Journal of Experimental Child Psychology 161 (2017) 19–31

Contents lists available at ScienceDirect

Journal of Experimental Child

Psychology
journal homepage: www.elsevier.com/locate/jecp

I like it by mere association: Conditioning

preferences in infants
Jenny L. Richmond a,⇑, Jenna Zhao a, Gabrielle Weidemann b
a
School of Psychology, University of New South Wales, Sydney, New South Wales 2052, Australia
b
School of Social Sciences and Psychology, Western Sydney University, Sydney, New South Wales 2000, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Preferences have a profound impact on our behavior; however, rel-
Received 23 September 2016 atively little is known about how preference formation works early
Revised 21 March 2017 in life. Evaluative conditioning occurs when the valence of an ini-
tially neutral object changes when it is paired with a positively
or negatively valenced stimulus. It is possible that evaluative con-
Keywords:
ditioning may account for early preference learning; however, the
Infant
Preference extent to which this kind of learning operates during infancy has
Affect not been empirically tested. The aim of the current studies was
Conditioning to assess whether infants’ preferences for neutral objects is influ-
Eye tracking enced by pairing them with affective stimuli (Experiment 1: happy
Associative learning vs. angry faces, N = 20; Experiment 2: mother vs. stranger faces,
N = 22). Infants’ preferences were tested using both looking time
and behavioral choice measures. The results showed that infants
tended to choose the object that had been paired with the positive
stimulus (Experiment 1: 13/20; Experiment 2: 14/22). Gaze behav-
ior at test did not differentiate between the two objects; however,
gaze behavior during conditioning predicted infants’ behavioral
preference. Only infants who looked longer at the affective stimu-
lus than at the object during learning chose the object that had
been paired with positive valence more often than chance. These
results suggest that infants’ preferences may be influenced by
learned associations between objects and affective stimuli, a pro-
cess akin to evaluative conditioning in adults.
Ó 2017 Elsevier Inc. All rights reserved.

⇑ Corresponding author. Fax: +61 2 9385 3641.

E-mail address: jrichmond@psy.unsw.edu.au (J.L. Richmond).

http://dx.doi.org/10.1016/j.jecp.2017.03.015
0022-0965/Ó 2017 Elsevier Inc. All rights reserved.
20 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31

Introduction

Human behavior is often guided by our preferences. Decisions about what we buy, the activities we
engage in, the food we eat, and the people we socialize with are influenced to a large extent by our
likes and dislikes. Although preferences have a profound impact on our behavior, relatively little is
known about how such preferences are formed, particularly early in life (De Houwer, Thomas, &
Baeyens, 2001).
One mechanism that may account for some instances of preference learning is evaluative condi-
tioning, or the conditioning of evaluations via association with liked versus disliked stimuli (Levey
& Martin, 1975; Martin & Levey, 1978). Evaluative conditioning occurs when the valence of an initially
neutral conditioned stimulus (CS) changes as the result of being paired with a positively or negatively
valenced unconditioned stimulus (US). For example, by pairing George Clooney (a positively valenced
stimulus) with the brand Nespresso (an initially neutral stimulus), advertisers aim to increase our lik-
ing of the Nespresso brand and ultimately increase sales.
Evaluative conditioning is a form of Pavlovian conditioning that involves a change in the response
to a CS as the result of it being paired with a US. A recent meta-analysis (Hofmann, De Houwer,
Perugini, Baeyens, & Crombez, 2010) showed that, just like traditional Pavlovian paradigms, changes
in liking of a CS are greater when adult participants report awareness of the CS–US contingency than
when participants are unaware. Similarly, effect sizes are greater when explicit measures of liking are
used (i.e., self-report or behavioral choice) relative to implicit measures (i.e., implicit association or
affective priming). In contrast to traditional Pavlovian conditioning, however, learned evaluations
may be more resistant to extinction and may depend on co-occurrence of the CS and US rather than
the predictive contingency between the CS and US. There was a relatively small body of research on
evaluative conditioning in children to draw from; however, the meta-analysis also showed that eval-
uative conditioning effects may be less robust in children than in adults (Hofmann et al., 2010). On the
one hand, effect sizes may be smaller because children are less aware of experimenter expectations
and therefore are less likely than adults to show evidence of demand effects. Alternatively, given that
explicit memory for the CS–US contingency influences evaluative conditioning independent of
demand effects, smaller effect sizes may be attributed to age-related changes in explicit memory
(O’Donnell & Brown, 1973).
In children, evaluative conditioning has been studied by pairing initially neutral stimuli (novel car-
toon characters) with food that children like (ice cream) or dislike (brussels sprouts) (Field, 2006). In
this study, children in the experimental group saw one Pokémon-like character paired with ice cream
and another paired with brussels sprouts; each CS–US pair was presented 10 times. Children in the
control group received the same amount of exposure to the CS (characters) and US (foods) as those
in the experimental group; however, the characters and foods were never paired. The results showed
that for participants in the experimental group, liking of the character paired with ice cream increased
following conditioning, whereas liking for the character paired with brussels sprouts decreased. For
children in the control group, there was no change in preferences for the characters as a result of expo-
sure. Interestingly, differences in preference ratings remained significant even after the CS was pre-
sented in the absence of the US during the extinction phase, suggesting that, much like in adults
(Vansteenwegen, Francken, Vervliet, De Clercq, & Eelen, 2006), evaluative conditioning in children
may be relatively resistant to extinction. This insensitivity to extinction suggests that, like in adults,
changes in liking may be the product of referential learning rather than expectancy learning
(Baeyens, Vansteenwegen, Hermans, & Eelen, 2001). The findings were replicated in a second exper-
iment in which an implicit measure (affective priming) was used as the index of stimulus liking. Field
(2006) concluded that evaluative conditioning operates during middle childhood and may be a mech-
anism via which preferences are formed early in life.
Most recently, evaluative conditioning has been demonstrated during early childhood. Using a sim-
ilar paradigm, Halbeisen, Walther, and Schneider (2016) showed that 3- to 6-year-old children’s pref-
erence for novel characters was influenced by pairing them with either liked and disliked animals
(Experiment 1) or liked and disliked foods (Experiment 2). The results showed that across this age
range, changes in preference were age invariant and did not depend on children’s recollection of
 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31 21

how the stimuli were paired. These findings suggest that during early childhood evaluative condition-
ing may be a simple associative learning (i.e., implicit) process that is independent of explicit memory
for the CS–US relationship (Halbeisen et al., 2016).
There are a number of phenomena in infant social cognition research that resemble evaluative con-
ditioning but have not been framed that way. In social referencing paradigms, for example, infants are
more willing to play with a toy that their mother has expressed a joyful or neutral expression toward
than a toy that she has expressed disgust toward (Hornik, Risenhoover, & Gunnar, 1987). Similarly,
infants prefer social partners who have previously spoken to them in infant-directed speech even
when the affective information in the speech is no longer available (Schachner & Hannon, 2011). In
addition, infants prefer characters that are paired with a bouncing action in a target over characters
that are not associated with bouncing irrespective of whether those characters help or hinder others
(Hamlin, Wynn, & Bloom, 2007; Scarf, Imuta, Colombo, & Hayne, 2012).
In each of these cases, changes in preference may be explained with reference to the kind of asso-
ciative learning that occurs in evaluative conditioning. In social referencing, it is possible that pairing a
novel toy (US) with either a joyful expression (CS+) or a disgust expression (CS ) influences infants’
liking of that stimulus and subsequent willingness to engage with it (Hornik et al., 1987). In speech
preference paradigms, it is possible that when a novel person (US) is paired with infant-directed
speech (CS+), infants come to prefer that individual because they were associated with a style of
speech that is ubiquitously positive (Kim & Johnson, 2013). Finally, Scarf et al. (2012) showed that
pairing an initially neutral character (US) with a positive bouncing event (CS+) is sufficient to shift
infants’ preferences toward that character both when that character is a helper in the scenario and
when it is a hinderer.
In sum, it seems that associations between neutral and affective stimuli can influence infants’ pref-
erences for those stimuli; however, the effect of merely pairing a liked or disliked stimulus with ini-
tially neutral stimuli has not been empirically tested during infancy. The goal of the current study was
to test whether pairing neutral stimuli (objects) with affective stimuli (happy and angry faces) would
affect infants’ preferences for those objects. For the current study, 6- and 7-month-old infants were
chosen because previous work has shown that changes in object preference can be measured via both
looking time and behavioral choice measures in this age group (Hamlin et al., 2007; Scarf et al., 2012).
Infants were habituated to object–face pairings and tested using both a visual paired comparison
(VPC) test and a behavioral choice test. We predicted that if evaluative conditioning is operating dur-
ing infancy, infants would look longer at, and be more likely to choose, the object associated with pos-
itive affect.

Experiment 1

Method

Participants
A total of 26 6- and 7-month-old infants (12 female and 14 male) were recruited from a database of
families who had registered their interest in participating in research. Infants were on average 194
days old (range = 175–218) at test. Children were of Caucasian (n = 16), Asian (n = 6), or Middle East-
ern (n = 4) descent. Parents were reimbursed $20 for their transport costs, and infants received a small
gift to thank them for their participation.

Apparatus and stimuli

Eye gaze was recorded using a Tobii XL60 eye tracker. The eye tracker is a 22-inch monitor with a
set of infrared cameras embedded in the edges of the screen. Once the cameras are calibrated, they
track where the infant is fixating on the screen. We used the standard five-point infant calibration
sequence that is built into the Tobii Studio software. Calibration was successful if all five points were
captured with minimal error. Eye gaze data were recorded at 60 Hz, and stimuli were controlled using
E-Prime (Psychology Software Tools).
22 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31

All stimuli used in the experiment were full-color photographs. Each image measured 180  230
pixels and appeared at 11.1  12.4 cm on the screen. Pairs of stimuli were presented 10 cm apart. Face
stimuli were chosen from the NimStim Face Stimulus Set (Tottenham et al., 2009) and consisted of six
individuals (3 male and 3 female) displaying happy expressions and the same six individuals display-
ing angry expressions. The neutral objects were a yellow triangle with googly eyes and a blue square
with googly eyes (Hamlin et al., 2007). During conditioning, one object was paired with happy faces
and the other object was paired with angry faces; the object–emotion pairings were counterbalanced
across infants.

Procedure
Infants were seated on their mother’s lap 60 cm from the Tobii eye tracker. The position of the
monitor was adjusted so that the corneal reflection was captured and a calibration sequence was ini-
tiated. The calibration involved a series of looming red balls that appeared at the four corners and cen-
ter of the screen. Calibration accuracy was checked and repeated if necessary.
A blue looming ball was presented in the center of the screen before each trial, and the experi-
menter initiated the trial when the participant was looking and eye tracking was verified. On each
trial, the object (CS) appeared on one side of the screen for 1 s before the face (US) joined it on the
other side of the screen (see Fig. 1 for sample trial sequence). We chose to present the object prior
to the face to ensure that infants fixated both the object and the face. The experimenter watched
through a live webcam feed and coded when the infant was looking at the screen using the computer
mouse, beginning when the face appeared and continuing until the infant looked away from the
screen. The stimuli disappeared when the infant had looked away, and an attention-getter was again
presented before the next trial began. E-Prime calculated the length of each trial and determined when

Fig. 1. Sample habituation trial sequence for Experiment 1.

 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31 23

the habituation criterion was met. The habituation criterion was met when total looking time over
three consecutive trials was less than half the total looking time during the first three trials.
After the criterion was met, infants were shown two 10-s test trials. In these test trials, infants were
shown the two objects, one that had been paired with happy faces and one that had been paired with
angry faces, on the left and right sides of the screen. The left/right position of the objects was reversed
for the second trial, and the order of these two test trials was counterbalanced across infants, to con-
trol for potential side biases.
After the eye-tracking portion of the experiment was completed, infants were tested in a behav-
ioral choice test. Infants sat on their parent’s lap, and physical versions of the two objects were
attached to a board and placed within reach of the infants. The experimenter who presented the
objects was blind to stimulus associations, and the left/right position of the objects was counterbal-
anced across infants. Infants were encouraged to touch the objects, and their behavioral choice was
video-recorded for later coding. The video coder was blind to the happy/angry object pairings, and
infants’ choice was coded as the first object they touched.

Results and discussion

A total of 26 infants were tested. Data from 3 infants were excluded due to parental interference
(n = 1), failing to habituate (n = 1), or experimental error (n = 1). An additional 3 infants were excluded
from analyses of behavioral choice because they refused to touch the stimuli during the behavioral
choice test. For all eye-tracking data, raw output was subjected to a fixation filter that defined a fix-
ation as a period of time in which the eye position does not move more than 50 pixels for at least
200 ms. Fixations falling within stimulus areas of interest (AOIs) were analyzed.

Habituation
Infants took on average 13.43 trials (SD = 8.23) to reach habituation criterion and accumulated a
total of 74.58 s (SD = 45.66) of looking time during the habituation phase. There was no significant dif-
ference in the total amount of fixation time accumulated during happy trials versus angry trials, t(22)
= 0.242, p = .811, d = .05. A repeated-measures analysis of variance (ANOVA) with trial type (happy vs.
angry) and stimulus (face vs. object) revealed a main effect of stimulus, F(1, 22) = 21.23, p = .000, par-
tial g2 = .49, reflecting that overall infants spent more time fixating the faces than the objects during
the habituation phase (see Table 1). There was no main effect of trial type, F(1, 22) = 0.02, p = .878, par-
tial g2 = .001, or a significant Stimulus  Trial Type interaction, F(1, 22) = 0.07, p = .796, partial
g2 = .003.

Visual paired comparison test

Immediately following habituation, infants were presented with 2  10-s test trials in which the
objects that had been paired with happy and angry faces were presented together. Data from 3 infants
were excluded from this looking time analysis because 1 infant did not have any looking data for Trial
1 and 2 infants did not look at one of the AOIs at all during Trial 2. Data from a total of 20 infants were
analyzed. For each infant, we calculated the proportion of time spent fixating the object that was
paired with happy faces. There was no difference in the magnitude of preferences across Trials 1
and 2, t(19) = 0.762, p = .455, d = .17, so data were averaged. A one-sample t test against chance per-
formance of .50 showed no significant preference, t(19) = 0.067, p = .947, d = .01. Infants spent on aver-

Table 1
Means and standard deviations of average looking times (ms) at face/object on happy and angry trial types during habituation
phase of Experiment 1 (n = 23).

Happy face Happy object Angry face Angry object

Mean 3972 2113 4102 2080
SD 1838 1366 3447 1289
24 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31

age the same amount of time fixating on the objects that were paired with happy faces (M = 3.06 s,
SD = 2.13) and angry faces (M = 3.01 s, SD = 2.36) during the test.
It is possible that the magnitude of visual preferences may change over the course of the 10-s VPC
trial. For this reason, we also looked at the time course of preferential looking across the first trial to
assess whether infants may have exhibited visual preferences within the first 4000 ms of the trial.
Fig. 2 plots the proportion of time spent looking at the object paired with happy faces parsed into
1000-ms time bins. Proportion scores did not differ significantly from chance in any time bin (all
ps > .05).

Behavioral choice test

Three infants were excluded from the analysis of behavioral choice for failing to touch the stimuli
at test. The behavioral choice test was video-recorded and coded. Interrater reliability was conducted
for 40% of the videos and resulted in a kappa of .844. Infants’ choice was coded as the first object the
infants touched. Of the 20 infants, 13 chose the object associated with happy faces and 7 chose the
object associated with angry faces. According to the binomial distribution, the probability of 13 of
20 infants randomly choosing the object associated with happy faces is .13.

Looking during habituation as a function of choice at test

Although there was a larger proportion of infants who chose the happy-associated object, this pref-
erence was by no means universal. Evidence from adults suggests that focusing on the valence is nec-
essary for evaluative conditioning to occur (Gast & Rothermund, 2011). For this reason, we were
interested in potential differences in the pattern of looking behavior during learning that might pre-
dict behavioral preferences at test. To determine whether the amount of time that infants fixated the
valence source (i.e., faces) relative to the objects predicted choice, we conducted a logistic regression
to predict behavioral choice at test (happy-associated object or angry-associated object) from the
mean percentage difference in the amount of time spent looking at faces versus objects (see Table 2).
The model was statistically significant, v2(1) = 9.793, p = .002, with the difference in looking at faces
over objects accounting for 51.6% of the variance (Nagelkerke R2) in behavioral choice. The model cor-
rectly classified 80% of infants’ choices (100% happy choices and 42.9% angry choices) from the mean
difference in looking time fixating faces versus objects during learning.
If greater face looking than object looking is a necessary condition for evaluative conditioning to
occur, we would expect that when this condition is satisfied infants would choose the happy-
associated object more frequently than the angry-associated object; however, when this condition
is not met, infants would perform at chance, choosing the happy- and angry-associated objects equally
often. To test whether infants who look longer at the faces than at the objects perform above chance,

Fig. 2. Mean proportion of time spent looking at the happy-associated object during the VPC test as a function of time bin.
 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31 25

Table 2
Variables in logistic regression equation predicting behavioral choice from mean proportion difference in face/object looking in
Experiment 1.

B SE Wald df p Exp(B)
Mean proportion difference in face/object looking 5.004 2.120 5.57 1 .018 148.935

whereas those who fail to look more at the faces than at the objects perform at chance, we needed to
group infants by the mean percentage difference in the amount of time that they spent looking at faces
versus objects (see Fig. 3). We used a criterion of at least 20% more time looking at faces than at objects
to categorize infants as face preferers (n = 12) versus non-face preferers (n = 8). This criterion has been
used in previous research on infant social perception (Connellan, Baron-Cohen, Wheelwright, Batki, &
Ahluwalia, 2000). As is illustrated in Fig. 3, 11 of the 12 infants who showed a distinct face preference
chose the happy-associated object; the binomial probability of this distribution of object choice occur-
ring by chance is .003. In contrast, among the infants who showed no distinct face preference, 2 chose
the happy-associated object and 6 chose the angry-associated object; this distribution of object choice
has a binomial probability of .15. These data suggest that face > object looking bias may be a necessary
condition for evaluative learning to occur; when infants fail to look longer at faces than at objects dur-
ing learning, they choose at random during the test.
In sum, whereas infants’ looking behavior in the paired comparison test did not differentiate
between the objects paired with happy and angry faces, during the behavioral choice test a greater
number of infants chose the object paired with happy faces than the object paired with angry faces.
Critically, looking behavior during learning predicted behavioral choice at test. Only infants who
showed a distinct face preference during habituation chose the happy object at a greater rate than
would be predicted by chance. In contrast, infants who did not show a distinct face preference were
equally likely to choose the happy- and angry-associated objects.

Experiment 2

The results of Experiment 1 showed that evaluative conditioning is operating during infancy and
that attentional biases toward faces, which are typical of 6- and 7-month-olds (Frank, Amso, &
Johnson, 2014; Frank, Vul, & Johnson, 2009), may be important in determining the extent to which

Fig. 3. Behavioral choice (happy vs. angry) as a function of mean difference in face versus object looking. Of 12 infants who
looked more than 20% longer at the face than at the object (face preferers), 11 chose the happy-associated object at test. In
contrast, for those who did not exhibit a distinct preference for faces during learning, 2 infants chose the happy-associated
object and 6 infants chose the angry-associated object.
26 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31

infants learn the association between objects and affective information. Infants showed a behavioral
preference for the object that was paired with positively valenced faces provided that they fixated the
source of the affective information—that is, the face—during conditioning.
In Experiment 1, we exploited the fact that infants prefer expressions of positive affect (i.e., happy)
over expressions of negative affect (i.e., angry) (LaBarbera, Izard, Vietze, & Parisi, 1976); however,
infants also show looking time preferences for stimuli that are very familiar versus unfamiliar. For
example, infants look longer at a picture of their own mother’s face when it is presented together with
that of a stranger. This preference emerges very early in life and manifests in both looking time pref-
erences (Field, Cohen, Garcia, & Greenberg, 1984) and in the magnitude of event-related potentials
(ERPs) (de Haan & Nelson, 1997). There is little argument that the mother’s face is a special case of
a positively valenced stimulus for infants; however, the impact of pairing the mother with an initially
neutral stimulus on subsequent liking for that stimulus is unknown. The goal of Experiment 2 was to
determine whether evaluative conditioning occurs when initially neutral objects are paired with a pic-
ture of the infant’s mother versus that of a stranger when both faces are displaying happy expressions.
We predicted that after conditioning infants would be more likely to choose the object that was asso-
ciated with their mother than the object associated with a stranger. In addition, based on the results of
Experiment 1, we expected that patterns of gaze during learning would predict the extent to which
affect/object associations were learned.

Method

Participants
A total of 26 6- and 7-month-old infants (11 female and 15 male) were recruited from the same
database of interested families. These infants were on average 201 days old (range = 176–239) at test.
Children were of Caucasian (n = 18), Asian (n = 5), or Middle Eastern (n = 3) descent. Parents were
reimbursed for transport costs, and infants were given a small gift to take home.

Apparatus and stimuli

The equipment and software used in this experiment were identical to those used in Experiment 1.
The neutral conditioned stimuli also remained the same. However, the happy and angry face images
were replaced with a picture of the infant’s mother and a picture of a stranger. At the beginning of the
session, a photograph of the mother, making her most natural happy face, was taken and inserted into
the stimulus presentation script. The same three female happy faces from Experiment 1 (Tottenham
et al., 2009) were used as ‘‘strangers” in this experiment. For each participant, the stranger who looked
most dissimilar to the mother was chosen to ensure that the faces would not be too difficult to dis-
criminate. The majority of babies (75%) were tested with a stranger of the same race as them. All
images were 180  230 pixels and measured 11.1  12.4 cm on the screen.

Procedure
Infants were seated in front of the eye tracker as in Experiment 1. The five-point calibration
sequence was run, and infants were presented with habituation trials pairing the pictures of their
mother and the stranger with either the blue square or yellow triangle stimulus using the same trial
procedure as Experiment 1. After the habituation criterion was met, the infants saw two 10-s test tri-
als in which the objects were presented together. After the eye tracking was completed, an experi-
menter showed the infants the physical objects on a board and their behavioral choice was recorded.

Results and discussion

A total of 26 infants participated in Experiment 2. Data from 4 infants were excluded from behav-
ioral choice analysis because they failed to touch the stimuli at test. As in Experiment 1, eye-tracking
data during both conditioning and test were subjected to a fixation filter and analyzed according to
predefined AOIs that represented each stimulus. Behavioral choice data were coded offline and taken
as the stimulus that the infant first touched.
 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31 27

Habituation
Infants in Experiment 2 spent on average 80.89 s (SD = 55.4) fixating the stimuli during the condi-
tioning phase and took on average 11.93 trials (SD = 6.78) to reach the criterion. There was no differ-
ence in the total amount of looking time devoted to trials depicting the infant’s mother versus those
depicting a stranger during habituation, t(25) = 0.562, p = .579, d = .11. Much like Experiment 1, overall
infants spent more time fixating the faces than the objects during habituation trials, F(1, 25) = 15.11,
p = .001, partial g2 = .38; there was no main effect of trial type (mother vs. stranger), F(1, 25) = 0.065,
p = .801, partial g2 = .003, and there was no interaction between stimulus type and trial type, F(1, 25)
= 0.092, p = .764, partial g2 = .004 (see Table 3).

Visual paired comparison test

Data from 3 infants were excluded from looking time analysis because 1 infant did not look at one
of the AOIs during Trial 1 and 2 infants did not exhibit sufficient looking time (i.e., >2500 ms) on both
Trials 1 and 2. Data from a total of 23 infants were analyzed.
The proportion of time that infants spent looking at the object that had been paired with their
mother was calculated out of total test looking for each test trial. There was no significant difference
in the magnitude of preference across trials, t(22) = 0.798, p = .433, d = .17, so preference scores were
averaged for analysis. One-sample t tests showed that the preference for the mother-associated object
(M = .48, SD = .13) was not significantly above chance, t(22) = 0.769, p = .450, d = .16.
As in Experiment 1, we also looked at the time course of preferential looking across the first trial to
assess whether infants may have exhibited visual preferences within the first 4000 ms of the trial.
Fig. 4 plots the proportion of time spent looking at the object paired with the mother parsed into
1000-ms time bins. Proportion scores did not differ significantly from chance except between 2000
and 3000 ms, t(24) = 2.37, p = .027, d = .53.

Behavioral choice test

During the behavioral test, infants were given the opportunity to choose between physical versions
of the objects that had been paired with their mother and the stranger. Interrater reliability was con-
ducted for 40% of the videos and resulted in a kappa of .848. Of the 22 infants in this experiment, 14
chose the object that had been associated with their mother and 8 chose the object that had been asso-
ciated with the stranger. According to binomial distribution, the probability of 14 of 22 infants choos-
ing the object associated with their mother was .143.

Looking during habituation as a function of choice at test

Again we calculated the mean percentage difference in looking at faces versus objects during learn-
ing and conducted a logistic regression to predict behavioral choice (see Table 4). The model was sta-
tistically significant, v2(1) = 5.008, p = .025, with the difference in looking at faces over objects
accounting for 27.1% of the variance (Nagelkerke R2) in behavioral choice. The model correctly classi-
fied 68.2% of infants’ choices (92.9% mother choices and 25% stranger choices) from the mean differ-
ence in looking time fixating faces versus objects during learning.
As in Experiment 1, we grouped infants by the mean percentage difference in the amount of time
spent looking at faces versus objects. There were 15 infants who exhibited distinct face preferences,
spending at least 20% more time looking at faces than at objects, and 7 infants who did not exhibit
a distinct face preference. As is shown in Fig. 5, 11 of the 15 infants who showed a distinct face pref-

Table 3
Means and standard deviations of average looking times (ms) at face/object on mother and stranger trial types during habituation
phase of Experiment 2 (n = 26).

Mother face Mother object Stranger face Stranger object

Mean 4998 2262 4797 2276
SD 3449 1451 4618 1653
28 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31

Fig. 4. Mean proportion of time spent looking at the mother-associated object during VPC test as a function of time bin.

Table 4
Variables in logistic regression equation predicting behavioral choice from mean proportion difference in face/object looking in
Experiment 2.

B SE Wald df p Exp(B)
Mean proportion difference in face/object looking 2.443 1.215 4.05 1 .044 0.087

erence chose the object associated with their mother; the binomial probability of this distribution of
object choice occurring by chance is .06. Of the 7 infants who showed no distinct face preference, 3
chose the mother-associated object and 4 chose the stranger-associated object; this distribution of
object choice has a binomial probability of .50.
In sum, looking behavior in the paired comparison test did not differentiate between the object
paired with the mother and the object paired with a stranger except between 2000 and 3000 ms. Dur-
ing the behavioral choice test, however, a larger number of infants chose the object that was associ-

Fig. 5. Behavioral choice (mother vs. stranger) as a function of mean difference in face versus object looking. Of 15 infants who
looked more than 20% longer at the face than at the object (face preferers), 11 chose the mother-associated object at test. In
contrast, for those who did not exhibit a distinct preference for the faces during learning, 3 infants chose the mother-associated
object and 4 infants chose the stranger-associated object.
 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31 29

ated with their mother. Furthermore, the amount of time that infants looked at the face versus the
object during habituation predicted which object the infants chose. Infants who showed a distinct face
preference during habituation chose the mother-associated object at a greater rate than would be pre-
dicted by chance. In contrast, infants who showed no distinct face preference during habituation chose
the mother- and stranger-associated objects equally. Consistent with the results of Experiment 1,
these data suggest that the amount of time infants spend looking at the face relative to the object
may constrain the extent to which infants are able to learn evaluative associations.

General discussion

Here we have shown that pairing neutral objects with valenced stimuli can influence infants’ pref-
erences for those objects. In the behavioral choice test, a larger proportion of infants chose the object
that was associated with positive valence (happy face or mother’s face) than the object that was paired
with negative/less positive valence (angry face or stranger’s face). These results suggest that the mere
pairing of an initially neutral object with a liked or disliked stimulus can result in the transfer of
valence to this neutral object and a shift in preference for it. Of interest, the pattern of looking dis-
played during habituation significantly predicted infants’ behavioral choice, although effects were
stronger in Experiment 1, where we used a more extreme valence contrast (happy vs. angry), than
in Experiment 2, where the contrast was less extreme (smiling mother vs. smiling stranger). Infants
who showed a distinct preference for faces during learning were more likely to choose the object asso-
ciated with the more positively valenced face than would be predicted by chance, whereas those who
did not show a distinct face preference were not. These results suggest that a clear preference for look-
ing at the source of the valence information during learning may be a necessary condition for evalu-
ative conditioning to occur.
We predicted that evaluative conditioning would manifest as preferences for the object paired with
positive valence in both the looking time and behavioral choice tests; however, we saw evidence of
preferential looking in only a single 1000-ms time bin. In Experiment 2, infants exhibited preferential
looking between 2000 and 3000 ms; however, contrary to predictions, preferences were for the object
that had been associated with the stranger. The lack of strong visual preferences during the VPC test
may simply reflect the fact that each of the objects was equally familiar at the end of habituation.
Infants accumulated the same amount of looking time to each of the objects during conditioning. It
is possible that when the familiarity/novelty of a stimulus is held constant, as in the current experi-
ments, the valence of the stimulus does not affect looking time. Other studies using a combination
of behavioral choice and looking time measures have also reported dissociations between these
indexes (Hamlin et al., 2007), suggesting that the behavioral choice test may be a more sensitive mea-
sure of preference than looking time.
In both experiments, the pattern of looking during learning predicted behavioral choice at test. It is
unclear, however, whether looking biases during learning play a causal role in preference formation or
are simply a consequence of the object preferences that infants bring to the experiment. It could be
argued that looking behavior during conditioning determines whether infants are able to extract
the affective information from the face and associate it with the object. For the valence to be trans-
ferred from the face to the object, infants must perceive and encode the affective information during
the conditioning experience (Jones, Fazio, & Olson, 2009). In the current experiments, each trial began
with a 1000-ms presentation of the object on its own before the face joined it. Consequently, it is likely
that the objects were well encoded; however, infants who spent relatively little time fixating the faces
may have failed to recognize and/or encode the affective information in the faces, which may have
prevented evaluative conditioning from occurring. In contrast, infants who exhibited distinct prefer-
ences for the face during learning may have been more likely to encode the affective information
and associate it with an object. Future research will assess whether attention during learning plays
a causal role in preference formation by manipulating infants’ looking behavior during conditioning
using a gaze-contingent paradigm.
It is also possible that looking biases reflect individual differences in infants’ preferences for, or
affective reactions to, faces; these preferences may place constraints on infants’ learning. For example,
30 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31

research has shown that pairing sweet tastes with neutral odors increases liking for those odors, but
only in people who like sweetness (Yeomans, Mobini, Elliman, Walker, & Stevenson, 2006). In much
the same way, it is possible that the relative differences in looking time at the faces and objects during
learning reflect infants’ natural inclination toward social versus nonsocial stimuli. There is evidence
for individual differences in infants’ visual preferences for faces compared with other equally visually
complex stimuli, and research has shown that female infants are more likely than male infants to
show preferences for faces (Connellan et al., 2000). Infants’ liking for faces may place constraints on
their learning in much the same way as adults’ liking for sweet tastes does; only those for whom faces
elicit a distinct affective response are able to associate that affect with the objects they are paired with.
Future research will use pretesting to determine whether individual differences in face preferences are
stable and, if so, whether they predict evaluative conditioning when affective information is commu-
nicated in the face.
Alternatively, rather than being causal, it is possible that looking behavior differences during learn-
ing simply reflect object preferences that infants bring to the experiment. There is some evidence from
adults to suggest that evaluative conditioning is less likely to occur when there are strong existing
preferences for the conditioned stimuli (Gibson, 2008). Similarly, infants who had a strong preexisting
preference for a particular object may have spent more time looking at the objects than at the faces
during the habituation phase. Unfortunately, we do not have the pretest data that would allow us
to determine whether object choice is independent of conditioning in infants who begin the experi-
ment with a strong preference for one object but is influenced by conditioning in infants who do
not. The fact that we had null preferences in the VPC test seems to speak against this possibility; how-
ever, future work will determine whether infants are less susceptible to evaluative conditioning when
there are strong preexisting preferences. In addition, without a control group, it is not possible to
determine whether preferences in the current study reflect increases in liking for the object that
was paired with positive valence or decreases in liking for the object that was paired with negative
valence or a combination of the two. Follow-up experiments pairing objects with happy versus neutral
faces and angry versus neutral faces is needed to address this issue.
Here we have shown that infants readily associate affect with objects. This process, which is akin to
evaluative conditioning in adults, may account for how infants learn preferences during early life.
Although this study was not designed to determine what kind of learning processes underlie early
evaluative conditioning, this will be a theoretically fruitful area for future research. Although evalua-
tive conditioning is a Pavlovian paradigm, much research in adults has shown that, unlike other forms
of conditioning, changes in liking depend more on conscious awareness of the CS–US relation than on
the extent to which the CS predicts the US (Hofmann et al., 2010). As such, propositional theories have
proposed that evaluative conditioning may rely on explicit learning processes rather than implicit
ones (Mitchell, De Houwer, & Lovibond, 2009). To date, these theories have been developmentally
silent, and research has not yet addressed the possibility that the mechanisms underlying evaluative
conditioning might change across the lifespan. Given the age-related changes that are seen in explicit
memory processes across early and middle childhood, it is possible that very early in life evaluative
conditioning is primarily subserved by a simple associative learning mechanism that is replaced by
a more complex propositional process as explicit memory processes improve. Alternatively, it may
be that, like other kinds of conditioning seen during infancy, the nature of the processes underlying
learning does not change; rather, learning improves in a continuous and quantitative way (Rovee-
Collier, 2007). Future research will address the extent to which variables that influence explicit but
not implicit memory (i.e. age, retention interval, memory load) affect evaluative conditioning perfor-
mance in infants and children. The advantage of testing theories of evaluative conditioning mecha-
nism by studying infants and children is that we might inform the debate under conditions that are
less contaminated by participants’ desire to conform to experimenter expectations.
Here we have shown that infants’ preferences can be influenced by pairing objects with positively
versus negatively valenced stimuli. We showed that individual differences in the time that infants
spent looking at the faces relative to the objects during learning predicted their behavioral choice
at test. From the current results, it is not clear whether these looking biases reflect stable individual
differences in preferences for social stimuli, which may place constraints on what it is possible for
infants to learn, or simply reflect attentional processes and individual differences in what infants learn
 J.L. Richmond et al. / Journal of Experimental Child Psychology 161 (2017) 19–31 31

in the moment. Future research will detangle these possibilities and uncover the mechanisms under-
lying evaluative conditioning early in life.

References

Baeyens, F., Vansteenwegen, D., Hermans, D., & Eelen, P. (2001). Chilled white wine, when all of a sudden the doorbell rings:
Mere reference and evaluation versus expectancy and preparation in human Pavlovian learning. In F. Columbus (Ed.).
Advances in psychology research (Vol. 4, pp. 241–277). Huntington, NY: Nova Science.
Connellan, J., Baron-Cohen, S., Wheelwright, S., Batki, A., & Ahluwalia, J. (2000). Sex differences in human neonatal social
perception. Infant Behavior and Development, 23, 113–118.
de Haan, M., & Nelson, C. A. (1997). Recognition of the mother’s face by six-month-old Infants: A neurobehavioral study. Child
Development, 68, 187–210.
De Houwer, J., Thomas, S., & Baeyens, F. (2001). Associative learning of likes and dislikes: A review of 25 years of research on
human evaluative conditioning. Psychological Bulletin, 127, 853–869.
Field, A. P. (2006). I don’t like it because it eats sprouts: Conditioning preferences in children. Behaviour Research and Therapy, 44,
439–455.
Field, T. M., Cohen, D., Garcia, R., & Greenberg, R. (1984). Mother–stranger face discrimination by the newborn. Infant Behavior
and Development, 7, 19–25.
Frank, M. C., Amso, D., & Johnson, S. P. (2014). Visual search and attention to faces during early infancy. Journal of Experimental
Child Psychology, 118, 13–26.
Frank, M. C., Vul, E., & Johnson, S. P. (2009). Development of infants’ attention to faces during the first year. Cognition, 110,
160–170.
Gast, A., & Rothermund, K. (2011). What you see is what will change: Evaluative conditioning effects depend on a focus on
valence. Cognition & Emotion, 25, 89–110.
Gibson, B. (2008). Can evaluative conditioning change attitudes toward mature brands? New evidence from the Implicit
Association Test. Journal of Consumer Research, 35, 178–188.
Halbeisen, G., Walther, E., & Schneider, M. (2016). Evaluative conditioning and the development of attitudes in early childhood.
Child Development. http://dx.doi.org/10.1111/cdev.12657. Advance online publication.
Hamlin, J., Wynn, K., & Bloom, P. (2007). Social evaluation by preverbal infants. Nature, 450, 557–559.
Hofmann, W., De Houwer, J., Perugini, M., Baeyens, F., & Crombez, G. (2010). Evaluative conditioning in humans: A meta-
analysis. Psychological Bulletin, 136, 390–421.
Hornik, R., Risenhoover, N., & Gunnar, M. (1987). The effects of maternal positive, neutral, and negative affective
communications on infant responses to new toys. Child Development, 58, 937–944.
Jones, C. R., Fazio, R. H., & Olson, M. A. (2009). Implicit misattribution as a mechanism underlying evaluative conditioning.
Journal of Personality and Social Psychology, 96, 933–948.
Kim, H. I., & Johnson, S. P. (2013). Do young infants prefer an infant-directed face or a happy face? International Journal of
Behavioral Development, 37, 125–130.
LaBarbera, J., Izard, C., Vietze, P., & Parisi, S. (1976). Four- and six-month-old infants’ visual responses to joy, anger, and neutral
expressions. Child Development, 47, 535–538.
Levey, A., & Martin, I. (1975). Classical conditioning of human ‘‘evaluative” responses. Behavioral Research & Therapy, 13,
221–226.
Martin, I., & Levey, A. (1978). Evaluative conditioning. Advances in Behavioral Research Therapy, 1, 57–102.
Mitchell, C. J., De Houwer, J., & Lovibond, P. F. (2009). The propositional nature of human associative learning. Behavioral and
Brain Sciences, 32, 183–198.
O’Donnell, J. M., & Brown, M. J. (1973). The classical conditioning of attitudes: A comparative study of ages 8–18. Journal of
Personality and Social Psychology, 26, 379–385.
Rovee-Collier, C. (2007). Dissociations in infant memory: Rethinking the development of implicit and explicit memory.
Psychological Review, 104, 467–498.
Scarf, D., Imuta, K., Colombo, M., & Hayne, H. (2012). Social evaluation or simple association? Simple associations may explain
moral reasoning in infants. PLoS ONE, 7(8), e42698.
Schachner, A., & Hannon, E. (2011). Infant-directed speech drives social preferences in 5-month-old infants. Developmental
Psychology, 47, 19–25.
Tottenham, N., Tanaka, J. W., Leon, A. C., McCarry, T., Nurse, M., Hare, T. A., ... Nelson, C. (2009). The NimStim set of facial
expressions: Judgments from untrained research participants. Psychiatry Research, 168, 242–249.
Vansteenwegen, D., Francken, G., Vervliet, B., De Clercq, A., & Eelen, P. (2006). Resistance to extinction in evaluative
conditioning. Journal of Experimental Psychology: Animal Behavior Processes, 32, 71–79.
Yeomans, M. R., Mobini, S., Elliman, T. D., Walker, H. C., & Stevenson, R. J. (2006). Hedonic and sensory characteristics of odors
conditioned by pairing with tastants in humans. Journal of Experimental Psychology: Animal Behavior Processes, 32, 215–228.