has
already happened, but the future cannot. Under this account,
the potentiality of the future would support mapping both
future and potential onto right space in Spanish speakers.
However, present data actually show the opposite mapping
(potential-left, factual- right), and therefore rule out this
account.
Another possibility is based on the polarity correspondence
hypothesis proposed by Proctor and Cho (2006). If both
potentiality and lateral space are polar dimensions, with a
marked and an unmarked (default) pole, the polarity
correspondence hypothesis would predict that processing
should be facilitated when the poles of the same sign are
Figure 3: Mean latencies (ms) for factual (Panel A) and mapped onto each other. It seems intuitively correct to assume
potential (Panel B) events in Experiment 1 (error bars show that the unmarked pole of the dimension of potentiality is the
Standard Error of the Mean). factual pole, and that the unmarked pole of the dimension of

52
lateral space is the right side (at least for right-handers). an implicit nonlinguistic task. Cognitive Science, 34(8),
Therefore, mapping factual on the right response and potential 1430–1451.
on the left response would facilitate processing as compared to Gilead, M., Liberman, N., & Maril, A. (2012). Construing
the reversed mapping. counterfactual worlds: The role of abstraction. European
This view can account for the observed space-potentiality Journal of Social Psychology, 42(3), 391–397.
congruency effect, and at present we believe it is the best Hegarty, M. (2004). Mechanical reasoning by mental
available explanation of it. However, it opens other simulation. Trends in Cognitive Sciences, 8(6), 280–5.
challenging questions. Recently, Santiago and Lakens (2014) Johnson-Laird, P. N. (1983). Mental Models. Cambridge, MA:
have shown that polarity correspondence cannot explain the Cambridge University Press.
mapping of time (nor numbers) onto lateral space. What are, Kaup, B., Yaxley, R. H., Madden, C. J., Zwaan, R. a, &
then, the factors that make some conceptual dimensions, such Lüdtke, J. (2007). Experiential simulations of negated text
as the potentiality dimension, able to generate a polarity information. Quarterly Journal of Experimental Psychology,
correspondence effect, and that distinguish it from the 60(7), 976–90.
dimension of time, which is not? Lakoff, G. & Johnson, M. (1980). The metaphorical structure
To conclude, the present study has shown that potential past of the human conceptual system. Cognitive Science, 4, 195-
and future events activate the lateral mental timeline to the 208.
same extent as factual events do. In doing so, it has also Núñez, R. E., & Sweetser, E. (2006). With the future behind
revealed an interesting new phenomenon: the mental them: Convergent evidence from Aymara language and
representation of the dimension of potentiality can also gesture in the crosslinguistic comparison of spatial
establish links to the lateral spatial dimension, at least under construals of time. Cognitive Science, 30(3), 401–450.
conditions in which potentiality is task relevant. More research Orenes, I., Beltrán, D., & Santamaría, C. (2014). How negation
is needed to clarify the exact nature of this relation. is understood: Evidence from the visual world paradigm.
Journal of Memory and Language, 74, 36–45.
Acknowledgments Ouellet, M., Santiago, J., Israeli, Z., & Gabay, S. (2010). Is the
This research was supported by the Uruguayan National future the right time? Experimental Psychology, 57, 308-
Research and Innovation Agency to Roberto Aguirre, and by 314.
grants P09-SEJ-4772 from Junta de Andalucía and European Ouellet, M., Santiago, J., Funes, M. J., & Lupiáñez, J. (2010).
Regional Development Fund, and PSI2012-32464 from the Thinking about the future moves attention to the right.
Spanish Ministry of Economy and Competitivity to Julio Journal of Experimental Psychology: Human Perception
Santiago. We would like to thank Claudia Maienborn, Rolf and Performance, 36, 17–24.
Ulrich, Marc Ouellet, and Juan Carlos Valle-Lisboa for their Proctor, R. W., & Cho, Y. S. (2006). Polarity correspondence:
help throughout this project. A general principle for performance of speeded binary
classification tasks. Psychological Bulletin, 132(3), 416–42.
Santiago, J., & Lakens, D. (2014). Can conceptual congruency
References effects between number, time, and space be accounted for by
Baumeister, R. F., & Masicampo, E. J. (2010). Conscious polarity correspondence? Acta Psychologica.
thought is for facilitating social and cultural interactions: doi:10.1016/j.actpsy.2014.09.016
How mental simulations serve the animal-culture interface. Santiago, J., Lupiáñez, J., Pérez, E., & Funes, M. J. (2007).
Psychological Review, 117(3), 945–71. Time (also) flies from left to right. Psychonomic Bulletin &
Casasanto, D., & Bottini, R. (2014). Mirror reading can Review, 14, 512–516.
reverse the flow of time. Journal of Experimental Santiago, J., Román, A., Ouellet, M., Rodríguez, N., & Pérez-
Psychology: General, 143, 473–479. Azor, P. (2010). In hindsight, life flows from left to right.
De Vega, M., León, I., Hernández, J. A., Valdés, M., Padrón, Psychological Research, 74(1), 59–70.
I., & Ferstl, E. C. (2014). Action sentences activate sensory Schneider, W., Eschman, A., & Zucolotto, A. (2002). E- Prime
motor regions in the brain independently of their status of User’s Guide. Pittsburgh: Psychology Software Tools Inc.
reality. Journal of Cognitive Neuroscience, 26(7), 1363– Torralbo, A., Santiago, J., & Lupiáñez, J. (2006). Flexible
1376. conceptual projection of time onto spatial frames of
Ferguson, H. J., Tresh, M., & Leblond, J. (2013). Examining reference. Cognitive Science, 30, 745–757.
mental simulations of uncertain events. Psychonomic Tversky, B., Kugelmass, S. & Winter, A. (1991). Cross-
Bulletin & Review, 20(2), 391–9. cultural and developmental trends in graphic productions.
Flumini, A., & Santiago, J. (2013). Time (also) flies from left Cognitive Psychology 23, 515–557.
to right... if it is needed! In M. Knauff, M. Pauen, N. Sebanz, Ulrich, R. & Maienborn, C. (2010). Left–right coding of past
& I. Wachmuz (Eds.), Proceedings of the 36th Annual and future in language: The mental timeline during sentence
Conference of the Cognitive Science Society (pp. 2315– processing. Cognition, 117, 126–138.
2320). Austin, TX: Cognitive Science Society. Weger, U., & Pratt, J. (2008). Time flies like an arrow: Space-
Fuhrman, O., & Boroditsky, L. (2010). Cross-cultural time compatibility effects suggest the use of a mental time-
differences in mental representations of time: Evidence from line. Psychonomic Bulletin and Review, 15, 426–430.

53
A Rational Model for Individual Differences in Preference Choice
Sheeraz Ahmad (sahmad@cs.ucsd.edu)
Department of Computer Science and Engineering
University of California, San Diego

Angela J. Yu (ajyu@ucsd.edu)
Department of Cognitive Science
University of California, San Diego

Abstract is irrational and sub-optimal (Tversky, 1972; Kahneman et


Human preference choice suffers curious contextual effects: al., 1982). More recently, it has been proposed that these
the relative preference between two multi-attribute options peculiarities may arise from specific idiosyncrasies in neural
(e.g. cars with differing safety and economy ratings) can dra- architecture or dynamics in brain areas that support prefer-
matically shift, depending on the presence/absence of addi-
tional options. This phenomenon defies any simple utility- ence choice (Busemeyer & Townsend, 1993; Roe et al., 2001;
based account of choice, and has been taken to imply Usher & McClelland, 2004; Trueblood, 2012).
irrationalities/sub-optimalities in human decision-making, or
reflect idiosyncrasies in neural processing. Recently, we used In contrast to a purely utility-based account of preference
a Bayesian model to show that these contextual effects are nor- choice, we recently proposed an alternative normative model
mative consequences of observers using options to learn about of decision-making (Shenoy & Yu, 2013), which assumed
the “market”. However, it had an unsavory implication that
all decision-makers asymptotically converge to the same be- that humans do not have fixed, perfectly known utility val-
liefs/behavior. Here, we propose a new model that uses both ues assigned to options, but instead may suffer uncertainties
market and personal utilities to make choices. This model still about how to assign utilities both within an attribution dimen-
captures the contextual effects, while also allowing asymptotic
differences in individual preferences and providing a general sion and also jointly for a combination of attribute values in
framework for explaining how consumption informs one’s be- a multi-attribute scenario. Consequently, observers use avail-
liefs and preferences. able options not only to make choices, as assumed by pre-
Keywords: decision making; preference choice; multi- vious utility-based models of preference choice, but also to
attribute; contextual effects; individual differences; Bayesian
learning learn about the range and distribution of attribute values gen-
erally available in the “market”, as well as some market-based
Introduction sense for how the attributes ought to be valued against each
Humans are regularly faced with decisions involving a choice other. By this account, the addition of a third option con-
between options with multiple attributes. For example, one fers extra information about the “market” and may therefore
may have to choose between a car that has a higher safety influence the relative preference between the two original op-
rating but lower mileage and another that has a lower safety tions, with the effect expected to be particularly strong when
rating but higher mileage; or one may have to choose be- the decision-maker has relatively little experience with the
tween a PhD applicant who has better grades but worse letters particular “market.” For example, in the PhD applicant exam-
and another that has worse grades but better letters. There ple, suppose a professor is evaluating applications of students
may not be a universal or obvious way to make these de- from a foreign country (whose academic structure is not well
cisions, and indeed humans often exhibit inconsistencies in known to him), and he must choose between an applicant A
their choices. One particular class of peculiarities in human who has a test score of 290 and a grade of 90, and an applicant
preference choice has garnered special attention in psychol- B with a test score of 300 and a grade of 80. Based on this
ogy research, namely a type of contextual effects whereby an data alone, he may not have a strong preference between the
individual’s relative preference between two options can be applicants, but if a third applicant C comes in with a test score
altered, or even reversed, when a third option (a ‘decoy’) is of 290 and a grade of 130, the professor may strongly shift
introduced into the choice-set (Kahneman & Tversky, 1979; his relative preference between A and B toward B (though
Kahneman et al., 1982; Tversky & Simonson, 1993). he may prefer C over both A and B), since C’s grade of 130
The discovery of these contextual effects has caused dif- shows that a grade of 90 over 80 is not really much of an ad-
ficulty for traditional, utility-based accounts of preference- vantage at all. Indeed, we showed that this inferential model
based decision-making in humans. If each option has an as- was able to account for all three classical contextual effects:
sociated scalar utility for an individual, and the probability attraction, compromise, and similarity (Shenoy & Yu, 2013).
of choosing option A monotonically depends on the utility This earlier model provided a normative explanation for
of option A in comparison to the utility of option B, then why contextual effects arise in rational decision-makers. It
the relative preference between A and B should not change also provided a means for modeling individual differences, by
when a third option, C, is added or removed (Luce, 1959; allowing different individuals to have different prior beliefs
Thurstone, 1954; Tversky, 1972). This contradiction has con- about the distribution of attribute values in different dimen-
tributed to the school of thought that human decision-making sions and how a combination of attribute values from multiple

54
dimensions jointly determine the utility function. However, it Consumption Model
makes the odd prediction that all decision-makers will even-
tually converge to the same beliefs about the market, given
sufficient exposure to the various options available in the mar- We consider an option space with two attributes, which are
ket; in Bayesian parlance, this is because the iteratively up- combined using a Cobb-Douglas utility function (Douglas,
dated posterior distribution will converge to the same (delta) 1976), parametrized by γ, which specifies the relative impor-
function regardless of the prior distribution. This prediction tance of the two attributes to the joint utility function. The
flies in the face of empirical data (Malaviya & Sivakumar, utility or value of an option (x, y) is v = xγ y1−γ . Note that
1998; Müller et al., 2012) and casual observations that dif- the two attributes nonlinearly combine to determine the util-
ferent individuals often exhibit systematic and long-lasting ity function: this captures diminishing marginal utility, or the
differences in their preference choices. idea that differential change in an already abundant attribute
In this work, we propose an alternate Bayesian account of contributes less to the overall utility than a similar change in
preference choice, which captures the notion that each person a scarce attribute (Hicks, 1932).
entertains beliefs about both the market-based utility func-
In our model, there are two parameters that contribute to
tion (learned from exposure to available options, or “win-
the attribute trade-off, in turn affecting the multi-attribute util-
dow shopping”) and a personal utility function (learned from
ity function. The market utility, vm = xγ y1−γ , is parametrized
choosing/consuming specific options), and combines the two
by market tradeoff parameter γ, and the personal utility, v p =
in making preference choices. The introduction of this per-
xλ y1−λ , is parametrized by the personal tradeoff parameter λ.
sonal utility function allows individuals to have persistent di-
We propose the following way to combine these towards a
versity in their preference choices. In the following, we first
net utility: v = vwm v1−w
p , since it leads to a simplified form:
describe the three classical contextual effects, then the new
v=x y ζ 1−ζ (ζ , wγ + (1 − w)λ), and intuitively provides an
Bayesian model, followed by simulation results that compare
our model against classical contextual effects and more subtle “average” of the two utilities (in this case, a generalized ge-
individual differences, as well as against the previous model ometric mean). The parameter w can be interpreted as a per-
(Shenoy & Yu, 2013) in the context of asymptotic learning; sonality trait dictating how much an individual values unique-
we conclude with a discussion of the broader implications of ness as opposed to conforming to the market. For example,
this work, relationship to related work, and potential direc- consider a consumer buying a car; even though she may be
tions for future research. more price-conscious herself, she might still want to buy a
trendy, more expensive car because of external considerations
Three Contextual Effects like status symbol, peer pressure or resale value (w > 0.5). A
Three classical contextual effects have been observed when more rebellious consumer may value uniqueness more and
starting from two originally equally preferable options (say make a decision primarily based on her own preference, giv-
X and Y), each with two attributes: (1) attraction effect ing little consideration to market preferences (w < 0.5).
(Fig. 1A) (Huber & Payne, 1982; Heath & Chatterjee, 1995), The generative model (Fig. 2A) also assumes that depend-
where introduction of an option Z, that is close to Y and is ing on whether the option is consumed or not (ci ∈ {0, 1}), the
dominated by it on both attributes, leads to an increased pref- individual gets different levels of satisfaction. If the option is
erence of Y over X. (2) compromise effect (Fig. 1B) (Simon- not consumed (ci = 0), a small level of stochastic satisfac-
son, 1989), where introduction of an extreme option Z, that is tion is received, which can be interpreted to be resulting from
better than Y on one attribute and inferior on the other, as well mental simulation (thinking about the consumption of the op-
as is farther from X, again leads to an increased preference tion) but which yields no information about the hidden pref-
of Y over X. (3) similarity effect (Fig. 1C) (Tversky, 1972), erence λ. On the other hand, if the option is consumed, the
where the introduction of an option Z, that is almost compa- satisfaction derived is a noisy version of the personal utility,
rable to Y on both attributes, leads to an increased preference v p = xλ y1−λ for an option (x, y), which in turn is informative
of X over Y. about λ. For example, after accepting a few of the PhD ap-
plicants, the professor may decide, after all, that a student’s
grades are a better predictive of success for conducting re-
search in her lab than recommendation letters; while another
professor may decide just the opposite.
For inference, when the options are not consumed (ci =
0 ∀ i), the node si does not depend on oi or λ, thus the directed
edges from these to si are effectively removed and the model
Figure 1: Three classical contextual effects. Options X and Y are
equi-preferable in the absence of a third option. Arrow denotes the reduces to a simplified version of our previous model (Shenoy
direction of the preference shift when a decoy, Z, is added. (A) & Yu, 2013) (for w = 1). More generally, the posterior belief
Attraction effect. (B) Compromise effect. (C) Similarity effect. about λ can be updated based on the observed options (o =
{o}i ) and satisfaction (o = {o}i ) as:

55
Figure 2: (A) Graphical model showing how options (oi ) are generated in the market, based on the hidden value (vi ), market preference (γ),
individual preference (λ), weight (w) and the parameter θi . Also describes the satisfaction as a noisy representation of the personal value,
depending on the individual preference (λ) and whether the option is consumed (ci ). (B) For a particular setting of the attribute trade-off
parameter (ζ), all options lying on the same curve have the same value/utility as denoted on the curve; different curves signify different
values. (C) For a fixed value (vi ), all options lying on the same curve have the same attribute trade-off (ζ), as denoted on the curve; different
curves signify different trade-offs.

options is expected because of the residual uncertainty in the


posterior distributions of the option values (Debreu, 1958;
P(λ|o, s) ∝ P(s|o, λ)P(o|λ)P(λ) Blavatskyy, 2008).
( )(Z
= ∏ P(si |oi , λ; ci ) P(γ)P(θ)P(v)P(o|v, γ, θ, λ; w) Results
i γ,θ,v
) In this section, we apply the proposed models to differ-
ent multi-attribute choice tasks, in which sometimes sub-
dγdθdv P(λ)
jects have to choose among just two options, and sometimes
( ) among three options whereby a “decoy” is added to the two
= ∏ P(si |oi , λ; ci ) original options (see Fig. 1). We first show how our model ac-
i counts for the three classical contextual effects; we then use
(Z Z  ) the model to capture several observed individual differences
P(γ) ∏ P(θi )P(vi )P(oi |vi , γ, θi , λ; w)dθi dvi dγ P(λ) in existing experimental literature.
γ i θi ,vi For all simulations, the market parameters for the prefer-
(1) ence γ are a = 2 and b = 2, and for the utility/value vi , k = 20
and µ = 50. Other parameters used are, σθ = 20 and obser-
Similarly, the joint posterior on the hidden values can be vation noise σ0 = 2. Lastly, the options are X = (40, 60), Y =
updated as: (60, 40), Z = (50, 30) for attraction, Z = (80, 20) for compro-
mise, and Z = (65, 35) for similarity effect. Since no simple
P(v|o, s) closed form expressions exist for the different posteriors (e.g.
(Z
Eq. 1) and approximations based on discretization of contin-
Z 
∝ P(γ)P(λ) ∏ P(θi )P(oi |vi , γ, θi , λ; w)dθi uous variables are inexact and inefficient, we use a program
γ,λ i θi
)( ) called JAGS (Plummer, 2003) that uses Gibbs sampling (Ge-
man & Geman, 1984) to generate samples from the posterior
[P(si |oi , λ; ci )] dγ dλ ∏ P(vi ) (2)
i
distribution of the desired model parameters (v, γ and λ in our
case).
As in our previous model, we assume that the decision pol- For the simulations in Fig. 3, we assume that there is no
icy involves sampling from the posterior P(v|o); the sample consumption (ci = 0), and that the individual relies solely on
v̂ is then used to choose an option: π(v̂) = argmax j v̂ j . Thus, the information from market options (w = 1). With these
stochasticity in choice upon presentation of the same set of settings, our model reproduces all three contextual effects,

56
Figure 3: Proportion of choices for different effects and their explanation. (A) X and Y are equi-preferable. (B) Adding an option Z inferior
to Y increases the preference for Y relative to X. (C) Adding an option Z even more extreme than Y, in relation to X, increases the preference
for Y. (D) Adding an option Z very similar to Y, but not clearly more or less preferable to Y, decreases the preference for Y relative to X. (E)
X and Y lie on the equi-preference curve which represents a fixed value; any option above this curve would be considered more valuable, and
below would be considered less valuable. (F) The attraction decoy makes Y appear better on average. (G) The compromise decoy make Y
appear better on average. (H) The valuation of Y and Z are highly correlated, such that they tend to be considered both better than X or both
worse than X, for different settings of the hidden variable γ. When they are both worse than X, X is chosen; when they are both better than
X, Y is chosen half the time (Z the other half). Thus, on average, X is chosen more often than Y, even though they are on average considered
about equally valuable.

so that even though the two options are equi-preferable by leading to overall higher frequency of choosing X than Y.
themselves (Fig. 3A), the preference shifts towards Y when We also investigate scenarios where individual exhibit dif-
an attraction or compromise decoy is added to the choice ferent behavior based on their previous experience. Experi-
set (Fig. 3B and C), and towards X when a similarity de- ments have show that contextual effects are not always robust,
coy is added (Fig. 3D). Next, in order to understand how the with individual differences emerging when subjects value at-
model works for all the contextual effects, we show the re- tributes differently (Malaviya & Sivakumar, 1998; Müller et
sulting equi-preference curves, the locus of options that all al., 2012). To show that our model can capture such devia-
have the same value for a given preference γ (in other words, tions, we simulate the scenario where an individual prefers
all the points on the curve xγ y1−γ = v). Clearly, there are in- attribute y more over attribute x (λ = 0.35). Furthermore, we
finitely many equi-preference curves, one corresponding to assume that the individual relies equally on self and market
each value of v, but we only plot the ones, for visual sim- preference (w = 0.5). With these settings, we observe that
plicity, that always pass through the option X. When there compromise effect becomes insignificant (Fig. 4A), which is
are only two options, such an equi-preference curve with γ what has been observed for consumers who are less quality
set to its posterior mean also passes through Y, making the conscious (attribute x) and more price conscious (attribute y)
two options equally attractive (Fig. 3E). For the attraction ef- when making brand choices (Müller et al., 2012).
fect, option Y lies above such an equi-preference curve , mak-
ing it appear relatively more lucrative (Fig. 3F); however op- Lastly, we show how consuming the options can help an
tion X is still selected owing to the stochasticity in the option individual discover self-preference, and how the the process
values. Similar explanation holds for the compromise effect can lead to divergence in the choice behavior of two indi-
(Fig. 3G). For the similarity effect, the equi-preference curve vidual who learn their preference (from consumption) along
with γ set to its posterior mean is not particularly informative with the market preference (from options). In the model pro-
(Fig. 3H, blue curve), since this effect arises due to the close posed earlier (Shenoy & Yu, 2013), the only way the indi-
correlation between the valuation of the option Y and Z in the viduals could differ is if they have different priors over the
model. As can be seen, the two options are likely to appear market preference parameter (γ). However, with increase ex-
better or worse than X together (green and red curves respec- perience, the individual beliefs would converge, consequently
tively). Therefore, when they appear better, the choice gets leading to the same choice for all individuals who are ex-
split between them; when they appear worse, X gets chosen, perienced. In the simulations, we consider two individuals
starting with different priors on γ, Beta(2, 3) and Beta(3, 2),

57
Figure 4: (A) Individual difference in contextual effects. Compromise effect becomes insignificant in the consumption model for individuals
that value attribute 1 less (λ = 0.35). (B) & (C) Evolution of belief about market preference with experience for two individuals starting
with different priors, using the previous simple model and our consumption model respectively. The translucent bounded region shows the
standard deviation around the mean belief.

respectively. When both of these individuals experience the preferences, thus leading to diversification of individual pref-
same set of options, we note that their beliefs eventually con- erences.
verge (Fig. 4B). This contradicts the every day observation, However, there are some experimental findings that still
where individuals with very similar experience with a prod- prove challenging for the new model. For example, the attrac-
uct category, can still converge on buying different brands. tion effect has sometimes been observed to diminish for con-
Next, we consider the same two individuals and simulate their sumers with a low level of experience (Malaviya & Sivaku-
beliefs under our consumption model. We additionally note mar, 1998), contradicting a straight-forward prediction of our
that individual 1 prefers attribute 2 more (λ = 0.3), whereas model. Another curious phenomenon is the phantom de-
individual 2 prefers attribute 1 more (λ = 0.7), both indi- coy effect (Pratkanis & Farquhar, 1992; Pettibone & Wedell,
viduals giving equal weight to market-preference and self- 2007), which is very similar to the attraction effect, except
preference (w = 0.5). The extra parameters are ks = 100 and the decoy Z is slightly better than one of the options, say Y
µs = 5. Now the individual beliefs about joint preference di- (rather than worse, as in the attraction effect) and that the sub-
verge, in accordance with their hidden preference (Fig. 4C). ject is informed that this decoy is not actually available as a
Thus, the model provides an explanation for why two individ- choice; in that case, human observers reliably shift their pref-
uals with the same experience may end up preferring different erence toward Y instead of away from it, as our model would
options, and more generally a framework for explaining dif- currently predict. These more nuanced cases require further
ferent framing effects, as well as variations thereof, based on attention and provide fruitful avenues for future research.
the individual’s experience, personality, and other causes of
individualized preferences. In the future, we also wish to investigate whether an in-
dividual has a fixed relative weight for personal and mar-
Discussion ket preferences, or whether this trade-off can change either
with experience or context. One possibility is, as the individ-
In this paper, we proposed a Bayesian model that takes both ual gains more experience, the trade-off starts to favor self-
individual preference as well as effects of consuming options preference, thus requiring a more sophisticated model where
into account. This model reduces to a simplified version of the trade-off parameter (w) is dependent on experience (per-
our previous proposed model (Shenoy & Yu, 2013) in the ab- haps tied to internal uncertainty/confidence). Another direc-
sence of consuming chosen options (and experiencing cor- tion is to extend the model to allow for “vicarious satisfac-
responding satisfaction levels), and can explain not only all tion” so that the high demand of an option, with say attribute
three contextual effects but also observed individual differ- 1 as the larger attribute, would signal that attribute to be more
ences arising from experience and personal preference. Fur- preferable. Such a model could provide a computational ex-
thermore, the model relaxes an assumption we made in our planation to the phantom decoy effect. Lastly, humans may
prior work that the observer needs to infer both the option actively seek which options to consume, in order to figure
values as well as the “fair market value”. Instead, it assumes out their self-preference, e.g. trying a Vietnamese restaurant
that the observer only need to infer the relative utility (value) after having tried a Thai restaurant to see if a slightly dif-
of the available options, and not what constitute “fair” in ferent spice level would be more satisfying, and our model
the market place in absolute terms. Therefore, our current can be extended to incorporate this active decision making
framework simplifies the previous model, as well as provides component. This can potentially be achieved by choosing op-
a general framework for explaining how consumption com- tions based on a more sophisticated criteria that takes into
bines with “window-shopping” to inform one’s beliefs and account not only the immediate values or satisfaction (as is

58
done in the current formulation), but also the long term value Müller, H., Vogt, B., & Kroll, E. B. (2012). To be or not
and informational goals as well as the cost of consuming an to be price consciousa segment-based analysis of compro-
option. Insights from the field of active learning (see (Settles, mise effects in market-like framings. Psychology & Mar-
2009) for a survey) can provide the foundation for such a pur- keting, 29(2), 107–116.
suit. In summary, our work provides a novel computational Pettibone, J. C., & Wedell, D. H. (2007). Testing alternative
framework to account for individual differences in a variety explanations of phantom decoy effects. Journal of Behav-
of observed preference choice behavior, and opens up venues ioral Decision Making, 20(3), 323–341.
for future investigations into more sophisticated models of Plummer, M. (2003). Jags: A program for analysis of
preference-based decision making. bayesian graphical models using gibbs sampling.
Pratkanis, A. R., & Farquhar, P. H. (1992). A brief history of
Acknowledgments
research on phantom alternatives: Evidence for seven em-
We thank Laurence Aitchison and Peter Dayan for stimulat- pirical generalizations about phantoms. Basic and Applied
ing discussions on ideas related to the current paper. This Social Psychology, 13(1), 103–122.
material is based upon work supported in part by the U. S. Roe, R. M., Busemeyer, J. R., & Townsend, J. T. (2001). Mul-
Army Research Laboratory and the U. S. Army Research Of- tialternative decision field theory: A dynamic connectionst
fice under contract/grant number W911NF1110391. model of decision making. Psychological review, 108(2),
370–392.
References
Settles, B. (2009). Active learning literature survey (Com-
Blavatskyy, P. R. (2008). Stochastic utility theorem. Journal puter Science Technical Report No. 1648). University of
of Mathematical Economics, 44(11), 1049–1056. Wisconsin–Madison.
Busemeyer, J., & Townsend, J. (1993). Decision field the- Shenoy, P., & Yu, A. (2013). Rational preference shifts in
ory: a dynamic-cognitive approach to decision making in multi-attribute choice: What is fair? In Proceedings of the
an uncertain environment. Psychological review, 100(3), 35th annual conference of the cognitive science society.
432–459. Simonson, I. (1989). Choice based on reasons: the case of
Debreu, G. (1958). Stochastic choice and cardinal utility. attraction and compromise effects. Journal of Consumer
Econometrica, 26, 440–444. Research, 16(158-74).
Douglas, P. (1976). The Cobb-Douglas production function Thurstone, L. (1954). The measurement of values. Psycho-
once again: its history, its testing, and some new empirical logical Review.
values. The Journal of Political Economy. Trueblood, J. S. (2012, October). Multialternative context
Geman, S., & Geman, D. (1984). Stochastic relaxation, effects obtained using an inference task. Psychonomic bul-
gibbs distributions, and the bayesian restoration of images. letin & review, 19(5), 962–8. doi: 10.3758/s13423-012-
Pattern Analysis and Machine Intelligence, IEEE Transac- 0288-9
tions on(6), 721–741. Tversky, A. (1972). Elimination by aspects: a theory of
choice. Psychological review, 79(4).
Heath, T., & Chatterjee, S. (1995). Asymmetric decoy ef-
fects on lower-quality versus higher-quality brands: Meta- Tversky, A., & Simonson, I. (1993). Context dependent pref-
analytic and experimental evidence. Journal of Consumer erences. Management Science, 39, 1179–89.
Research, 22, 268–284. Usher, M., & McClelland, J. L. (2004). Loss aversion and
inhibition in dynamical models of multialternative choice.
Hicks, J. R. (1932). The theory of wages.
Psychological review, 111(3), 757.
Huber, J., & Payne, J. (1982). Adding asymmetrically domi-
nated alternatives: Violations of regularity and the similar-
ity hypothesis. Journal of Consumer Research.
Kahneman, D., Slovic, P., & Tversky, A. (Eds.). (1982).
Judgement under uncertainty: Heuristics and biases. Cam-
bridge, UK: Cambridge University Press.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An
analysis of decisions under risk. Econometrica, 47, 313-
27.
Luce, R. D. (1959). Individual choice behavior: A theoretical
analysis. New York: Wiley.
Malaviya, P., & Sivakumar, K. (1998). The moderating effect
of product category knowledge and attribute importance on
the attraction effect. Marketing Letters, 9(1), 93–106.

59
Motion event expressions in language and gesture: Evidence from Persian
Niloofar Akhavan (nakhavan13@ku.edu.tr)
Department of Psychology, Koç University
Rumelifeneri Yolu Sariyer 34450
Istanbul - TURKEY

Nazbanou Bonnie Nozari (nozari@jhu.edu)


Department of Neurology, Department of Cognitive Science, Johns Hopkins University,
1629 Thames Street, Suite 350, Baltimore, MD 21231, USA

Tilbe Göksun (tgoksun@ku.edu.tr)


Department of Psychology, Koç University
Rumelifeneri Yolu Sariyer 34450
Istanbul - TURKEY

Abstract
Motion Events
How do people conceptualize motion events and talk about
them? The current study examines how gestural
Languages vary in how they encode motion elements. A
representations of motion events arise from linguistic motion event consists of four semantic components; figure,
expressions in Persian, which has characteristics of both ground, path, and manner. Figure refers to a particular point
Talmy’s satellite- and verb-framed languages. We examined in space with respect to another object. Ground refers to
native Persian speakers’ speech and gestures in describing 20 another physical object, which serves as a reference point
motion events. We focused on two motion event components: with respect to which the figure is located. Path refers to the
path (trajectory of motion like up) and manner (how the translational motion and manner refers to motor pattern of
action is performed like jumping). Results indicated that when
expressing motion, Persian speakers produced path in both the movement of the figure (Slobin, 1996). Talmy (1985,
speech and gesture, whereas manner was conveyed only 1991) categorizes most of the world’s languages into two
through speech (mostly as adverbs). Additionally, dynamic major types of satellite-framed (S-framed) and verb-framed
gestures tended to occur in the same order they were uttered. (V-framed) languages based on the core elements of path
The difference between path and manner findings asks for and manner. S-framed languages such as English
further research to examine language-gesture interaction in (Germanic), Mandarin (Sino-Tibetan), and Russian (Slavic),
detail among different languages. Results also suggest
integrate motion with manner in the main verb and express
refinement in gesture theories that argue for one-to-one
correspondence between speech and gesture. path with a verb particle or a satellite thus leaving the verb
free to encode manner (e.g., run down (the hill)). On the
Keywords: motion events, gesture, language and thought, other hand, V-framed languages such as Spanish
Persian, Farsi
(Romance), Turkish (Turkic), and Hebrew (Semitic)
incorporate motion with path in the main verb and express
Introduction manner in the subordinated verb (e.g., in Turkish, koşarak
The relation between language and thought has been a çıktı ‘go up runningly’) thus, using two verbal clauses to
question for decades. Throughout the history of philosophy express both path and manner.
it has been implied that the limits of language are the limits After studying various languages, Slobin (1996)
of thinking and people of different languages have different concludes that lexicalization patterns are presumed to
thought processes (Wittgenstein, 1921; Whorf, 1956). More strongly impact thinking and formation of visuo-spatial
recently, Berman and Slobin (1994) stated, “the particular representations. However, language may not directly
ways of filtering and packaging information is shaped by influence event apprehension. Individuals’ attention for
one’s native language” (p. 613). This hypothesis, “thinking encoding motion events can be allocated to their language-
for speaking,” argues that thinking is provoked by the specific components only when they need to speak about
requirements of a linguistic code. In particular, the these events. For example, in a study comparing English
information to be expressed has to be tailored to speaking and Greek speakers, Papafragou, Hulbert, and Trueswell
and must be compatible with the lexical and constructional (2008) found that language on cognition effects arise only
resources of a given language (Slobin, 1996). However, when language is recruited to achieve a task, but not during
others argue that language underspecifies thought and event perception in general. Thus far, using various
cognitive organization is independent of language (e.g., methodologies, many languages have been analyzed through
Gleitman & Papafragou, 2005). In this paper, we investigate Tamly’s approach. To our knowledge, there is only one
how Persian speakers conceptualize motion events in both
speech and gesture.

60
study examining how Persian speakers encode motion Nevertheless, there has been an unresolved debate about
events in narratives (Feiz, 2011). whether speech and gesture form a tightly integrated
Feiz (2011) claims that Persian exhibits a mixed typology, communication system or whether they originate from the
with characteristics of both S-framed and V-framed same representational system or two separate but
languages. The similarity to S-framed languages is apparent interrelated systems (Alibali, Kita, & Young, 2000;
in cases where path information is expressed in path Butterworth & Hadar, 1989; Goldin-Meadow, 2003; Kita &
satellites. An example is: (az tappe) baala davidan ‘to run Özyürek, 2003; Krauss, Chen, & Gotfexnum, 2000;
up (the hill),’ in which baala ‘up’ is a satellite, and davidan McNeill, 1992). Research by McNeill (1992, 2005) supports
‘to run’ is a verb containing manner information. There are the view that speech and gesture originate from the same
cases, in which path information is coded in the verb (e.g. representational system. Along these lines, McNeill (1992)
charkhidan ‘to pirouette’), but these are not common, and suggested that since gesture conveys information not
most of them need an additional preposition to become a explicitly encoded in speech, it provides a unique window to
transitive verb (e.g., dor -e- […] charkhidan ‘to circle view underlying thought.
around’). On the other hand, the number of verbs that Other theories suggest that speech and gesture are
contain manner information (e.g., davidan ‘to run’) is also generated by two separate but highly interrelated systems
not high in Persian (Feiz, 2011), leaving manner (Alibali et al., 2000; Kita, 2000; Kita & Özyürek, 2003;
information to be expressed mostly in other parts of speech, Krauss et al., 2000). For example, Kita (2000) proposed that
such as adverbs, davan davan raftan ‘to go runningly’ gestures help to organize and package visuo-spatial
(davan davan = runningly; raftan = to go). In this sense, information into units of language. Moreover, Kita &
Persian more closely resembles a V-framed language. Özyürek (2003) proposed the Interface Model that also
In general, many Persian verbs contain neither path nor predicts priming between language and gestures. They
manner information. This is due to the special structure of emphasize the influence of language on gestures, but
most verbs, which are a combination of a noun + a light suggest independent systems for speech and gesture.
verb (e.g., harekat ‘motion’ + kardan ‘to do’ = to move). According to this model, language-specific aspects can also
The light verbs that appear in such compounds are limited in be represented in the gestures people use.
number and have different levels of fidelity to their original Cross-linguistic studies suggest that speakers of different
meaning, for example, kardan in harekat kardan preserves languages produce different gestures for the same concept,
its original meaning ‘to do (motion)’ but the verb zadan ‘to and these gestures follow the linguistic structure of the
hit’ means something very different when used as a light utterances in their language (e.g., Kita, 2000; Kita &
verb in ghadam zadan ‘to stroll’. Thus, the core semantics Özyürek, 2003; McNeill, 2000; McNeill & Duncan, 2000).
of the light verbs are rarely interpreted literally, and the For example, English speakers express manner together
meaning of the verb relies heavily on its noun component. with path in their speech and gesture. In contrast, Spanish
These noun components also vary in how much semantic speakers omit manner in their speech but express it in a
information they convey. Some, like harekat ‘motion’ are compensatory way in their gesture, and their path gestures
broad and underspecified, thus, harekat kardan can mean follow the verbs (McNeill & Duncan, 2000). Further studies
any type of motion. Some, like ghadam ‘(slow) step’, have with English, Turkish, and Japanese speakers have revealed
more specific semantics, thus, conveying a little more than that the gestural representations mainly corresponded to
just the basic action. But since many nouns do not carry language-specific encodings of motion events (Kita &
detailed information, peripheral details like path and manner Özyürek, 2003; Kita et al., 2007; Özyürek et al., 2005). In
are usually left to other parts, such as prepositions and particular, English speakers use one verbal clause to express
adverbs. The construct described above makes Persian a both elements of path and manner with one manner + path
unique case for studying the relationship between language conflated gesture (e.g., ‘running up’ is represented by a
and gesture. gesture of making index and middle fingers move upward
direction while alternating fingers), whereas, Turkish
Gesture use in Motion Events speakers use two verbal clauses thus they more likely use
Spontaneous co-speech gestures are bodily motions that two separate gestures for path and manner (e.g., ‘going up
embody a meaning related to the accompanying speech. runningly’ is expressed by an upward motion for ‘go up’
These gestures are commonly used for thinking and and then alternating index and middle fingers without
communicating information that are visuospatial in nature upward movement for ‘run’). In the Turkish case path is in
(Alibali, 2005; Kita & Özyürek, 2003), providing a great the main clause (go up) and manner is in the subordinate
deal of information about the internal structure of the (adverbial) clause (running).
speech. They also reflect internal cognitive process and The close correspondence of linguistic structure to
provide a window on the embodied nature of mind gesture, however, has not been universally supported. In a
(Hostetter & Alibali, 2008). Co-speech gestures are closely recent study comparing English and Turkish monolinguals
linked, both in meaning and time, to the speech they with controlled stimuli (similar to the ones used in this
accompany (McNeill, 2005). study), Karaduman et al. (2015) found that English speakers
produced more manner and path combinations in their

61
speech compared to Turkish participants, as expected. Participants watched 20 dynamic movie clips, depicting
Interestingly, this difference was not apparent in their different motion events with randomized combinations of
spontaneous gestures. In contrast to the previous findings, 10 manners (hop, skip, walk, run, cartwheel, crawl, jump,
they found that speakers of both languages used twirl, march, step) and 9 paths (through, to, out of, under,
predominantly path gestures in their gesture use, despite the over, in front of, around, across, into). The actions were
differences in their utterances. performed by a woman in an outdoor area (see Figure 1 for
sample stimuli).
The Current Study
We reviewed evidence on the sensitivity of gestures to the
structure of the language that they accompany. These results
point to a close correspondence between the linguistic and
gesture systems. The question is whether there are other
factors that limit this one to one correspondence. Results of
Karaduman et al. (2015), which show similar gesture
production in spite of linguistic differences, point to a
common component to gestures, one that may mirror
universals of human cognition, rather than specifics of a
language.
Figure 1: Sample stimuli from the experimental task. The
The current study aims to investigate how gestural
picture is a still frame from the movie clip of a motion
representations of motion events stem from linguistic
event: jump over. The yellow arrows indicate the direction
expressions in Persian, the unique characteristics of which
of the movement.
we reviewed earlier. This is the first controlled study to
examine Persian in terms of differences in spatial language
Procedure
characteristics and the way these differences are manifested
in spontaneous co-speech gestures. All participants were tested individually in their home
Due to the structure of verbs, discussed earlier in the environment. Before each task, two practice trials were
paper, Persian speakers are expected to express path of given. Participants were then presented 20 trials in a
motion with prepositions and manner of motion as verb or randomized order. After watching each video, they were
adverb together with using auxiliary verbs. Our critical asked to describe the action in the clip. No instructions were
prediction concerns gestures: if linguistic forms correspond given regarding gesture use. Participants’ hands and torsos
very closely to gestures, as expected by the Interface Model were videotaped.
(Kita & Özyürek, 2003), we predict that Persian speakers
would use two types of gestures: (1) when the speech Coding
resembles English expressions conflating path and manner
information, such as baala davidan ‘to run up,’ there would Speech. All the speech was transcribed verbatim by the first
be one conflated gesture representing both path and manner author (a native Persian speaker). The transcribed speech
of motion; (2) when the speech resembles Turkish as in the was coded for the correct use of manner (how the action is
case of davan davan bala raftan ‘to go up runningly’ there performed) and path (the trajectory of action). First, for
would be two separate gestures; one referring to path and each trial, the coder assessed whether there was any manner
the other referring to manner of motion. If factors other than and/or path information mentioned. Second, the pattern of
linguistic form influence the production of gestures in speech responses in terms of path and manner information
Persian speakers, we might instead see dissociation between was categorized into groups of manner only, path only or
gesture and speech. If this arises due to a cognitive manner + path together. Each trial containing a path or
universal, we may observe the same pattern reported by manner received a subcode as follows: For manner, it was
Karaduman et al. (2015), with predominance of path coded as expressed in (1) a verb (davidan ‘to run’), (2) an
gestures. adverb (Bodo bodo [lit. ‘run run’] ‘in a running fashion’; ley
ley konan ‘hop hop doing’), and (3) the noun in a compound
Method verb containing a light verb (donbaal kardan ‘to chase).
Path was also categorized into path as (1) a preposition
Participants (kenare ‘side of’), (2) a verb (charkhidan ‘to pirouette’), (3)
a verb + a preposition (dor charkhidan ‘to circle around’),
15 monolingual Persian speakers between the ages of 18 and (4) a light verb (baala raftan ‘to go up’; dar aamadan ‘to
30 (7 females and 8 males) were tested in Iran. emerge’), (5) a light verb + a preposition (az bein rad
shodan ‘from pass do’).
Task and stimuli
Video clips of different motion events developed and Gesture. Participants’ spontaneous gestures were
standardized by Göksun et al. (under review) were used. transcribed from the video. First, for each trial, the number

62
of gestures was coded. Second, the gestures were classified Table 1: Number and percentages of manner and path
as static or dynamic. Static gestures referred to objects or expressions in speech
locative properties of objects (e.g., pointing finger to refer to
the preposition ‘above’). Dynamic gestures involved the Manner Number Percentage
movements of hands that could represent the action of the (1) Verb 40 15.0
person such as ‘moving the index from left to right to (2) Adverb 187 72.5
display the direction of the motion’. Third, the dynamic (3) Light verb 31 12.3
gestures were classified into (1) manner only, (2) path only, Total 258
and (3) path + manner together. Manner only gestures are
those that enact the style of a motion without emphasizing Path Number Percentage
the trajectory of the movement, the path (e.g., circular (1) Preposition 45 16.8
movement of the index finger in place to represent (2) Verb 2 .7
cartwheeling). Path only gestures show a direction without (3) Verb+Preposition 29 10.8
representing the manner (e.g., movement of the index finger (4) Light verb 14 5.2
in an arc pattern along the horizontal axis from right to left (5) Light verb+Preposition 178 66.4
to represent ‘across’). Path + manner gestures constitute Total 268
both components simultaneously (e.g., circular movement
of index finger along the horizontal axis from right to left to Gesture analyses
represent ‘cartwheeling across’). Figure 2 represents these
Participants produced a total of 364 gestures in 237 out of
three types of gestures.
300 trials. On average, 72.5% of gestures were identified as
dynamic, 9.3% of gestures were static, and 19.5% as beat
gestures. In this paper, we only focused on dynamic gestures
that referred to motions in the clips.
Next, we analyzed the overall pattern of dynamic gestures
in terms of expressing manner and path information. Results
showed that participants expressed significantly more path
information in their gestures than manner information or
path + manner information together (conflated), x2 (2, N =
(a) (b) (c) 264) = 157.36, p < .001.
Last, we analyzed how participants used path and manner
Figure 2: Sample gestures that represent (a) a path only information in each trial. In these trial-based analyses, we
motion (e.g., across), (b) a manner only motion (e.g., coded whether participants used only path, only manner or
cartwheeling), and (c) a path + manner (cartwheeling both in each trial. For the trials where participants used both
across). manner and path we also coded the order of their
occurrence. The majority of dynamic gestures were
Results identified as path only (M= 59.7%, SD =17.25) compared to
manner only (M=11.2%, SD =16.98), manner + path
Speech analyses conflations (M=8.7%, SD =10.41), or their combinations
Participants expressed manner (M= 85.67%, SD = 8.42) and (M=20.4%, SD =12.56), x2 (3, N = 206) = 138.58, p <.001.
path (M=87.33%, SD =10.83) information similarly with no A closer look at the combined expressions indicated that
statistically significant difference between them, t(14) = - people often used gestures for manner information before
.418, p = .682. Next, we analyzed how participants encoded path information, the same order in which they were uttered,
manner in speech. We found that people produced manner x2 (1, N = 42) = 34.38 p<. 001. All numbers and
in adverbial form more frequently than in any other forms corresponding percentages for the following analyses are
(M= 72.48%, SD =14.10), x2 (2, N = 258) = 178.39, p < provided in Table 2.
.001. For example, manner information was expressed as
‘bepar bepar’ (in hop hop fashion) for hopping. We then Table 2: Number and percentages of manner only, path
analyzed path expressions and found that paths were mostly only, and manner +path expressions in gesture
encoded with preposition + light verb (‘dor -e- derakht
mire’ lit. = around tree goes, ‘goes around the tree’; ‘az Dynamic Gesture Number Percentage
khiyaban rad shod’ lit. = from street cross did, ‘crossed the Manner Only 23 11.2
street’), x2 (4, N = 268) = 380.32, p <.001 (see Table 1 for Path Only 123 59.7
all numbers and corresponding percentages Manner + Path (conflated) 18 8.7
Manner and Path 42 20.4
Total 206

63
Speech – gesture relations Additionally, in this study we used naturalistic stimuli (as
To further explore the information represented for motion opposed to the cartoon events in the previous studies) and
event expressions, we analyzed whether path and manner 20 sentences all containing different combinations of paths
were conveyed in both speech and gesture or in isolation. and manners. This imposes a high load on both language
We found dissociation between the coexistence of the two and gesture systems. While there might be close
gesture types and linguistic information. Participants tended correspondence between representations in the two systems,
to encode path information in both speech and gesture the two may have different capacities and limits. For
whereas manner was mostly produced within speech only, example, dual sequential representations might be harder to
x2 (3, N=474) = 58.91, p < .001 (see Table 3). represent in the gesture system. If so, when faced with such
demands, the system may drop the gesture that is manually
Table 3: Number and percentages of speech and gesture more demanding.
combinations Our results, however, provide support for Interface Model
in 3 ways. First, overall there were very few manner verbs
Combinations Number Percentage and manner as a verb + path as a preposition combinations
in speech. As a result participants produced manner + path
Path Speech only 86 18.1
conflated gestures only in 9% of the gesture trials. This
Manner Speech only 167 35.23
finding matches with the S-framed language characteristic
Path Speech-gesture 152 32
of Persian (like English). Second, because path and manner
Manner Speech-gesture 69 14.5
information were mostly separated in two clauses as a
Total 474
property of V-framed languages (like Turkish), manner and
path information were displayed in separate gestures, if any.
Discussion Third and novel to this study, gesture sequences followed
To our knowledge, this is the first study on motion event the same order as their linguistic counterparts. Past research
conceptualization in speech and gesture in Persian. We has mostly ignored the effects of word order on gesture use.
investigated how dynamic gestures contributed to motion In Persian, subject–object–verb is the formal word order,
expression in speech in a language that has characteristics of but there is high flexibility in ordering words. However,
both Talmy’s S- and V-framed languages. adverbs usually do not come after the main verb
As expected, Persian speakers frequently used adverbs, (Megerdoomian, 2001). In keeping with this, we found that
prepositions, and light verbs to describe both manner and manner gestures that are expressed as adverbs in speech
path of the events. Interestingly, however, people’s dynamic occurred before path gestures that were mainly expressed as
gestures mainly referred to path of motion, and not its a combination of preposition and light verbs at the final part
manner. Manner + path conflated gestures made up only 8% of the sentence. This finding illustrates the role of language-
of dynamic gestures. When looking at the overall and trial specific encoding on gesture use, as claimed by the Interface
based gesture frequencies, Persian speakers tended to Model.
gesture for path information, whereas manner information In summary, the study of Persian, a language unique in its
was expressed in speech only. large number of noun + light verb compounds, and
The key question was whether variation in speech possessing the characteristics of both S- and V-framed
corresponded to the gestural expressions. The Interface languages, revealed the same pattern of correspondence
Model suggests that there is an online interface between between path gestures and the utterances describing them, as
linguistic and gestural representations in utterance English and Turkish. The dominance of path gestures across
generation, in which spatial imagery is packaged into verbal languages may point to the universality of language-gesture
units (Kita & Özyürek, 2003; Özyürek et al., 2005). Our interaction. On the other hand, other expressions such as
results are only partially compatible with this account. manner + path conflations with manner + preposition
The majority of dynamic gestures described path of utterances, manner and path information production in two
motion (60% of gesture) without including any manner separate clauses as in speech, and manner-path gesture
information. This finding is in line with recent research by orders paralleling word order in speech are compatible with
Karaduman et al. (2015), which indicated the predominant the influence of language-specific structures on gesture.
use of path gestures in contrast to manner gestures among These findings call for closer inspection of factors involved
both English and Turkish speakers. This supports a common in language-gesture interaction, and refinement of the
and possibly a universal pattern to gesture production that Interface Model.
may not be sensitive to linguistic structure. Why do English,
Turkish and Persian speakers in our studies prefer path References
gestures to manner gestures? We cannot answer this
Alibali, M. W. (2005). Gesture in spatial cognition:
question with certainty, but put forth possible reasons,
Expressing, communicating, and thinking about spatial
leaving a more definite answer to future studies.
information. Spatial Cognition and Computation, 5(4),
The easiness of manual movements for paths could be a
307-331.
factor. In particular, perhaps path is easier than manner to be
displayed by hands due to its spatial configuration.

64
Alibali, M. W., Kita, S., & Young, A. J. (2000). Gesture and McNeill, D. (2005). Gesture and Thought. Chicago:
the process of speech production: We think, therefore we University of Chicago Press.
gesture. Language and cognitive processes, 15(6), 593- McNeill, David & Susan D. Duncan (2000). Growth points
613. in thinking-for-speaking. In D. McNeill (Ed.), Language
Whorf, B.L. (1956). Language, thought, and reality. and gesture. Cambridge: Cambridge University Press.
Cambridge, MA: MIT Press. Megerdoomian, K. (2001). Event Structure and Complex
Berman, R. A., & Slobin, D. I. (Eds.). (2013). Relating Predicates in Persian: Special issue on the Syntax of
events in narrative: A crosslinguistic developmental Iranian languages. Canadian Journal of
study. Psychology Press. Linguistics 46(1/2), 97-125.
Butterworth, B., and Hadar, U. (1989). Gesture, speech and Özyürek, A., Kita, S., Allen, S. E., Furman, R., & Brown,
computational stages: A reply to McNeill. Psychological A. (2005). How does linguistic framing of events
Review, 96, 168-74. influence co-speech gestures?: Insights from
Feiz, P. (2011). Traveling through space in Persian and crosslinguistic variations and similarities. Gesture, 5(1-2),
English: a comparative analysis of motion events in 219-240.
elicited narratives. Lang. Sci., 33 (3), 401–416. Papafragou, A., Hulbert, J., & Trueswell, J. (2008). Does
Gleitman, L., & Papafragou, A. (2005). Language and language guide event perception? Evidence from eye
thought. In K. Holyoak & R. Morrison (Ed.), Cambridge movements. Cognition, 108(1), 155-184.
Handbook of Thinking and Reasoning. Cambridge: Slobin, D.I. (1996). Two ways to travel: verbs of motion in
Cambridge University Press. English and Spanish. In: Shibatani, M., Thompson, S.A.
Gleitman, L.R. & Papafragou, A. (2013). Relations between (Eds.), Grammatical Constructions: Their Form and
language and thought. In D. Reisberg (Ed.), Handbook of Meaning. Oxford University Press, Oxford.
Cognitive Psychology. New York: Oxford University Talmy, L. (1985). Lexicalization patterns: Semantic
Press. structure in lexical forms. In T. Shopen (Ed.), Language
Goldin-Meadow, S. (2003). Hearing gesture: How our typology and semantic description. Cambridge,
hands help us think. Cambridge: Harvard University Cambridge University Press
Press. Talmy, L. (1991). Path to realization: A typology of event
Göksun, T., Lehet, M., Malykhina, K., & Chatterjee, A. conflation. Proceedings of the Seventeenth Annual
(under review). Spontaneous gesture and spatial language: Meeting of the Berkeley Linguistics Society. Berkley, CA.
Evidence from focal brain injury. Wittgenstein, L. (1922). Logisch-Philosophische
Hostetter, A. B., & Alibali, M. W. (2008). Visible Abhandlung (Ed.). New York, NY: Hardcourt, Brace &
embodiment: Gestures as simulated action. Psychonomic Company, INC.
bulletin & review, 15(3), 495-514.
Karaduman, A. N., Çatak, E. N., Bahtiyar. S., Göksun, T.,
(March, 2015). The role of gestures in describing motion
in English and Turkish. Poster accepted to present at the
International Conference on Psychological Science,
Amsterdam, Netherlands.
Kita, S. (2000). How representational gestures help
speaking. In D. McNeill (Ed.), Language and gesture.
Cambridge: Cambridge University Press.
Kita, S., & Özyürek, A. (2003). What does cross-linguistic
variation in semantic coordination of speech and gesture
reveal?: Evidence for an interface representation of spatial
thinking and speaking. Journal of Memory and
language, 48(1), 16-32.
Kita, S., Özyürek, A., Allen, S., Brown, A., Furman, R., &
Ishizuka, T. (2007). Relations between syntactic encoding
and co-speech gestures: Implications for a model of
speech and gesture production. Language and Cognitive
Processes,22(8), 1212-1236.
Krauss, R. M., Chen, Y., & Gottesman, R. F. (2000).
Lexical gestures and lexical access: A process model. In
D. McNeill (Ed.), Language and gesture. Cambridge:
Cambridge University Press
McNeill, D. (1992). Hand and Mind: What Gestures Reveal
about Thought. Chicago: University of Chicago Press.

65
Daxing with a Dax: Evidence of Productive Lexical Structures in Children
Sara Al-Mughairy (sara.m@berkeley.edu)
Department of Psychology, University of California, Berkeley

Ruthe Foushee (foushee@berkeley.edu)


Department of Psychology, University of California, Berkeley

David Barner (barner@ucsd.edu)


Department of Psychology, University of California, San Diego

Mahesh Srinivasan (srinivasan@berkeley.edu)


Department of Psychology, University of California, Berkeley

Abstract follow patterns, children could form generalizations about


these patterns and spontaneously apply them to novel words.
In English, many words can be used flexibly to label arti-
facts, as nouns, or functional uses of those artifacts, as verbs: This would promote word learning, because children would
We can shovel snow with a shovel and comb our hair with a only need to learn one phonological form to express multiple
comb. Here, we examine whether young children form gen- meanings. For instance, if children learned the pattern that
eralizations about flexibility from early in life and use such
generalizations to predict new word meanings. When chil- labels of artifacts can also describe the uses of those artifacts,
dren learn a new word for an artifact, do they also expect it they could infer that a new artifact name would also apply to
to label its functional use, and vice versa? In Experiment 1, the artifact’s functional use. Children could form such gen-
we show that when four- and five-year-olds are taught a first
novel word to label a familiar action—e.g., that bucking means eralizations early in life, thus facilitating early word learning.
shoveling—they exclude the artifact involved in this action— This would be especially plausible if patterns of flexibility
i.e., the shovel—as the meaning of a second novel word (e.g., marked conceptual relations that children find salient, like
gork). This suggests that children spontaneously expected the
first novel word—which referred to the action—to also refer to the relationship between an artifact and its function (Caslet
the artifact. In Experiment 2, we show that this pattern extends & Kelemen, 2007).
to words that label novel actions involving novel artifacts, sug- An alternative possibility—borrowing from usage-based
gesting that children expect any word for an action to label the
artifact that helps carry out that action. Experiment 3 traces theories of language development (e.g., Tomasello, 2003)—is
how such generalizations may arise in development. In partic- that children only gradually form generalizations about flex-
ular, we show that while four- and five-year-olds each expect ibility, instance by instance. By this account, children might
words to label artifacts and their functional uses, three-year-
olds may not. initially treat the noun and verb meanings of words like shovel
Keywords: Language acquisition; polysemy; mutual exclu- and hammer as homophones. After this, they would sepa-
sivity; class-extension rules rately represent the relationships between shovel/shovel and
hammer/hammer as “islands of flexibility”, prior to uniting
Introduction them under the same abstract linguistic principle. Critically,
Language gifts us with the resources to innovate in order to on this account, children only form productive generaliza-
express our ideas. One such resource is the potential to flex- tions after exposure to many pairs of flexible words, limiting
ibly extend words to new meanings (Copestake & Briscoe, the potential of such generalizations to constrain children’s
1995; Pustejovsky, 1995). For example, many verbs in En- hypotheses about word meaning.
glish have been formed from nouns (see Clark & Clark, One approach to distinguish between these two possibili-
1979): when we shovel the snow, bike to school, or comb ties is to explore whether young children spontaneously ex-
our hair, we describe our actions in terms of the artifacts that pect words to be used flexibly: e.g., for words that label tools
help us carry out those actions. Similarly, there are many to also label their functional uses, and vice versa. Although
instances of nouns in English that have been derived from studies have provided evidence that young children com-
verbs: when we take a long walk, use an eraser, or cheer prehend the relationship between flexible nouns and verbs
for a wrestler, we use nouns defined by their corresponding (Berman, 1999; Bushnell & Maratsos, 1984; Lippeveld &
actions. Adult speakers have productive knowledge of these Oshima-Takane, 2010), and even produce new ones—e.g.,
patterns, and can use them in a systematic way to meet com- ”Don’t broom my mess” (example from Clark, 1982), they do
municative demands—e.g., She will Wikipedia the answer. not provide unequivocal evidence that children expect verbs
But when and how does this generative talent arise? to be formed from nouns.
The present study explores whether productive knowledge For example, in one study (Lippeveld & Oshima-Takane,
of flexibility emerges early in life, and might allow children 2010), two-and three-year-old children watched a video in
to predict new word meanings, thereby supporting lexical de- which a bottle opener was used to open a bottle, while a novel
velopment. When flexible words, like shovel, bike, and comb, noun vop was used to refer to the bottle opener—e.g., “This

66
is a vop! Look at what it can do to the bottle!” Later, chil- Experiment 1
dren were tested on their interpretation of the innovative verb Here, we examine whether children have formed lexical
formed from this noun—e.g., children were asked to “find structures to allow familiar words like shovel to be ex-
the one that is voping.” The three-year-olds (but not the two- tended between their different uses. Using familiar words al-
year-olds) responded in a way suggesting that they under- lowed us to test whether children treat existing corresponding
stood the new verb—they looked longer at an action that de- noun/verb pairs as unrelated lexical items, like homophones
picted bottle-opening than a different action. However, it is such as bat/bat. Such a lack of distinction between these two
possible that although children seemed to identify the bottle- varieties of phonological overlap would be expected accord-
opening action as the likely referent of the verb “voping,” ing to a usage-based account in the period prior to having
they may not have expected the bottle-opening event to be re- formed “islands of flexibility.” Previous studies using a sim-
ferred to as voping—they may only have formed a verb from ilar method to the one used here have shown that when chil-
vop when asked to “find voping.” Further, children could have dren learn that a novel word labels one homophone (e.g., that
succeeded even without forming a verb from a noun. When dax labels a baseball bat), they do not expect the novel word
children were taught what a vop was, they could have initially to also label the other homophone (e.g., an animal bat; Srini-
linked vop not just to the bottle-opener but also to the bottle- vasan & Snedeker, 2011). Thus, we reasoned that if chil-
opening event that it participated in, looking then at that event dren treat familiar noun and verb meanings as unrelated ho-
due to the phonological overlap between vop and voping. mophones, they should not expect a novel label for a verb
The present studies address the limitations of previous meaning (e.g., dax to label hammering) to also label the cor-
work by adapting a mutual exclusivity method to probe chil- responding noun meaning (e.g.,dax to label a hammer). In
dren’s expectations of how words can be used. In one form contrast, if familiar noun and verb meanings arise from a
of a mutual exclusivity task (e.g., Diesendruck & Markson, common, generative structure, children should expect a novel
2001), children are presented with two items for which chil- label for the verb meaning to also label the noun meaning.
dren do not know labels, and are taught that a novel word Methods
labels one of those items (e.g., blicket). They are then asked
to choose the referent of a second novel word (e.g., Give me Participants The participants were 20 children (7 girls) be-
the wug). Children tend to choose the item that has not yet tween the ages of 4;1 and 5;11 (M = 59 months). Four addi-
been labeled, excluding the item that already has a label on tional children participated but were excluded due to missing
the grounds that it should not have a second. the initial trials that gauged their understanding of the task
(1), for failing to identify any critical noun-verb pairs during
Here, we adapt this method to explore whether children ex- the warm-up (1), or because they didn’t want to continue (2).
pect that a novel verb—e.g., daxing—that is taught to refer to All children were either brought into the lab or recruited from
an action will also refer, as a noun, to the artifact that helps daycares in the San Diego area. All children received a small
carry out that action. If so, children should expect that a sec- gift for participating.
ond novel word—e.g., blicket—cannot refer to the artifact,
because the first novel word already does. This would be pre- Materials and Procedure We used a mutual exclusivity
dicted if children have formed productive lexical structures to task to examine whether children spontaneously extend novel
support flexible uses of verbs as nouns. If, on the other hand, labels between the noun and verb uses of words like shovel.
children treat words for actions and words for artifacts as sep- We familiarized the children to the task by first introducing
arate lexical items, they should not exclude artifacts as the them to a character named Monkey (by showing them a pic-
meanings of the second novel words. Critically, this method ture of Monkey). We told the children that one special thing
probes children’s own expectations about the extension of the about Monkey is that, because he is a Muppet, he speaks a
novel verb—children are given no evidence that this word can special Muppet language. We told the children that in the
be shifted, as in previous studies. game, they would learn some Muppet words.
Each of the trials included a training phase and a judgment
Using this method, we explore the productivity of the lex- phase. We initiated the training phrase of each critical trial by
ical structures three-, four-, and five-year-old children have asking the children if they knew the meaning of a novel Mup-
formed to support corresponding noun and verb pairs. In Ex- pet verb—e.g., “Sometimes, Monkey likes to buck stuff from
periment 1, we assess whether children treat nominal and ver- one place to another. Do you know what bucking is?” Hav-
bal uses of familiar words as separate lexical items by exam- ing established that the novel verb was an unfamiliar word
ining whether children expect a novel word that refers to a from Muppet language, we showed them a video in which the
familiar action—e.g., shoveling—to also refer to the artifact novel verb was used to refer to a familiar action. For example,
that carries out that action—e.g, a shovel. In Experiment 2, children were shown a video in which Monkey was shoveling
we explore whether this tendency also extends to novel ac- sand from a plate into a bowl, while hearing the verb buck
tions performed by novel artifacts. In Experiment 3, we in- used to describe the action in a number of ways. Children
vestigate how knowledge of and expectations about flexible heard the verb used in the infinitive to refer to what Monkey
words change throughout early development. was going to do (“He’s going to use it to buck something into

67
the bowl”), in the progressive to refer to the action as it was item were given feedback, but all children had to get two out
ongoing (“Wow, Monkey’s bucking it into the bowl”), and in of three of these items correct without feedback to proceed.
the past tense to refer the completed action (“Monkey bucked Children also received one foil warm-up trial that required
some stuff into the bowl”). The video also described how the them to choose between two items that were different from
affordances of the artifact facilitate the action (“What Mon- the one they were trained on. For example, in one trial, chil-
key has is pretty long and it can carry the stuff well”). Crit- dren were taught that spado referred to a knife and then had
ically, children did not receive any evidence that the novel to decide whether a table or chair was the referent of parma.
word could refer to the artifact itself (e.g., the shovel). These items were included to prevent children from expect-
Immediately following the video, the experimenter initi- ing that the first novel word could always be extended to one
ated the judgment phase of the trial: e.g., “So that’s what of the two pictures presented in the judgment phase.
bucking is. Now we know what bucking is. But now, I
want the gork. Show me the gork!” The child was then pre- Table 1: Experiment 1 Critical Items.
sented with a slide containing two pictures. The pictures
included an instance of (1) the artifact used in the video Training Phase Event Judgment Phase
(e.g., a shovel of a different color) and (2) the patient ob- Moop (bike) to school Tima = bike or school
ject/substance that had been acted upon (e.g., a second ex- Dax (hammer) nail into wood Kiv = hammer or nail
emplar of sand). This was to ensure that children were us- Buck (shovel) sand into bowl Gork = shovel or sand
ing the novel words to refer to categories, rather than indi- Tig (brush) hair on head Lum = brush or hair
vidual tokens. The child’s choice—which they indicated by Wug (tape) picture to box Koon = tape or picture
pointing—was then recorded. We reasoned that if children Kraz (lock) box Bip = lock or box
expect a word for an action to also refer to the artifact that Lorp (button) sweater Zot = button or sweater
carries out that action, they should exclude the artifact as the Jop (comb) hair on head Raj = comb or hair
referent of the second novel word and instead choose the pa-
tient. Critically, the English names of artifact and the patient
were not provided during the training phase, such that they After the warm-up items, the children were shown between
could each serve as candidate referents of the second novel one and four critical items, depending on how many noun-
word (e.g., we referred to the shovel as a “thing” or “what verb pairs they had produced in the pre-test. These items
Monkey has,” and referred to the sand as “stuff” or “some- were administered in a fixed order; see Table 1 for a descrip-
thing”). tion of the training and judgment phases of the critical items.
On average, children received 3.8 critical items, with 6 tested
Children only received a particular critical item if they had on biking, 19 tested on hammering, 9 tested on shoveling, 14
been able to accurately produce the English noun and verb tested on brushing, 12 tested on taping, 2 tested on locking,
uses related to that particular item in an earlier pre-test (e.g., 10 tested on buttoning, and 4 tested on combing. Because
use shovel as both a noun and verb). In the pre-test, we tested the three mutual exclusivity warm-up trials provided a ceil-
children’s knowledge of the critical nouns by showing them ing measure of children’s ability to make mutual exclusivity
pictures of the artifacts and asking them to name them. After judgments, we report children’s performance on these trials
testing the critical noun uses, we tested children on the critical below. Finally, we constructed two versions of the task that
verb uses, by showing them pictures of people using those varied with respect to whether the pictures in the judgment
artifacts and asking them to describe what the people were phases of the trials were presented to the left or right of the
doing. If children did not immediately name these actions, child.
we prompted them—e.g., for the shoveling item: “What is
she doing to the sand?” We only accepted responses that used Results and Discussion
the target artifact verb—e.g., “She’s picking up the sand with Because of our small number of items and the categorical na-
a shovel” wasn’t accepted. Children were tested on the noun ture of our data, we present only non-parametric analyses.1
and verb forms of comb, shovel, tape, bicycle, button, brush, On the warm-up trials, children reliably chose the different-
hammer, and lock. type foils, M = .80, SE = .08, Wilcoxon T = 174.5, N = 20,
Before receiving any critical trials, the children received p < .005. This indicates that the children were readily able
three mutual exclusivity warm-up items to measure children’s to make mutual exclusivity judgments when doing so simply
ability to make mutual exclusivity judgments when doing required shifting a noun between two exemplars of the same
so only required shifting a noun between two exemplars of type— e.g., between two books.
the same type. For example, in one trial, children were Would children also make mutual exclusivity judgments on
taught that blicket referred to a book, and then had to choose the critical trials, when doing so would require shifting be-
whether tima referred to another book (same-type) or to a tween a verb and noun? Our dependent measure on the criti-
CD (different-type). We expected that children would reli- 1 Preliminary analyses for Experiments 1 and 2 did not find sig-
ably choose the foil item (e.g., the CD) on these trials if they nificant effects of gender, age, or side of presentation. We have thus
understood the task. Children that did not do so on the first excluded these factors from the analyses reported here.

68
cal trials was the proportion of times children excluded the ar-
Figure 1: Experiment 2 Critical Items (The pictures shown
tifact and chose the patient of the event—the object/substance
were displayed during the judgment phases of the trials).
the artifact was used to act upon. Children chose the patient
reliably more than chance (.5), M = .73, SE = .07, Wilcoxon
T = 148.5, N = 20, p < .005. This suggests that children
expected the first novel word—which had referred, as a verb,
to the action (e.g., shoveling)—to also refer, as a noun, to the
artifact that helped carry out the action (e.g., to the shovel).
Due to this expectation, children may have excluded the arti-
fact as the meaning of the second novel word, believing that
the first novel word already referred to it. This finding sug-
gests that children do not represent the nominal and verbal
uses of familiar words like shovel as separate lexical items.
Rather, children have formed lexical structures to support the
flexible extension of these words across their noun and verb
meanings.
Critically, children could not have succeeded on the criti- Each of the critical artifacts and patients appeared unique
cal trials simply by mapping the first novel word to the entire in shape and color, and also possessed novel functions (see
event they watched in the training video. Although such a Figure 1). Children learned of the functions of the novel ar-
mapping would allow the first novel word to apply to the arti- tifacts when the novel verbs were modeled in the videos. As
fact of the event, it would also allow it to apply to the patient before, children heard the novel verb used in the infinitive to
of the event. Thus, the fact that children preferentially ex- refer what Monkey was going to do (“He’s going to use it to
tended the first novel word to the artifact and not the patient wug this stuff to make a shape out of it”), in the progressive
suggests that children expect a verb for a familiar event (e.g., to refer to the action as it was ongoing (“Monkey’s wugging
shoveling) to refer to a specific constituent of that event—its the stuff”), and in the past tense to refer the completed ac-
artifact (e.g., the shovel) but not its patient (e.g., the sand). tion (“Monkey wugged the stuff and made a shape out of it”).
The video also described how the affordances of the artifact
Experiment 2 facilitate the action: e.g., “What monkey has is sharp on the
Here, we examine whether the structures that encode the rela- bottom and it has a handle that Monkey can hold onto.”2 Im-
tions between the nominal and verbal uses of familiar words mediately after the training phase, we initiated the judgment
are productive. If they are, children should expect any word phase of the trial: e.g., “So that’s what wugging is. Now we
for an action to also refer to the artifact that helps carry out know what wugging is. But now I want the lum. Show me
that action. To test this possibility, we taught children novel the lum!” The pictures of the artifact and patient were then
words for actions involving novel artifacts that acted on novel presented and the child’s choice was recorded. To see how
patients. Would children expect the novel words for the ac- robust these extensions are, the pictures depicted different to-
tions to also refer to the artifacts? If so, they should exclude kens of artifacts and patients of the same category as those
the artifacts and choose the patients when asked to determine used in the videos. For example, the artifact in the wugging
the referents of the second novel words. item had different colored blocks attached to it, and the pa-
tient in the daxing item was composed of a different-colored
Methods clay (see Figure 1).
Participants The participants were 20 children (8 girls) be- Results and Discussion
tween the ages of 4;0 and 5;10 (M = 58 months). Four addi-
tional children participated but were excluded for failing the As in Experiment 1, children were readily able to make mu-
initial trials that gauged their understanding of the task. tual exclusivity judgments when doing so required shifting a
noun between exemplars from the same type. On the trials
Materials and Procedure All aspects of the materials and where there was a same-type option, children reliably chose
the procedure were the same except that different critical the different-type foils over the same-type matches, M = .94,
items were used. Rather than depicting actions involving fa- SE = .03, Wilcoxon T = 136, n = 17, p < .001. Our de-
miliar artifacts (e.g., shovels) acting upon familiar patients pendent measure on the critical trials was the proportion of
(e.g., sand), the critical items depicted novel artifacts that
2A control condition was run to rule out the possibility that the
acted upon novel patients. Because the critical items did not
active artifact was more salient, leading the child to associate it with
involve familiar artifacts, we did not pre-test children on their the novel word and succeed at test even without attending to the
knowledge of the nominal and verbal uses of familiar artifact syntax of the narration. In this condition, the same videos were pre-
words. Thus, the critical trials that children received were sented with alternate narrations in which the patient was labeled with
the novel word. When asked to identify the referent of the second
not restricted by their performance on an earlier pre-test—all novel label at test, children reliably chose the artifact (Wilcoxon T =
children in Experiment 2 received the same four critical trials. 37, n = 16, p =.06)

69
times children excluded the novel artifact and chose the novel noun/verb pairs in parental input.
patient of the event. Children chose the patient reliably more
than chance (.5), M = .78, SE = .06, Wilcoxon T = 126, n = Results and Discussion
17, p < .005. Children may have excluded the novel arti- The data were entered into a mixed effects logistic regression
facts because they expected the first novel words to instead with Age and Vocabulary (parent-reported familiarity with in-
refer to them. This suggests that children expect a word that strument noun/verb pairs) as between-subjects variables, and
labels a novel action to also label a particular constituent of Trial Type (Control or Critical) as a within-subjects variable.
that action—the novel artifact that helps carry it out. This Age (X2 (1, N=74) = 10.10, p < 0.01)) and Trial Type (X2
finding strengthens the conclusions of Experiment 1, by sug- (1, N=74) = 9.61, p < 0.01)) emerged as significant on this
gesting that the structures that support familiar flexible pairs analysis, indicating that children’s performance on the task
are productive and generalize to novel cases. Such produc- improved as they got older, and differed on Critical and Con-
tivity could facilitate children’s initial acquisition of corre- trol trials. Given the number of children for whom we did not
sponding noun/verb pairs, and could help explain why lexical have a vocabulary measure (10), and its lack of influence, we
innovations—like to Wikipedia—are often created. removed this variable, enabling us to analyze all the children
tested in Experiment 3 (n=84). This new model found signif-
Experiment 3 icant effects of Age (X2 (1, N=84) = 17.2, p < 0.001)) and
In Experiment 2, we established that four- and five-year-olds Trial Type (X2 (1, N=84) = 17.40, p < 0.001)), as well as an
are able to extend a novel word that labels a novel action to interaction between the two (X2 (1, N=84) = 4.28, p < 0.05)).
the artifact performing that action, suggesting that the struc- The interaction captures the difference in rate of improvement
tures that encode the relationships between the nominal and across this age range for the two trial types. While perfor-
verbal uses of familiar words are productive. Here, we ex- mance on control trials dramatically improved from three to
amine the developmental trajectory of this generative ability five years, critical trial performance did so much more slowly.
and explore whether it may be linked to prior experience with A one-way ANOVA revealed that though the vocabulary mea-
flexible noun/verb pairs, like hammer and shovel. sures were not predictive of trial performance, they did im-
prove significantly with age F(1, 74) = 2.54, p < 0.001.
Methods Non-parametric tests confirm the success of later ages on
Participants The participants were 84 monolingual chil- the task. While four- and five-year-olds reliably chose the
dren (42 girls) between the ages of 3;0 and 5;11 (M = 53.8 patient of the verb, rather than the artifact (four-year-olds:
months, SD = 10.2), including 29 three-year-olds (M = 42.6 Wilcoxon W = 16240, n = 29, p < 0.001; five-year-olds: W
months, SD = 3.9), 29 four-year-olds (M = 53.8 months, SD = 13802, n = 26, p < 0.001), three-year-olds performed at
= 3.2), and 26 five-year-olds (M = 66.4 months, SD = 3.8). chance on the critical trials (0.5). That three-year-olds per-
Fourteen participants were excluded due to experimenter or formed above chance for the controls (Wilcoxon W = 16472,
technical error (10), interference resulting from a distracting n = 29, p < .01) suggests either that our task was not sensi-
testing environment (3), or because they had witnessed an- tive enough to reflect their knowledge of this productive re-
other child participate previously (1). lationship, or that the expectation of this predictable lexical
extension might emerge sometime in the third year of devel-
Materials and Procedure The procedure for this experi- opment.
ment was identical to Experiment 2, with the following alter-
ations. An additional mutual exclusivity warm-up trial was
added. These four warm-up trials were instead used as con- Figure 2: Proportion of patient and different-type choices on
trol trials, serving as a predictor of children’s understanding critical and control trials, respectively, by age.
of the task in our analyses. Consequently, unlike in Experi-
ments 1 and 2, children did not receive feedback on any of
their choices in these trials, and children were not excluded
on the basis of their performance. A final distinction in the
procedure for Experiment 3 is that the majority of the parents
of participants filled out a vocabulary survey, which consisted
of 33 flexible noun/verb pairs (derived in part from Clark &
Clark, 1979). The survey probed parental report of produc-
tion and comprehension separately (e.g., for the noun/verb
pair bike, parents were asked to report whether their child
understood the noun form bike, the verb form to bike, and
whether they produced each of these forms). The addition
of this survey was motivated by the finding in Lippeveld and
Oshima-Takane (2010) that extension between verb and ar-
tifact labels could be correlated with frequency of flexible

70
General Discussion referent of the second novel word. If we find that younger
children do not perform these spontaneous generalizations,
English includes many examples of words that label artifacts,
we could then examine whether children understand the re-
as nouns, and functional uses of those artifacts, as verbs. The
lationship between the different meanings of familiar flexi-
present studies indicate that these examples are not etymolog-
ble verb/noun pairs or whether they treat these words as un-
ical relics, but instead reveal a productive linguistic structure.
related homophones (as we did in Experiment 1). In doing
In Experiment 1, we showed that four-and five-year-old so, we will investigate whether children initially form usage-
children do not represent the nominal and verbal uses of fa- based “islands of flexibility” or whether these productive lex-
miliar words as separate, unrelated words, but instead de- ical structures are present to support word-learning.
rive them from common lexical structures. Specifically, after
learning that a novel verb referred to a familiar action, chil- Acknowledgments
dren excluded the artifact involved in that action when deter- We thank our undergraduate researchers at UC Berkeley and
mining the referent of a second novel word. Children instead UC San Diego for assistance with stimuli construction, partic-
chose the patient of the action, suggesting that they expect the ipant recruitment, and data collection. We are grateful to the
word labeling an event to also refer to a specific constituent of families of the children who participated in the San Diego,
that event—its artifact. In Experiment 2, we showed that the Vancouver Island, and Berkeley areas. This work was sup-
structures four-and five.year-olds deploy to capture the flex- ported by a grant from the James S. McDonnell Foundation
ibility of familiar verbs are productive: after learning that a to DB.
novel verb referred to a novel action, children in Experiment
2 excluded the novel artifact of that action as the meaning of References
a second novel word. This suggests that four-and five-year- Berman, R. (1999). Children’s innovative verbs versus
olds spontaneously expect any word for an action to also re- nouns: Structured elicitations and spontaneous coinages.
fer to the artifact that helps carry out that action. Finally, In L. Menn & N. B. Rattner (Eds.), Methods for studying
in Experiment 3, we found that while four-and five-year-olds language production (pp. 69–93). Mahwah, NJ: Erlbaum.
seem to have this productive expectation, three-year-olds do Bushnell, E. W., & Maratsos, M. P. (1984). Spooning and
not. However, given that we observed a significant relation- basketing: Children’s dealing with accidental gaps in the
ship between age and performance on the control trials, it is lexicon. Child Development, 55, 893–890.
possible that three-year-olds, and even younger age groups Caslet, K., & Kelemen, D. (2007). Reasoning about arti-
are able to generalize between noun and verb meanings, but facts at 24 months: The developing teleo-functional stance.
that our experimental measures were not sensitive enough to Cognition, 103(120–130).
detect this. One reason to believe that three-year-olds may Clark, E. V. (1982). The young word maker: A case study of
be capable of forming such generalizations is that children of innovation in the child’s lexicon. In E. Wanner & L. Gleit-
this age often create new verb from nouns (and nouns from man (Eds.), Language acquisition (pp. 390–428). Cam-
verbs) in their spontaneous speech (e.g., Clark, 1982). bridge: Cambridge University Press.
At stake in the question of when children begin to form Clark, E. V., & Clark, H. H. (1979). When nouns surface as
generalizations about flexible word use is whether such gen- verbs. Language, 55(4), 767–811.
eralizations could play a role in facilitating lexical develop- Copestake, A., & Briscoe, T. (1995). Semi-productive poly-
ment. As noted in the Introduction, if children recognize the semy and sense extension. Journal of Semantics, 12, 15–
special relationship between shared labels for actions and ar- 67.
tifacts early on, this could be quite powerful for language ac- Diesendruck, G., & Markson, L. (2001). Children’s avoid-
quisition because children would need to learn only one word ance of lexical overlap: A pragmatic account. Develop-
to express multiple meanings. If, on the other hand, such mental Psychology, 37(5), 630–641.
inferences arise only later in life—and are constructed only Lippeveld, M., & Oshima-Takane, Y. (2010). The dax is dax-
gradually after exposure to several instances of such flexible ing the cheese: When do children acquire class extension
word pair patterns—they will not play as large of a role in rules for denominal verbs? Boston, MA: Paper presented
lexical development. at the 35th Annual Boston University Conference on Lan-
In future research, we intend to explore how children come guage Development.
to form generalizations about flexible word use, by using a Pustejovsky, J. (1995). The generative lexicon. Cambridge:
more sensitive dependent measure, such as preferential look- MIT Press.
ing. For example, we could teach children a novel word to Srinivasan, M., & Snedeker, J. (2011). Judging a book by
label a novel action (as in Experiments 2 and 3), and then its cover and its contents: The respresentation of polyse-
instruct them to look at the referent of a second novel word. mous and homophonous meanings in four-year-old chil-
If children are able to spontaneously generalize that the first dren. Cognitive Psychology, 62, 245–272.
novel word can refer to both the action and the artifact per- Tomasello, M. (2003). Constructing a language: A usage-
forming that action, then we would expect them to gaze at based theory of language acquisition. Harvard University
the acted-upon item (not the artifact) when told to look at the Press.

71
Visuo-Spatial Memory Processing and the Visual Impedance Effect
Rebecca Albrecht (albrechr@cs.uni-freiburg.de)
Center for Cognitive Science, University of Freiburg

Holger Schultheis (schulth@informatik.uni-bremen.de)


Cognitive Systems, University of Bremen

Wai-Tat Fu (wfu@illinois.edu)
Department of Computer Science, University of Illinois at Urbana-Champaign

Abstract Barkowsky, 2013, for an in-depth discussion of this point)


remains largely to be determined.
Models of spatial reasoning often assume distinct visual and
spatial representations. In particular, the visual impedance ef- In this paper we argue that previous research has not suf-
fect – slower response time when more visual details are rep- ficiently considered the role of memory when studying and
resented in three-term series spatial reasoning tasks – has been
taken as evidence for the distinctive roles of visual and spa- comparing visual and spatial representations. We demon-
tial representations. In this paper, we show that a memory strate our case by presenting a memory model that explains
model of spreading activation based on the ACT-R architec- the visual impedance effect without the assumption that vi-
ture can explain the visual impedance effect without the as-
sumption of distinct visual and spatial representations. Using sual and spatial representations have distinctive functional
the same memory representation, varying levels of visual fea- roles in spatial reasoning. Instead, the model assumes that vi-
tures associated with an object are represented in the model. sual and spatial information can both be represented similarly
The visual impedance effect is explained by the spreading ac-
tivation mechanism of ACT-R. The model not only provides as memory items. However, the model predicts that some-
a more parsimonious explanation to the visual impedance ef- times additional visual details may slow down the mainte-
fect, but also leads to testable predictions of a wide range of nance of the memory representations of the informaton. This
memory effects in spatial reasoning.
is different from the argument by Knauff and Johnson-Laird
Keywords: Visual impedance, memory processing, scalable Knauff and Johnson-Laird (2002), who argued that visual
representation, spreading activation, ACT-R, relational reason-
ing, mental model theory. representations of information may slow down the reason-
ing process. As a result, contrary to their arguments Knauff
Introduction and Johnson-Laird (2002), the visual impedance effect does
Processing visual and spatial information is among the most not provide any support to the claim that visual and spatial
crucial human abilities, because it permeates virtually every- relations are represented distinctively, nor does it imply that
thing we do (imagine moving in / through the environment an abstract spatial mental model can lead to a faster reason-
without being able to process the visual and spatial informa- ing process. Our model not only provides a more parsimo-
tion available from you surroundings). nious explanation to the visual impedance effect, but it also
In a seminal paper Ungerleider and Mishkin (1982) argue has the advantages of having more generality and continuity
that in the primate brain two separate pathways are responsi- with other theories in cognitive sciences.
ble for processing visuo-spatial information: The what path-
Visual Impedance
way and the where pathway which are associated with the
temporal and parietal lobe, respectively. The what pathway Three-term series problems (P. N. Johnson-Laird, 1972) have
mainly processes information related to object identification played a prominent role in investigating spatial and visual
and recognition (e.g., color), while the where pathway mainly representations (e.g., Shaver, Pierson, & Lang, 1975; Knauff
processes spatial information (e.g., object location or move- & Johnson-Laird, 2002; Rauh, Hagen, Kuss, Schlieder,
ment). This distinction has subsequently received additional & Strube, 2005; Schultheis, Bertel, & Barkowsky, 2014;
support from many behavioural and neuroscientific studies Schultheis & Barkowsky, 2013; Sima et al., 2013). A three-
(e.g., Milner & Goodale, 2008; Klauer & Zhao, 2004). term series problem constitutes a deductive relational reason-
The existence of these two distinct neural pathways has ing problem in which the relation between two objects, A and
given rise to the assumption that visuo-spatial information C, has to be inferred given the relations between objects A
processing in humans draws on two distinct types of men- and B as well as the relation between B and C. For example,
tal representations: Visual and spatial representations. Al- given the information that (a) the dog is left of the cat and
though this assumption is shared by the two main theories of (b) the mouse is left of the dog, participants may be asked to
visuo-spatial information processing, the mental model the- verify the statement that the mouse is left of the cat. Simi-
ory (P. Johnson-Laird, 1998) and the theory of mental im- larly, knowing that (a) the dog is dirtier than the cat and (b)
agery (Kosslyn, Thompson, & Ganis, 2006), the nature of the mouse is dirtier than the dog, one can verify the statement
the representations and, in particular, the relation between that the mouse is dirtier than the cat.
the two types of representations (see Sima, Schultheis, & Knauff and Johnson-Laird (2002) conducted experiments

72
in which they compare participants performance in solving relation-type more-than
such three-term series for different types of relations. Specif-
ically, Knauff and Johnson-Laird (2002) distinguish between o1 p o2 o1 p o2
visual, visuo-spatial, and control relations: Visual relations
are relations that are easy to envisage visually (e.g., dirtier);
visuo-spatial relations are relations that are easy to envisage A content B hat left tie
spatially and visually (e.g., to the left of); control relations
(a) Abstract (i.e. not instan- (b) Instantiated relation with
are relations that are hard to envisage both spatially and vi- tiated) representation of rela- relation-type more-than and
sually (e.g., better). The main finding reported by Knauff tional information. property (p) left.
and Johnson-Laird (2002) is that reasoning about visual rela-
tions takes significantly more time than reasoning about ei- Figure 1: Uniform representation of relational information. A
ther visuo-spatial or control relations. This comparatively relation consists of a relation-type, two objects (o1 , o2 ), and a
poor performance of reasoning with visual representations property (p).
has been termed the visual impedance effect.
The explanation provided for the visual impedance effect sentation with ACT-R’s spreading activation mechanism. We
assumes that for all types of relations the actual reasoning illustrate how our model explains the visual impedance ef-
process involves (spatial) mental models (Knauff, Fangmeier, fect and present an ACT-R implementation and simulation of
Ruff, & Johnson-Laird, 2003). For visuo-spatial relations our model. We conclude in highlighting the contribution of
and control relations the given information is directly repre- our modeling work as well as interesting questions for future
sented in such a spatial mental model and, thus, can imme- work.
diately be used for reasoning. For visual relations, however,
the given information is initially represented by a visual rep- Memory Representation
resentation (e.g., a visual mental image) that does not sup-
In order to understand the dependency between a represen-
port reasoning. To solve the reasoning problem, an additional
tation of relational information in memory on the one hand
step for building a spatial mental model is required (Knauff,
and a reasoning process on the other we will first introduce a
2009). Note that this explanation of the visual impedance ef-
scalable, abstract representation for relational information. In
fect assumes that the comparatively poor performance with
particular, our abstract representation distinguishes between
visual relations is due to problems associated with the rea-
the relation in a mathematical sense, i.e., as it is deemed suit-
soning process. Against this background it seems remark-
able for reasoning, and the meaning of a relation.
able that all available computational models that formalise
To abstractly represent relational information of the type
reasoning with spatial mental models do not explain the vi-
employed in three-term series problems, we consider more-
sual impedance effect (Krumnack, Bucher, Nejasmic, Nebel,
than relation types. A more-than relation type consists of
& Knauff, 2011; Ragni & Knauff, 2013; Khemlani, Trafton,
three different pieces of information, two objects (o1 , o2 ) and
Lotstein, & Johnson-Laird, 2012).
a property (p), i.e. more-than(o1 , p, o2 ). An example is de-
We propose that the visual impedance effect is not a rea- picted in Figure 1). Intuitively, our representation can be un-
soning effect, but a memory effect. Following Schultheis and derstood as “object o1 has more of a property p than object
Barkowsky (2011), we assume that spatial and visual repre- o2 ”. Or more concretely, “the hat is more left than the tie”
sentations are not two distinct, qualitatively different types for the visuo-spatial relation “left” and “the hat is more dirty
of representations, but that visuo-spatial representations lie than the tie” for the visual relation “dirty” (cf. Figure 1).
along a continuum that characterizes how specifically amodal We further assume that for a concrete relational statement
/ spatial or modality-specific / visual a representation is. De- each of the three arguments is associated with features in
pending on the current task context, representations can flex- memory that represent the arguments’ meaning. We define
ibly scale along this continuum (i.e., become more or less vi- the content of an object (or property) as the tuple of fea-
sual) without a need to modify the reasoning processes work- tures necessary to represent the object o, i.e. content(o) =
ing on the representations. Given such scalable representa- p p
( f1o , . . . , fno ), and property p, i.e. content(p) = ( f1 , . . . , fn ),
tions, we assume that visual relations give rise to more com- in memory. Figure 2 shows graphically how relational state-
plex representations, because more visual details are repre- ments and their contents are represented. This representa-
sented. Specifically, due to spreading activation of memory tion includes an abstract representation suitable for reasoning
items, the additional visual details slower down the access to (Figure 2, above the red line) and the memory representation
the information that needs to be processed during the later defined by the content of the relational information (Figure 2,
reasoning process; and, thus, slows down the overall reaction below the red line).
times when solving the three-term series problem. Scalability of this representation is defined in terms of the
In the following we first introduce a scalable, abstract rep- number of features involved in representing the relational
resentation for relational information that supports reason- statement. In particular, the representation can scale to be-
ing. We then present our model that combines this repre- come more or less visual depending on how much visual fea-

73
more-than more-than

o1 p o2 o1 p o2

A relation B relation C

hat dirty tie dirty shoe

f1hat ... fnhat dirty


f1 ... dirty
f1tie ... fntie dirty
f1 dirty
f1shoe ... fnshoe
fn fn

Figure 2: Representation of relational information ’the hat is dirtier than the tie’ and ’the tie is dirtier than the shoe’ in memory
assuming a hierarchical representation of information. Above the red line is a representation which supports reasoning with
relations. Below the red line is the memory representation of the objects and property used to elaborate the content of relational
information. Scalability is defined in terms of the number of features necessary to represent relational information.

tures are associated with the relational statement. For exam- d which are a part of the current working memory state (i.e.,
ple, the relation “dirtier” may be associated with features like assigned to a buffer). Formally, the signal strength between
“dirt”, “mud”, “black dots”, etc. and, thus yield a more visual chunk c and a chunk d is computed as Sd,c = S − ln( f and ).
representation than the relation “to the left of”, which may The signal strength depends on the number of outgoing con-
only be associated with a single feature “position”. nections of chunk d, a concept which has been termed f an of
chunk d. The signal strength additionally depends on a global
Cognitive Model constant S which has been interpreted psychologically as an
In this section we describe an ACT-R model that explains the approximation of the declarative memory size. The complete
visual impedance effect. Employing the above described rep- spreading activation of a chunk c is calculated by 1
resentation, the model explains the effect as a memory phe-
nomenon arising from spreading activation. sa(c) = ∑ Sd,c
d in working memory
ACT-R Spreading Activation
The time to retrieve a chunk c from declarative memory
ACT-R realises working memory as a structured set of buffers
is defined with respect to the activation of chunk c, in our
(Anderson, 2007). Buffers hold declarative information, so-
case RT (c) = a · e−sa(c) , where a is a constant. The higher
called chunks. A chunk is a set of key-value (or slot-value)
the activation of a chunk the lower the response time. For
pairs. For example, a chunk representation of the introduced
spreading activation a greater fan implies a lower activation
’more-than’ relation has three slots (o1 , p, o2 ) to which values
and, thus, a higher response time.
(often also chunks) can be assigned. Behaviour in ACT-R is
produced by the repeated application of production rules that
fit a current working memory state and change the working Example. It may be helpful to more closely consider
memory state according to their definition. Changes to the how the spreading activation mechanism explains the visual
working memory come about by requests to modules that are impedance effect. The visual impedance effect is measured
associated with buffers. Modules process requests by updat- in the time to verify a given conclusion. When a conclusion
ing the chunks contained in their buffers. This processing is needs to be verified, information from the first and second
associated with a time cost and, in some cases, has an uncer- relational statement have already been integrated in a mental
tain outcome. representation. Depending on assumptions stated in a reason-
The ACT-R declarative module (sometimes called declar- ing theory this mental representation may, for example, be a
ative memory) holds all declarative information known to a mental model (Ragni & Knauff, 2013) or a relational infer-
model such as, for example, the complete representation de- ence, i.e., a relational statement, (Braine & O’Brien, 1998).
picted in Figure 2. The time it takes the declarative module to In either case, this mental representation needs to be retrieved
process requests depends on the activation values assigned to from declarative memory in order to verify the conclusion. In
candidate chunks. While a number of mechanisms can influ- 1 For the sake of representation simplicity we assume that spread-
ence the chunks’ activations, we focus our analysis to spread- ing activation is enabled for every buffer and that all buffers are as-
ing activation. signed the same weight, which sum up to a total of 1. Addition-
ally, in our scenario the working memory holds the same number of
The spreading activation of a chunk c in ACT-R is defined chunks in every request. Thus, we leave out the weight of the ACT-R
in terms of a signal strength S between chunk c and all chunks spreading activation equation.

74
the following, we will use a mental model as the mental repre- f1h
f1h
sentation. The argument for a relational inference is analogue hat ...
...
hat
and the simulation results for the retrieval time only differ by fnh
fnh
a constant factor due to the different representation.
Consider our scalable representation defined for relational f1h model:
model:
pos1 : hat
information, i.e., more-than(o1 , p, o2 ). When a mental model dirtier ... pos2 : tie
pos1 : hat
left of f1 pos2 : tie
is requested from the declarative module the conclusion is fnh
pos3 : shoe
pos3 : shoe
axis: dirtier
represented in the model’s working memory (e.g., the goal axis: le f to f
buffer). Thus, o1 , o2 and p are potential sources for spread- f1h
f1h
ing activation (cp. Figure 3). The signal strength between the shoe ...
shoe ...
mental model chunk (model) and the content of a relation p
fnh
is then S p,model = S − ln( f an(p)) + r. The fan is influenced fnh
by the number of features associated with the relation, i.e. (a) Fan for chunks in work-
f an(p) = content(p), and the constant r approximating the ing memory for conclusion re- (b) Fan for chunks in work-
lation ”dirtier”. The content ing memory for conclusion re-
fan associated with reasoning representation (e.g. a second lation ”to the left of”. The con-
of relation ”dirtier” needs to be tent of relation ”to the left of”
mental model in declarative memory). The signal strength represented by more than one
can thus be calculated as S p,model = S − ln(content(p)). Con- feature (e.g., dirty, mud, black can be represented by a single
spots, etc. feature (e.g., the position).
sequently, the more features are necessary to represent the
content of a relation, the higher the retrieval time of a target
Figure 3: Example illustration of memory representations for
chunk (due to the higher fan).
visual (a) and visual-spatial relations. Chunks active in the
Now consider the concrete relations introduced for the vi-
working memory are ”hat”, ”shoe”, and ”dirtier” (3a), and
sual impedance effect, that is, visual relations like “dirtier”,
”to the left of” (3b). The target chunk is a mental model that
visuo-spatial relations like “to the left of” and control rela-
needs to be retrieved in order to verify the conclusion. Due to
tions like “better”. If we assume that the visuo-spatial relation
the higher fan the target chunk (here a mental model) receives
“to the left of” can be represented using one feature (e.g., the
more spreading activation for the conclusion relation ”to the
position), the fan is f an(left of) = content(left of) = 1 (Fig-
left of” than for the conclusion relation ”dirtier”. Thus, the
ure 3b). For visual relations like “dirtier”, on the other hand,
retrieval time is higher for the conclusion relation ”dirtier”.
more features need to be represented (e.g., dirt, mud, etc.)
Therefore, the fan associated with “dirtier” is higher than
the fan associated with “to the left-of”, i.e. f an(dirtier) > Brüssow, 2010), which assumes that exactly one retrieval is
f an(left of), and the signal strength between chunks “dirtier” necessary to verify a given conclusion.
and model is lower than between “to the left of” and model, For a prototypical task such as “the hat is dirtier than the
i.e. Sdirtier,model < Sleft of,model (Figure 3a). Assuming that the tie”, the “tie is dirtier than the shoe”, “is the hat dirtier than
features representing objects o1 and o2 are the same in both the shoe?” we define the mental representation as either a
cases2 , the spreading activation that the mental model chunk mental model chunk or a relational inference chunk which is
receives from the visual relation “dirtier” is given by stored in declarative memory, e.g.,
sa(model) = S − ln(k) + S − ln(n) + S − ln(l) • (r1 ISA model pos1 hat pos2 tie pos3 shoe rel dirtier)
| {z } | {z } | {z }
Shat,model Sdirtier,model Sshoe,model • (r2 ISA inference o1 hat o2 shoe rel dirtier)

Additionally we represent the features associated with ob-


and from the spatial relation “to the left of” is given by
jects and the property of a relation as content chunks, e.g.,
sa(model) = S − ln(k) + S − ln(1) + S − ln(l) • (l1 ISA content id left-of feature l1)
| {z } | {z } | {z }
Shat,model Sleft of,model Sshoe,model • (d1 ISA content id dirtier feature dd)

Obviously, the mental model chunk receives more spreading Source of spreading activation is a representation of the
activation for visuo-spatial relations. Therefore, the retrieval conclusion in the goal buffer, e.g.,
time is lower (see Figure 4). (rel1 ISA more-than o1 hat p dirtier o2 shoe)
We define one production rule which requests a mental
ACT-R implementation. Our ACT-R model is based on representation chunk (model or inference) from declarative
the ACT-R implementation of PRISM (Ragni, Fangmeier, & memory. This request does not specify any restrictions on
slot values other than the type being either a mental model
2 Instead of keeping the features of the objects fixed and varying or an inference. The time it takes the declarative module to
the features of the relation, it would also be possible to represent answer this request depends on the fan associated with the
the objects by more of less features depending on the relation type.
This would not impact the explanatory power of our model w.r.t the objects and the property of the relation, that is, the number of
visual impedance effect. content chunks associated with the objects and the property.

75
Our work shows that combining the concept of scalable
representation structures with spreading activation provides

a more parsimonious explanation to the visual impedance
effect, as the proposed model does not assume distinct vi-
560

● sual and spatial representations or a specific reasoning pro-


550
Retrieval Time (ms)

cess. The current model uses memory representation of ob-


jects and memory processes that have been used to explain a
540

wide range of memory effects (e.g., in previous ACT-R mod-


530


els of memory tasks). The current model therefore has the
potential to lead to a wide range of testable predictions on
520

the effects of memory in spatial reasoning, such as effects of


510

● individual differences in working memory capacity, interfer-


1 2 3 4 5
ence effects, or effects of memory decay. In addition to the
original visual impedance effect (Knauff & Johnson-Laird,
Number of Features
2002), our modeling work also explains moderations of the
effect that have been reported. If the visual impedance ef-
Figure 4: Retrieval time for a mental representation chunk fect is due to memory processing as assumed in the proposed
(here mental model chunk) when a conclusion needs to be model, it should scale with the model’s ability and necessity
verified. The response time increases independently of a con- to represent specific features in order to maintain a represen-
crete task or reasoning theory with the number of visual fea- tation suitable for reasoning. Consistent with our model, re-
tures associated with the objects and the relation. search shows that blind people show no visual impedance ef-
fect (Knauff, 2009)— perhaps because they are less inclined
to represent objects with visual features, or the number of
All ACT-R parameters are set to their default values. We visual features tend to be lower for blind people. Further-
approximate parameter S by the logarithm of the average size more, people who have a higher tendency to visualize object
of the model’s declarative memory (i.e. S = 3). Figure 4 details show a stronger visual impedance effect (Castañeda
shows how the retrieval time increases with the number of & Knauff, 2013), because they tend to represent more visual
features associated with the content of a relation. features as other groups.
Accordingly, our model accounts for the visual impedance The proposed cognitive model investigates the impact of
effect by predicting that verification of conclusions for three- the memory representations of visual features on spatial rea-
term series problems involving visual relations such as “dirt- soning. The model, however, does not make any assump-
ier” take more time than for problems involving visuo-spatial tion on the reasoning process, as the reasoning tasks are the
relations. Interestingly, our model predicts that the visual same across the different conditions in the studies on vi-
impedance effect should not be restricted to the use of visual sual impedance (Knauff & Johnson-Laird, 2002). In other
relations, but should arise whenever the reasoner is inclined words, the explanation of the visual impedance effect by our
to associate multiple (visual) features with a relational state- model is independent of the reasoning process. For exam-
ment. As discussed below, existing evidence supports this ple, if we apply the reasoning process in the PRISM model
prediction. (Ragni et al., 2010), in which the main difference in level
of difficulty in spatial reasoning tasks is characterized by the
Conclusion number of focus operations on the represented objects, we
will have the same number of focus operations in each con-
Knauff and Johnson-Laird argued that the reason why items
dition, and the only difference is how quickly the model can
that could easily be envisaged would lead to slower response
asscess the objects represented in memory as the focus op-
times was that visual representations of irrelevant features
erations are applied. However, we should point out that the
slowed down the reasoning process. We provided an alter-
PRISM model by itself does not seem to be able to explain
native explanation: the easily envisaged items took longer to
the visual impedance effect. On the other hand, our model
be accessed in memory because they were associated with
can be used with other reasoning theories (e.g., (Krumnack
more visual features, which slowed down their access time
et al., 2011; Braine & O’Brien, 1998)) to explain the visual
as predicted by the spreading activation mechanism. Con-
impedance effects. In other words, our model suggests that
trary to the argument by Knauff and Johnson-Laird, we did
the visual impedance effect can be explained by memory pro-
not find that the visual impedance effect provided any sup-
cesses rather than reasoning processes.
port to the claim that easily envisaged items were represented
by a visual representation that was funcational different from The goal of this paper is to show that the visual impedance
a (spatial) mental model, nor did the results support the claim effect can be explained without committing to a unique spa-
that “visual imagery as the medium for reasoning would be tial representation that is distinct from visual representation.
implausible” (Knauff & Johnson-Laird, 2002). This is consistent with the idea that the long debate about the

76
role of visual imagery in spatial reasoning can be resolved by Johnson-Laird, P. N. (1972). The three-term series problem.
considering the visuo-spatial representation as a continuum Cognition, 1, 57–82.
Schultheis and Barkowsky (2011), with varying levels of vi- Khemlani, S., Trafton, J. G., Lotstein, M., & Johnson-Laird,
sual (and spatial) features represented in memory. Another P. (2012). A process model of immediate inferences. In
advantage of this approach is that by utilizing memory repre- Proceedings of the 11th international conference on cogni-
sentations and mechanisms, the model is more readily com- tive modeling (p. 151).
pared and tested against a wide range of cognitive phenomena Klauer, K. C., & Zhao, Z. (2004). Double dissociations in
beyond spatial reasoning. We believe that our modeling ap- visual and spatial short-term memory. Journal of Experi-
proach and results constitute an important first step towards mental Psychology: General, 133(3), 355–381.
studying the impact of memory processing in human reason- Knauff, M. (2009). A neuro-cognitive theory of deductive
ing and the nature of spatial and visual representations. While relational reasoning with mental models and visual images.
previous theories and studies mostly restricted considerations Spatial Cognition & Computation, 9(2), 109–137.
to the reasoning process or the representation, we define a Knauff, M., Fangmeier, T., Ruff, C. C., & Johnson-Laird,
link between these concepts. As a result, our approach also P. N. (2003). Reasoning, models, and images: Behav-
highlights promising avenues for future work, both empirical ioral measures and cortical activity. Journal of Cognitive
and computational, to shed more light on aspects of reasoning Neuroscience, 15(4), 559-573.
processes and representations. Knauff, M., & Johnson-Laird, P. (2002). Visual imagery can
Empirically, we propose experiments that explicitly con- impede reasoning. Memory & Cognition, 30(3), 363–371.
trol the number of represented features, both of objects and Kosslyn, S. M., Thompson, W. L., & Ganis, G. (2006). The
relations. Such an experiment would yield valuable results case for mental imagery. New York: OUP.
on the effect size with respect to the number of features nec- Krumnack, A., Bucher, L., Nejasmic, J., Nebel, B., & Knauff,
essary to represent concepts. A possible approach is using a M. (2011). A model for relational reasoning as verbal
high and low similarity conditions similar to Folk and Luce reasoning. Cognitive Systems Research, 11, 377-392.
(1987), that is, where more or less features need to be rep- Milner, A., & Goodale, M. (2008). Two visual systems re-
resented in order to draw conclusions (e.g., “the red hat is viewed. Neuropsychologia, 46(3), 774 - 785.
dirtier than the tie” vs. “the red hat is dirtier than the red tie” Ragni, M., Fangmeier, T., & Brüssow, S. (2010). Deductive
vs. “the red hat is dirtier than the blue tie”). spatial reasoning: From neurological evidence to a cogni-
Computationally, assuming visual impedance is in fact an tive model. In Proceedings of the 10th international con-
effect in memory processing our results can be used to further ference on cognitive modeling (pp. 193–198).
examine reasoning theories. The ACT-R implementation of Ragni, M., & Knauff, M. (2013). A theory and a compu-
the PRISM model represents the complete mental model in tational model of spatial reasoning with preferred mental
one chunk and approximates focus operations as a constant models. Psychological review, 120(3), 561.
factor. However, according to the ACT-R theory information Rauh, R., Hagen, C., Kuss, T., Schlieder, C., & Strube, G.
is usually represented as linked lists. Thus, if a mental model (2005). Preferred and alternative mental models in spatial
was represented as a linked list a focus operation would in fact reasoning. Spatial Cognition & Computation, 5(2), 239–
be a request to declarative memory. In this case, our memory 269.
model would predict a linear increase in the response time for Schultheis, H., & Barkowsky, T. (2011). Casimir: an archi-
visuo-spatial relations with the number of focus operations. tecture for mental spatial knowledge processing. Topics in
Cognitive Science, 3(4), 778–795.
References Schultheis, H., & Barkowsky, T. (2013). Variable stability
of preferences in spatial reasoning. Cognitive Processing,
Anderson, J. R. (2007). How can the human mind occur in 14(2), 209 - 211.
the physical universe? Oxford University Press. Schultheis, H., Bertel, S., & Barkowsky, T. (2014). Modeling
Braine, M. D., & O’Brien, D. P. (1998). Mental logic. Psy- mental spatial reasoning about cardinal directions. Cogni-
chology Press. tive Science, 38(8), 1521–1561.
Castañeda, L. E. G., & Knauff, M. (2013). Individual dif- Shaver, P., Pierson, L., & Lang, S. (1975). Converging evi-
ferences, imagery and the visual impedance effect. In dence for the functional significance of imagery in problem
M. Knauff, M. Pauen, N. Sebanz, & I. Wachsmuth (Eds.), solving. Cognition, 3(4), 359 - 375.
Proceedings of the 35rd annual conference of the cognitive Sima, J. F., Schultheis, H., & Barkowsky, T. (2013). Dif-
science society. Austin, TX: Cognitive Science Society. ferences between spatial and visual mental representations.
Folk, M. D., & Luce, R. D. (1987). Effects of stimulus Frontiers in psychology, 4.
complexity on mental rotation rate of polygons. Journal Ungerleider, L., & Mishkin, M. (1982). Two cortical systems.
of Experimental Psychology: Human Perception & Perfor- Analysis of Visual Behavior, 549–586.
mance, 13, 395 - 404.
Johnson-Laird, P. (1998). Imagery, visualization, and think-
ing. Perception and cognition at centurys end, 441–467.

77
Change your Mind: Investigating the Effects of
Self-Explanation in the Resolution of Misconceptions
Laura K. Allen (LauraKAllen@asu.edu)
Danielle S. McNamara (DSMcnama@asu.edu)
Arizona State University, Department of Psychology, P.O. Box 872111
Tempe, AZ 85281 USA

Matthew T. McCrudden (Matt.McCrudden@vuw.ac.nz)


Victoria University of Wellington, School of Education, P.O. Box 17-310
Wellington 6147, New Zealand

Abstract held by the reader (Kendeou & van den Broek, 2007; van
We investigated the differential effects of self-explaining a
den Broek & Kendeou, 2008). The co-activation of both
refutational text, compared to thinking aloud or rereading. correct and incorrect information presumably increases the
Undergraduate students (n = 105) read a refutational text likelihood that readers recognize inaccuracies in their
about natural selection and were asked to either self-explain, understandings and work to revise their misconceptions.
think-aloud, or re-read the text. Then they completed a Support for the co-activation hypothesis primarily stems
posttest that assessed general knowledge of natural selection. from research investigating the cognitive processes that take
Students who self-explained the refutational text subsequently place while reading refutational texts (Kendeou, Muis, &
outperformed their peers on a test of their knowledge of
natural selection. Additionally, the results suggest that both Fulton, 2011; Kendeou & van den Broek, 2007; McCrudden
instructional and performance differences were significantly & Kendeou, 2014; van den Broek & Kendeou, 2008). Prior
linked to the degree of causal cohesion present within studies, for instance, have demonstrated that readers allocate
students’ natural language responses to the text (i.e., self- more time to target sentences within refutational texts as
explanations and think-alouds). opposed to control versions of these texts. Additionally,
Keywords: comprehension; conceptual change; students who read refutational texts generate think-aloud
computational linguistics; cohesion; self-explanation; statements that are more indicative of conceptual change
strategies strategies (Kendeou & van den Broek, 2007; Kendeou et al.,
2011; van den Broek & Kendeou, 2008).
Introduction Despite these online processing differences, research on
Misconceptions emerge from our attempts to understand the the efficacy of these specialized texts to promote conceptual
world around us (Guzzetti et al., 1993). As a result, they change has been mixed. Although some research has shown
tend to be relatively intuitive and relate to our prior positive effects of these texts on retention of science
(accurate) knowledge reasonably well. Not surprisingly knowledge and inference-level performance (Ariasi &
then, these misconceptions can be difficult to recognize and Mason, 2011; Diakidoy, Kendeou, & Ioannides, 2003;
extremely resistant to change (van den Broek & Kendeou, Mason & Gava, 2007), other studies have reported null
2008). Importantly, misconceptions cause interference when results (Kendeou et al., 2011; Kendou & van den Broek,
we attempt to learn new and related information (Feltovich, 2007; Palmer, 2003). These mixed findings indicate that
Couson, & Spiro, 2001), which can pose serious problems conceptual change from refutational texts is not a simple or
in our academic and everyday lives. Hence, researchers straightforward process. Rather, this learning process likely
investigate processes involved in resolving misconceptions depends on a number of other factors, and in particular, the
(conceptual change) and means to promote conceptual cognitive processes in which readers engage while reading.
change most effectively (Vosniadou, 2003).
One method proposed to enhance conceptual change is Text Comprehension Processes
through the development of specific types of educational Comprehension of texts is a complex activity that involves
texts. In particular, refutational texts are commonly knowledge of the language and the domain, as well as
employed in classroom and laboratory settings because they interactions among lower and higher-order skills used to
encourage students to alter their beliefs about concepts by: process this knowledge. Not surprisingly, then, individuals
(a) explicitly defining common misconceptions of a given vary a great deal in the cognitive processes that they employ
topic, (b) stating the inaccuracies in these beliefs, and (c) during comprehension (McNamara, Jacovina, & Allen, in
following these statements with correct explanations of the press; McNamara & Magliano, 2009). One explanation for
topic (Dole, 2000; Guzzetti et al., 1993). According to the inconsistent results regarding the effects of refutational texts
co-activation hypothesis, refutational texts are effective is that, despite reading the same text, students may engage
because they promote the simultaneous activation of the in vastly different processes depending on the particular
correct information in the text and the incorrect information circumstances (instructions, goals, prior knowledge, etc.).

78
Deep comprehension relies on a reader’s ability to Current Study
activate prior knowledge and make connections among this
In the current study, we examine whether instructing
prior knowledge and information in a text (McNamara &
individuals to self-explain a refutational text will
Magliano, 2009). The result of these processes is the mental
differentially affect their understanding of natural selection
representation. Readers develop coherent representations of
in comparison to thinking-aloud or rereading. We also
text material to the degree that they establish connections
examine the extent to which these differences manifest in
using inferences (Oakhill & Yuill, 1996). Thus, the
the cohesion of students’ verbal responses while reading the
generation of inferences is key to successfully
text. Our research questions are listed below:
comprehending text information (McNamara, 2004;
McNamara & Magliano, 2009). 1) Does self-explanation of a refutational text enhance
Self-explanation is a strategy that has been used to comprehension of natural selection in comparison to
promote these coherence-building processes (McNamara, thinking-aloud or rereading the text?
2004), which in turn enhances understanding of complex
2) Does the cohesion of students’ natural language
concepts (Chi et al., 1989). In particular, self-explanation
responses vary as a function of instructional
fosters the activation of prior knowledge, the generation of
condition (i.e., self-explanation vs. think aloud)?
inferences, and places a greater focus on causally relevant
information, rather than perceptually relevant information 3) Does the cohesion of students’ natural language
(Chi et al., 1989; Legare & Lombrozo, 2014; Walker et al., responses predict post-reading performance on a test
2014). Causal information focuses on mechanistic relations of natural selection knowledge?
between people or objects (e.g., X caused Y to happen), We first hypothesize that students in the experimental
whereas perceptual information refers to the characteristics conditions will vary in the degree to which they are able to
of those people or objects (e.g., their color or shape). learn from the refutational text. In particular, we
Our principal claim is that conceptual change may not hypothesize that students who engage in self-explanation
rely solely on the type of texts presented to students, but will outperform the students who think-aloud or reread on a
also (and more importantly) on the comprehension post-reading measure of natural selection knowledge.
processes that students employ while reading these texts. In Second, we hypothesize that students instructed to self-
particular, we suggest that refutational texts will be explain the text will significantly differ from students
successful to the extent that students generate inferences, instructed to think-aloud in their use of causal cohesion, but
which will consequently increase the coherence of their text not in their use of referential cohesion. This hypothesis
representations. Thus, instructing students to self-explain follows from the assumption that self-explanation primarily
will promote coherence-building processes and increase the enhances the construction of causal connections between
efficacy of refutational texts to promote conceptual change. events, rather than referential connections among concepts.
As an initial step, we investigate coherence-building This hypothesis is in line with previous research that has
processes while students read refutational texts. We do so shown that self-explanation promotes greater processing of
by examining both referential and causal cohesion in causal information (Legare & Lombrozo, 2014; Walker et
students’ responses to the text (i.e., self-explanations and al., 2014) and promotes more coherent mental
think-alouds). Referential cohesion emerges from cues such representations of text (Allen et al., 2015; McNamara, 2004;
as overlap in objects (i.e., nouns) or people, indicating that McNamara & Magliano, 2009).
these referents are the same or different across sentences. Our third hypothesis relates to the link between the
Causal cohesion is signaled by overlapping actions (i.e., cohesion indices and performance on a post-reading test of
verbs) and connectives (e.g., because, therefore), which natural selection. We hypothesize that the cohesion indices
serve to explicitly describe connections among events, that significantly differentiate the self-explanation and
actions, people, and objects. think-aloud conditions will also relate to students’ test
Although these cohesion indices are not direct measures performance. This finding would suggest that the potential
of coherence (e.g., McNamara et al., 2014), studies have benefits of self-explanation are related (at least in part) to
shown that cohesion can serve as a proxy for coherence, and the degree to which self-explanation promotes specific
the cohesion of students’ self-explanations is a strong cognitive processes during comprehension.
predictor of their ability to comprehend texts (Allen, Snow,
& McNamara, 2015; Varner et al., 2013). Thus, we predict Method
that in comparison to normal reading processes reflected in
students’ think-alouds, instructing students to self-explain
text will lead them to place a greater emphasis on causal Participants
relationships, which will increase the degree to which their Participants were 105 introductory psychology students
text responses are causally cohesive. Additionally, we from a university located in the southwestern United States
predict that these cohesion differences will relate to their who participated for course credit. The students were
performance on a posttest knowledge measure. predominantly in their first year of college (66.3%); 67.3%
were male; 53% were Caucasian, 20% were Hispanic, 18%
were Asian, 9% were African American, and 4% reported

79
other; 18.3% of participants reported that they were second aggregated all of the responses provided by each individual
language speakers of English. Seven participants were student (this aggregation method is discussed in greater
excluded from the analyses due to missing data. detail in Varner et al., 2013). Thus, for each student, one
aggregated response file was created, which contained all of
Study Procedure the text that they generated while reading. Paragraph breaks
Students first completed a brief demographics questionnaire were added to each of these aggregated files to preserve the
and then read a refutational text related to natural selection. paragraph structure of the text; thus, each file contained
The text was presented one sentence at a time, with previous eight paragraphs.
text remaining on the screen. Students completed a posttest
that assessed general knowledge of natural selection. Computational Analysis of Text Cohesion Students’
Students were randomly assigned to one of three aggregated response files were analyzed using Coh-Metrix
conditions, which related to the instructions they were given (McNamara et al., 2014), which calculates linguistic text
for reading the text: self-explanation condition (n=33), properties, ranging from lower-level word indices to higher-
think-aloud condition (n=35), or reread condition (n=30). level indices about coherence and rhetorical language use.
Students in the self-explanation and think-aloud conditions For the purposes of the current study, we used Coh-Metrix
were prompted to generate typed responses on 16 separate to measure referential cohesion and causal cohesion. For
occasions throughout the text. The students in the self- each of these groups, three indices were selected.
explanation condition were asked to explain the information The referential cohesion indices included: argument
in the text that they had just read to themselves, whereas overlap, stem overlap, and content word overlap. Argument
students in the think-aloud conditions were told to state overlap refers to the degree to which sentences in the text
whatever they were thinking. Students in the reread contain overlapping nouns and pronouns. Stem overlap is
condition did not generate responses while reading. To similar to argument overlap, but it matches all words with
control for time on task, they read the text twice. similar stems. Thus, overlapping stems will be counted even
if one is a noun and the other is an adjective. Finally,
Measures content word overlap refers to the proportion of explicit
content words that overlap between two sentences.
Refutational Text The text assigned to students (n=716
Therefore, this variable helps to control for the varying
words; 8 paragraphs) was adapted from an excerpt in Steven
lengths of sentences in a text.
Pinker’s book, How the Mind Works and describes the
The causal cohesion indices included: causal ratio,
concept of natural selection and refutes intelligent design. In
explicit verb overlap, and semantic verb overlap. The causal
particular, it explains how the world can appear to be a
ratio is assessed by calculating the ratio of causal verbs to
product of intelligent design, but does not, in reality, have a
causal particles within a text. The causal verb measure is
designer. To make this point, the text uses the example of
based on the frequency of main causal verbs in a text (as
how the eye evolved. The text was adapted to be
identified by WordNet; Felbaum, 1998) and the causal
refutational, in that it explicitly acknowledged commonly
particle count is based on a pre-defined set of causal
held alternative conceptions about a topic (here, natural
particles (e.g., because, as a result). This index reflects the
selection) and directly refuted them by providing more
degree to which students are explicitly explaining causal
satisfactory explanations.
events by expressing the directionality of cause-effect
relationships. Verb overlap is also calculated with WordNet,
Conceptual Inventory of Natural Selection (CINS)
and measures the degree to which verbs (which have strong
Natural selection is a topic for which students commonly
links to actions, events, and states) are repeated in the text.
have misconceptions. Thus, in this study, we examined
Verb cohesion indicates the degree to which a text makes
students’ understanding of this topic. The CINS is a 20-item
explicit connections among events (rather than objects).
multiple-choice assessment developed to measure general
Semantic verb overlap is calculated using Latent Semantic
knowledge of natural selection (Anderson, Fisher, &
Analysis and refers to the degree to which sentences in a
Norman, 2002). The CINS was specifically developed to
text contain verbs that have similar semantic meaning.
capture students’ common misconceptions of natural
selection. The items address understanding of the five
factors and three inferences identified by Mayr (1982) as
Results
important for understanding the logic underlying the theory Statistical analyses were conducted to examine whether
of natural selection. The CINS uses common alternative self-explanation of a refutational text enhanced students’
conceptions as distractors. This assessment is considered to performance on a post-reading measure of natural selection
be a valid and reliable measure of knowledge about natural knowledge, as well as whether the cohesion of their typed
selection (e.g., Nehm & Shonfeld, 2008). responses played a role in this effect.

Text Analyses CINS Performance As predicted, students in the self-


To prepare the students’ natural language responses (i.e., explanation condition (M=51.36%, SD=20.13%)
their self-explanations or think-alouds) for text analysis, we significantly outperformed students in both the think-aloud

80
(M=41.43%, SD=15.93%), and reread (M=41.50%, think-aloud (TA). A DFA model was first developed for the
SD=20.09%) conditions, F(1,95)=43.17, p<.001 (see Figure entire set of students and this model was then used to
1). Hence, self-explanation enhanced students’ knowledge predict group membership of the students using leave-one-
of natural selection in comparison to thinking aloud or out-cross-validation (LOOCV) in order to ensure that the
rereading the text. model was stable across the dataset.
The stepwise DFA retained all three variables related to
Figure 1: Conceptual Inventory of Natural Selection Scores causality. The results revealed that the DFA using these two
as a Function of Condition indices correctly allocated 50 of the 68 students in the total
set, χ2 (df=3, n=68)=25.00 p<.001, for an accuracy of
73.5% (chance level for this analysis is 50%). The reported
Cohen’s Kappa was .473, indicating a moderate agreement.
For the LOOCV analysis, the DFA allocated 47 of the 68
students for an accuracy of 69.1% (see the confusion matrix
reported in Table 2). The results of these analyses confirm
our second hypothesis and indicate that students in the self-
explanation condition generated text responses that were
more causally cohesive, but exhibited no differences in
terms of referential cohesion.

Table 2: Confusion Matrix for DFA classifying Task


Instructions

SE TA
Table 1: Descriptive Statistics [(Means and (SD)] for Whole Set SE 25 8
Referential and Causal Cohesion TA 10 25
Index Self- Think-
Explanation Aloud SE TA
Referential Cohesion LOOCV SE 23 10
Argument overlap 0.31 (0.13) 0.26 (0.18) TA 11 24
Stem overlap 0.03 (0.17) 0.02 (0.02)
Content word overlap 0.06 (0.02) 0.06 (0.03) Performance Differences Related to Causal Cohesion We
Causal Cohesion last examined the degree to which the indices of causal
Causal ratio 1.00 (0.53) 0.69 (0.40) cohesion that significantly differed according to
Explicit verb overlap 0.58 (0.09) 0.48 (0.11) experimental condition also related to students’ performance
Semantic verb overlap 0.13 (0.03) 0.10 (0.04) on the CINS test. Pearson correlations were calculated
between these causal cohesion indices and students’ scores
Cohesion Indices Our second research question regarded on the CINS. This analysis revealed that one of the three
whether students in the self-explanation and think-aloud causal cohesion indices was significantly correlated with
conditions differed in their use of referential and causal scores: explicit verb overlap (r=.36, p< .01).
cohesion within their verbal responses. A MANOVA A final DFA analysis was calculated to investigate
analysis was first conducted to investigate whether whether this measure of verb overlap accurately classified
referential cohesion differed for students in the two the students according to their CINS performance group. A
conditions (see Table 1 for descriptive statistics). This median split was calculated on students’ CINS scores to
analysis yielded a non-significant multivariate model, produce two groups: Low Score (M=6.26, SD=1.25) and
F(3,64)=2.21, p=.10, with none of the indices High Score (M=12.42, SD=2.69). The results of the DFA
demonstrating significant effects. A similar MANOVA revealed that the model correctly allocated 50 of the 68
analysis was conducted to examine whether causal cohesion students in the total set, χ2 (df=1,n=68)=15.832 p< .001, for
differed for the two conditions. This analysis yielded a an accuracy of 73.5% (the chance level for this analysis is
significant model, F(3,64)=10.10, p<.001, with all of the 50%). The reported Cohen’s Kappa was .471, indicating a
indices demonstrating significant effects: causal ratio moderate agreement. For the LOOCV analysis, the DFA
F(1,66)=7.58, p<.01, explicit verb overlap F(1,66)=16.65, also allocated 50 of the 68 students for an accuracy of
p<.01, and semantic verb overlap F(1,66)=11.25, p=.001. 73.5% (see the confusion matrix reported in Table 3). These
A follow-up Discriminant Function Analysis (DFA) was results partially confirm our final hypothesis that the degree
next calculated to investigate whether the three significant of causal cohesion in students’ text responses was related to
causal indices (i.e., causal ratio, explicit verb overlap, their performance on the CINS test. Specifically, students
semantic verb overlap) accurately classified the students with higher verb overlap in their responses also had higher
according to whether they were asked to self-explain (SE) or

81
scores on the CINS test. This suggests that the benefits of be given explicit instructions on how to engage with the text
self-explanation may be attributable (at least in part) to its in order to benefit from this generation process.
promotion of text processes that emphasize developing The results from the current study are important because
connections among actions and events. they indicate that self-explanation can promote beneficial
comprehension processes that can help to increase students’
Table 3: Confusion Matrix for DFA classifying CINS comprehension of complex science concepts. In particular,
Performance this study suggests