Вы находитесь на странице: 1из 9

Organizational Behavior and Human Decision Processes 116 (2011) 163–171

Contents lists available at ScienceDirect

Organizational Behavior and Human Decision Processes


journal homepage: www.elsevier.com/locate/obhdp

How do you fake a personality test? An investigation of cognitive models


of impression-managed responding
Mindy K. Shoss a,⇑, Michael J Strube b
a
Department of Psychology, University of Houston, TX, USA
b
Department of Psychology, Washington University in St. Louis, MO, USA

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

Article history: Because faking poses a threat to the validity of personality measures, research has focused on ways of
Received 3 February 2010 detecting faking, including the use of response times. However, the applicability and validity of these
Accepted 14 May 2011 approaches are dependent upon the actual cognitive process underlying faking. This study tested three
Available online 28 June 2011
competing cognitive models in order to identify the process underlying faking and to determine whether
Accepted by Paul Levy
response time patterns are a viable method of detecting faking. Specifically, we used a within-subjects
manipulation of instructions (respond honestly, make a good impression, make a specific impression)
Keywords:
to examine whether the distribution of response times across response scale options (e.g., disagree,
Cognition
Faking
agree) could be used to identify faking on the NEO PI-R. Our results suggest that individuals reference
Personality questionnaire a schema of an ideal respondent when faking. As a result, response time patterns such as the well-known
inverted-U cannot be used to identify faking.
Ó 2011 Elsevier Inc. All rights reserved.

Introduction Particularly troublesome, 74% of applicants believe that other


applicants fake on selection assessments (English et al., 2005 as ci-
Personality measures have re-emerged in personnel selection ted in Griffith & McDaniel, 2006), suggesting considerable pressure
because research has supported their validity in such contexts for applicants to fake in order to not be at a significant disadvan-
(Ones, Dilchert, Viswesvaran, & Judge, 2007). The ease and low cost tage. Not surprisingly, the self-help literature now has many
of administering these assessments, as well as the lower adverse ‘‘how to’’ guides for manipulating responses to personality assess-
impact compared to other selection tools (Hough, Oswald, & ments to increase the chances of being hired (Griffith & McDaniel,
Ployhart, 2001), adds to their attraction. However, researchers 2006).
have questioned the validity of these measures in light of research In response to these issues, researchers have sought to identify
suggesting that individuals can (Griffith & McDaniel, 2006; ways to detect fakers, and have suggested that response times
Viswesvaran & Ones, 1999) and do (Birkeland, Manson, Kisamore, might be useful towards this end (e.g., Holden & Kroner, 1992;
Brannick, & Smith, 2006) effectively distort their responses in Holtgraves, 2004). However, research findings on the differences
situations where it is advantageous to do so. Recent research sug- in response times between those faking and those responding hon-
gests that distortions can come at a considerable cost to predictive estly have largely been inconsistent. Not only does this call into
validity (Komar, Brown, Komar, & Robie, 2008), especially for those question the idea that response times may be useful for identifying
who score at the high end of the distribution (Mueller-Hanson, fakers, but also it raises a broader issue of identifying the cognitive
Heggestad, & Thornton, 2003). process that underlies faking on personality questionnaires that
Thus, faking undermines the use of personality scales as re- might lead to differences in response times.
search instruments and in applied settings (Mesmer-Magnus & In the current study, we examine the fundamental assumption
Viswesvaran, 2006). Faking subverts the purpose of content-driven underlying the suggestion to use response times to detect faking—
personnel selection and instead may lead to the selection of appli- that the cognitive process engaged while faking is different enough
cants who are especially skilled at impression management rather from honest responding that response times will be consistently
than especially noteworthy on the intended attribute. At an even different when an individual is faking than when an individual is
more basic level, faking on selection instruments raises questions responding honestly. Specifically, we test three competing models
about the basic honesty of applicants and potential employees. of faking—the self-schema, adopted-schema, and semantic evalua-
tion models—by examining patterns of response times for different
⇑ Corresponding author. Address: Department of Psychology, University of response options. Our approach to testing these models is a novel
Houston, Houston, TX 77024, USA. one. In contrast to prior studies that have examined average re-
E-mail address: mmkrischer@uh.edu (M.K. Shoss). sponse times across items, our approach allows us to explicitly test

0749-5978/$ - see front matter Ó 2011 Elsevier Inc. All rights reserved.
doi:10.1016/j.obhdp.2011.05.003
164 M.K. Shoss, M.J Strube / Organizational Behavior and Human Decision Processes 116 (2011) 163–171

the nuanced predictions these models make for how response time being assessed relates to either a favorable or stereotypic image.
patterns associated with faking compare to the well-known in- Responses are selected solely on the basis of social desirability,
verted-U response time pattern for honest responding (Kuiper, rather than by the extent to which they are self-referent.
1981). As a result, we are not only able to test competing models Research examining these models has primarily examined aver-
of faking, but are also able to examine whether response time pat- age response times across items, rather than examining average re-
terns can be used to identify faking. sponse times for particular response options, and has resulted in
Our study, therefore, advances the literature on faking by shed- inconsistent and conflicting findings. Supporters of the self-schema
ding light on the cognitive processes underlying faking and, in model cite research indicating that the social desirability implica-
doing so, indicating whether response time patterns are a viable tions of questionnaire items have been found to correlate with re-
way of detecting faking. As these models of faking take as their sponse times, reflecting the delayed judgment stage implied by
starting point the cognitive processes presumed to underlie honest this model (McDaniel & Timm, 1990). For example, Holtgraves
responding, we begin by discussing cognitive models of honest (2004) varied the social desirability signals of instructions for a
responding. We then describe the models of faking and the predic- personality assessment and found that response times were posi-
tions that each makes for response time patterns across response tively associated with social desirability concerns. In contrast, Hol-
options. den and colleagues (Holden & Hibbs, 1995; Holden & Kroner, 1992)
found that impression-managed response times for responses con-
Cognitive models of honest responding sistent with an ideal applicant were identical to honest-responding
response times, which they note is consistent with the use of an
Despite the fact that reading a 20-item scale takes about adopted-schema. Finally, Hsu, Santelli, and Hsu (1989) found that
1.5 min, the average administration time for the same length scale impression-management was associated with faster responses
is about 10–15 min (Griffith & McDaniel, 2006). Therefore, other than honest responding, consistent with the semantic-exercise
cognitive processes are likely engaged when an individual re- model, which suggests a judgment stage without a retrieval stage.
sponds to a personality assessment. For example, Holtgraves Contributing to the confusion, Vasilopoulos et al. (2000) sug-
(2004) argued that these cognitive processes include a retrieval gested that the variation in findings may be a result of differences
stage, during which one gathers the information necessary to for- in the manipulation of the impression management task. They pre-
mulate a response, and a judgment stage, during which one maps dicted that when the ideal job characteristics are not familiar or an
the retrieved information onto the response options and selects the individual is merely trying to make a global positive impression,
appropriate response. response times would be slower than honest-responding response
More specifically, others have argued that honest responding times because no clear impression is being constructed. Consistent
requires accessing a self-schema in the retrieval phase and com- with these predictions, they found that impression-management
paring item responses to the exemplars that define the individual was associated with slower response times than honest responding
when making judgments about the response option to select (Kui- for a low job familiarity (fake-good) condition, whereas a high job
per, 1981). The endpoints of a response scale (e.g., strongly agree, familiarity (fake-specific) condition was associated with slightly
strongly disagree) are more likely to correspond to simple, clear, quicker responses. These results indicate that schema strength,
and easily retrieved exemplars, leading to fast responses because characterized as familiarity with the job, may influence the average
comparison to self-relevant information is accomplished with little speed with which individuals respond to personality inventories.
effort (see also Popham & Holden, 1990). In contrast, middle re- These findings, however, do not lend clear support for a given mod-
sponse options (e.g., agree somewhat, slightly disagree) corre- el and make the interpretation of prior research on the response
spond to more complex, conflicting, or context-dependent times associated with faking difficult to interpret.
exemplars, requiring more thought (e.g., extensive retrieval) and We suggest that past research contains a number of problems
resulting in longer response times. The resulting ‘‘inverted-U ef- that limit the ability to test model validity. One major problem is
fect’’ has now been confirmed by numerous studies (e.g., Casey & the use of between-subjects designs. These designs require the
Tyron, 2001; Mueller, Thompson, & Dugan, 1986). Recently, use of a standardization technique to control for individual differ-
Akrami, Hedlund, and Ekehammar (2007) reported its presence ences in reading speed and simple response time that might ob-
for responses to the NEO Personality Inventory, the measure used scure comparisons across conditions. For example, Holden (1998;
in the current study. Holden, Fekken, & Cotton, 1991) used a double-standardization
When individuals fabricate responses to questionnaire items, technique to analyze the response latency data for a number of
additional processes are likely to be engaged as attempts are made studies that assigned participants to either a fake or honest
to manage or project an impression that deviates from the true or responding condition. In this approach, response times are first
honest self. Several models of impression-managed responding standardized across items within each subject, in order to control
have been proposed, each with implications for how response for reading speed, and then response times are standardized across
times might deviate from the inverted-U pattern. subjects within each item, to control for item-specific factors such
as item length. Casey and Tryon (2001) attempted to improve on
Cognitive models of impression-managed responding Holden’s technique by having participants push a key after they
had finished reading the question, then a key for an answer. The re-
Vasilopoulos, Reilly, and Leaman (2000) discussed three dis- sponse time for the individual was the time difference between
tinct, mutually exclusive models of the cognitive processes that both key pushes, assumed to just measure psychological response
might guide impression-managed responding. In the self-schema time. However, individuals may begin the retrieval process as they
model, individuals access their self-schemas but decide to make re- are reading or may not remember or be able to separate when
sponses inconsistent with those schemas, which requires some reading the question ends and thinking about a response begins.
deliberation while the social desirability of responses is consid- In other words, the standardization process may remove variability
ered. The adopted-schema model is similar to the self-schema mod- due to response distortion—the variability of primary interest in
el except that impression-managed responses involve referencing response fabrication research. Indeed, empirical results indicate
a schema of an ideal respondent, not one’s own self-schema. The higher response-distortion estimates for within-subjects designs
semantic-exercise model suggests that impression-managed re- (that require no standardization) than for between-subjects de-
sponses require a simple semantic evaluation of how the trait signs (Mesmer-Magnus & Viswesvaran, 2006).
M.K. Shoss, M.J Strube / Organizational Behavior and Human Decision Processes 116 (2011) 163–171 165

Table 1
Predicted response time patterns for models of impression-managed responding.

Model Schema Response process Predicted response time distribution for Predicted difference between fake-good and fake-specific
fake good condition conditions
Self-schema Self Retrieval and prolonged Flatter and slower compared to honest No difference
judgment stages responding
Adopted- Ideal Retrieval and judgment Indistinguishable from honest Fake-specific will show a faster and more pronounced
schema Respondent stages responding inverted-U distribution than fake-good
Semantic- None Judgment stage only Flatter and faster than honest No difference
exercise responding

Note. The honest responding condition is expected to produce a significant inverted-U pattern, consistent with past research.

A second, related problem is that these studies generally com- Semantic-exercise model
pare average response times between items (i.e., averaging to- The semantic-exercise model involves a social desirability
gether response times regardless of the response option assessment of each item. In Holtgraves’ (2004) conceptualization,
selected), eliminating the possibility of using more subtle patterns this model corresponds to the elimination of the retrieval stage.
such as the inverted-U effect to detect honest versus faked re- Responses are selected solely on the basis of social desirability,
sponses. Moreover, using response time distributions allows com- rather than by the extent to which they are self-referent or reflect
peting tests of the previously described cognitive models, as each an ideal schema. Response times may show an inverted-U distribu-
has distinct implications for the way impression-managed tion to the extent that extreme responses more readily convey a
responding should emerge with regard to the inverted-U pattern. good impression. However, given that a semantic evaluation en-
If response time distributions across response options are different gages relatively few cognitive processes, moderate responses
when individuals fabricate their responses compared to respond- (e.g., somewhat agree) should also be made relatively quickly.
ing honestly, then response time distributions could be a viable Thus, although an inverted-U may emerge, this will be less pro-
method to detect faking in applied settings. We describe implica- nounced than in the honest-responding condition, and response
tions of the three models for response latency distributions in times will be generally faster than is true for honest responding.
the following discussion. This information is also summarized in Because selection merely requires a semantic evaluation, this
Table 1. model suggests there will no differences between the response
time distributions when individuals are trying to make a general
Self-schema model good impression versus a specific impression.
Recall that the self-schema model of impression-managed
responding suggests that individuals access their self-schemas The current study
but decide to make responses inconsistent with these schemas in
an attempt to appear more socially desirable. This deliberation The present study employed a within-subjects design in which
about social desirability prolongs the judgment stage in Holt- participants responded to the NEO PI-R under each of three condi-
graves’ (2004) conceptualization. Moreover, the self-schema model tions: (a) honestly, (b) to make a good impression, and (c) to make
suggests that extreme dissimulated responses should engage rela- a specific impression. Both general and specific impression-man-
tively more cognitive processes than less extreme responses. Re- aged conditions were included because the three models make
sponses on the end-points of the scales (e.g., strongly agree), different predictions for the response time distributions under
which ordinarily are retrieved and selected easily when the re- fake-general and fake-specific instructions. Thus, including a
sponse is honest, require additional thought, especially if an indi- fake-specific condition is necessary to provide a complete test of
vidual’s self-schema indicates that a moderate response would the models.
be honest. On average, this process produces a flatter response dis- Within the specific impression condition, participants were gi-
tribution than occurs for honest responding and one that is slower ven one of two sets of instructions—to respond as if applying for
overall due to the added deliberation about social desirability of re- a job in accounting or to respond as if applying for a job in public
sponses relative to the self. Additionally, this model suggests no relations. Given that popular stereotypes of these positions involve
differences between the distributions when individuals are trying greater differences in levels of extraversion than in levels of other
to make a general good impression versus a very specific impres- dimensions (e.g., openness), we expected this condition would pri-
sion—both involve additional deliberation that takes additional marily involve a manipulation of the extraversion dimension. Prior
time, especially for extreme responses. research has found increased extraversion scores as a result of fak-
ing (Bagby & Marshall, 2003), but this may depend on the impres-
Adopted-schema model sion one is trying to make. Although faking to make a generally
According to the adopted-schema model, individuals access a favorable impression and specific impressions appropriate for
schema of an ideal respondent. Provided that individuals have a many jobs likely involve portraying oneself as extraverted, some
schema of what a good applicant might look like, the adopted sche- jobs may call for a more introverted nature. This study, in addition
ma model predicts a pattern of response times that should be also to testing the aforementioned models, determined if respondents
be inverted-U shaped because the responses are ‘‘honest’’ with re- could dissimulate in this way when the circumstances called for it.
spect to the referenced schema—the referenced schema is adopted As noted previously, the cognitive models have different impli-
as if it were one’s own. The adopted-schema model can be taken to cations for response time distributions, both their shape and their
make a more extreme prediction, however, in line with the results overall level. Prior research does not make a strong case for a par-
reported by Vasilopoulos et al. (2000) suggesting that schema ticular model, so we note here the predicted pattern for each of the
strength may affect response times. To the extent that a clear sche- models (see Table 1). The self-schema model predicts a flatter re-
ma is available and is simpler than the self-schema or a general sponse time pattern for the fake-good and fake-specific conditions
good impression schema, the adopted schema model predicts an compared to the honest response condition and overall response
even more pronounced inverted-U pattern for impression-man- times that are significantly slower. The adopted-schema model
aged responding and overall responses times that are faster. predicts an inverted-U pattern for the fake-good condition that
166 M.K. Shoss, M.J Strube / Organizational Behavior and Human Decision Processes 116 (2011) 163–171

should be indistinguishable from that found in the honest- Instructions for non-specific good impression
responding condition. But, the adopted-schema model also makes Please answer the questions as if your responses were going to
a more specific prediction with regard to the specific instruction be read by potential employers for a job you really want or gradu-
conditions (especially for the extraversion dimension that under- ate school admissions committees for a school you really want to
lies these instructions) that allows it to be differentiated from hon- go to. You should try to present yourself in the best light possible
est responding. For the extraversion dimension, the inverted-U for so, for each item, select the response that you feel will give you the
the fake-specific public relations condition should be more pro- highest score and make you look like the best applicant.
nounced than for the other conditions. The instruction to fake
appropriateness for a public relations job represents a very specific Instructions for specific impression—public relations
and clear ideal that, if adopted, should produce a particularly pro- Please answer the questions as if you were the ideal candidate
nounced inverted-U pattern. Moreover, because both the public for a position in public relations. According to the top public rela-
relations and the accountant jobs likely represent clearer and more tions firms, an ideal candidate has the following qualities: outgo-
specific ideals than faking to make a general positive impression, ing, confident, credible, creative, takes initiative, and exercises
the overall response times for them should be significantly faster good judgment.
than for honest responding or general fake-good responding. The
semantic-exercise model predicts generally faster response times
Instructions for specific impression—accounting
for the faking conditions than for honest responding, and a less
Please answer the questions as if you were the ideal candidate
pronounced inverted-U effect.
for a position in accounting. An ideal candidate in accounting is de-
tailed-oriented, punctual, has good time management, and is
Method disciplined.

Participants Measures

Sixty-two participants were recruited from the Psychology Personality


Department Volunteer Participant Pool. Twenty-five participants The Revised NEO Personality Inventory (Costa & McCrae, 1992)
identified themselves as European Americans, 6 as African Ameri- is a 240-question assessment measuring the Five Factor Model
cans, 4 Hispanic, 11 Asian Americans, and 4 as other ethnicities. dimensions of personality: Neuroticism (Cronbach’s a = .90), Extra-
The average participant age was 19.38 (SD = 1.19). Two partici- version (a = .93), Openness to Experience (a = .85), Agreeableness
pants neglected to return for the second session, resulting in their (a = .91), and Conscientiousness (a = .91). Each dimension is mea-
first session data not being used in the analyses. Of the 60 partic- sured using 48 items, each rated on a 5-point Likert scale ranging
ipants included in the analyses, 22 were men and 38 were women. from strongly disagree (1) to strongly agree (5). This measure was
developed for use with adults without overt psychopathology
Design and has been found to have a high level of internal consistency
and test–retest reliability over long periods of time. For this reason,
This experiment used a within-subjects manipulation of re- it has been translated into many different languages and used in
sponse instructions: honest, fake-good, and fake-specific. The many different cultures.
fake-specific response instructions were manipulated between
subjects, resulting in two fake-specific conditions that instructed Results
participants to dissimulate so as to obtain a position in either
accounting or public relations. Although we expected the fake-spe- Faking estimates and manipulation check
cific conditions to be primarily a manipulation of the extraversion
dimension, we used job positions in order to increase realism and The effectiveness of the instructions was assessed by comparing
generalizability of our findings to selection contexts. In addition, scores from the NEO scale administered under the three instruc-
individual difference assessments (NEO-PI-R dimension scores) tion conditions in a 2 (Specific Instructions: Extravert versus Intro-
from the honest responding condition were incorporated into sta- vert)  3 (Type of Instruction: Honest, Good Impression, Specific
tistical analyses to determine if response-time patterns were mod- Impression) analysis of variance with the latter factor treated as
erated by particular personality characteristics. a repeated measure. Results and specific comparisons among con-
ditions for all scales are presented in Table 2. Based on previous re-
Procedure search demonstrating that applicants responding to the NEO PI-R
exhibited higher mean conscientiousness than non-applicants
The experiment was administered by computer and required (Griffin, Hesketh, & Grayson, 2004), we expected a significant in-
approximately one hour for the first session and half an hour for crease in conscientiousness scores when participants attempted
the second session. In the first session, participants were asked to make a good impression compared to their honest responses.
to respond honestly to the NEO PI-R (Costa & McCrae, 1992) and Also, based on prior faking research (McFarland, Ryan, & Ellis,
response times were recorded without participants’ knowledge. 2002; Vasilopoulos, Reilly, & Leaman, 2000), we expected elevated
At the end of the session, participants responded to a brief demo- extraversion and agreeableness scores and lower neuroticism
graphic questionnaire. scores in the fake-good condition compared to the honest respond-
During the second session, participants completed the NEO PI-R ing condition. However, we did not expect a significant difference
twice: (a) under non-specific instructions to give a good impres- in openness scores across conditions based on evidence suggesting
sion, and (b) under specific instructions to respond as an ideal can- that openness to experience is considerably difficult to fake (Griffin
didate for one of two jobs. All participants responded to the general et al., 2004).
instructions to fake good. Participants were randomly assigned to The most relevant NEO scale in our experiment was extraver-
just one of the specific instruction conditions. The computer pro- sion and the analysis of variance revealed a significant specific
gram counterbalanced the instructions and measured response instruction by type of instruction interaction, F(2, 116) = 20.77,
times. p < .001. As expected, extraversion scores were significantly higher
M.K. Shoss, M.J Strube / Organizational Behavior and Human Decision Processes 116 (2011) 163–171 167

Table 2
NEO dimension scores as a function of instruction condition.

Scale Honest (H) Fake-good (FG) Fake-specific


Extravert (FSE) Introvert (FSI)
M SD M SD M SD M SD
Neuroticism1 148.47 20.65 94.35 17.56 96.44 22.60 93.57 18.27
Extraversion2 170.12 24.48 192.65 13.02 202.80 14.60 164.30 23.19
Openness3 175.50 15.99 178.83 14.50 185.36 17.75 145.00 24.09
Agreeableness4 162.28 20.43 212.50 17.05 203.12 21.55 220.29 14.50
Conscientiousness5 165.02 20.20 187.30 15.31 174.44 23.40 163.37 21.10
1
F(2, 116) = 3.24, p < .01; t-tests revealed significant differences (p < .01) between H&FG, H&FSE, H&FSI.
2
F(2, 116) = 20.77, p < .01; t-tests revealed significant differences (p < .01) between H&FG, H&FSE, FSE&FSI.
3
F(2, 116) = 36.94, p < .01; t-tests revealed significant differences (p < .01) between H&FSI, FSE&FSI.
4
F(2, 116) = 9.37, p < .01; t-tests revealed significant differences (p < .01) between H&FG, H&FSE, H&FSI, FSE&FSI.
5
F(2, 116) = 4.90, p < .01; t-tests revealed significant differences (p < .01) between H&FG.

in the fake-good condition than in the honest responding condi- RTri is the average response time given by participant i for response
tion, in the fake-specific-extraversion condition than in the honest option r. ROri refers to the response option (1–5) chosen for ques-
responding condition, and in the fake-specific-extraversion condi- tions from one dimension of the NEO PI-R. The response option var-
tion than the fake-specific-introvert condition. Additionally, neu- iable was centered so that RO could take on values of 2, 1, 0, 1,
roticism scores were lower in all of the faking conditions than in and 2 for each individual. Homogeneity of error variance was as-
the honest responding condition, F(2, 116) = 3.24, p < .01. Similarly, sumed for each individual and confirmed by analysis of residuals.
agreeableness scores were higher in all of the faking conditions HLM assumes that residuals are randomly and normally distributed.
than in the honest responding condition, F(2, 116) = 9.37, p < .01. To achieve these assumptions, response time data were trimmed
Conscientious scores were significantly higher in the fake-good such that responses over 8 s were excluded from the analysis.
condition than in the honest responding condition, F(2, 116) = The Level 2 model for analysis of honest responding treats the
4.90, p < .01. The pattern of results clearly shows that participants Level 1 regression parameters (i.e., the intercept, linear, and qua-
successfully implemented both the general instructions to make a dratic effects of response options on response times) as outcomes
favorable impression and the more specific instructions targeted to to be predicted. Note that the Level 2 parameters are at the individ-
particular job characteristics. ual level and reflect the extent to which individual variability ex-
ists in the Level 1 parameters.
HLM analysis
p0i ¼ b00 þ r0i
We used hierarchical linear modeling (HLM) to analyze the re-
sponse time data because observations (i.e., response times) were
p1i ¼ b10 þ r1i
nested within individuals (Osborne, 2000). Specifically, each indi-
vidual had response times for each response option that he or p2i ¼ b20 þ r2i
she selected while responding to the NEO questionnaire. After re- Each parameter can be construed as fixed, randomly varying, or
verse coding appropriate items, we averaged the response times nonrandomly varying (i.e., moderators are present). Tests were
for each response option (e.g., agree) for a given scale (e.g., extra- conducted to determine the best model and in all cases either a
version) for each individual participant (and separately by condi- fixed or randomly varying model (shown above) was found to be
tion). We examined the NEO scales individually so that the optimal. The latter implies the presence of individual difference
honest and fake-good conditions could be easily compared to the moderators.
fake-specific condition, which we expected manipulated primarily Estimates and standard errors of the fixed coefficients are pre-
the extraversion dimension. Thus, each person had 25 dependent sented in Table 3, and results for the extraversion scale are also dis-
variables (response time averages; 5 per each of the 5 NEO scales) played graphically in Fig. 1. The analyses show a robust replication
for each experimental condition. HLM uses the data available in the of the inverted-U response time effect for each of the NEO PI-R sub-
case of individuals who did not select a given response option scales with negative and significant quadratic coefficients in each
when responding to items in a particular scale. The inverted-U ef- case. There were several significant linear effects as well; however,
fect was examined by testing for a quadratic trend in average re- these effects were weak when compared with the quadratic in-
sponse times across the five response options. Then, we used verted-U response time effect. Additionally, as can be seen in the
moderator variables to determine if the relationship between re- intercept parameter (p0i), the openness subscale was responded
sponse times and response options was general to all participants to more slowly overall than other subscales, supporting previous
or depended on the experimental manipulations and individual research suggesting that these items are more difficult to respond
differences. to than the other Big Five scales (Griffin & Hesketh, 2004). The ro-
bust replication of the inverted-U response time effect in the hon-
Honest responding model specification and analysis est responding condition provides the baseline against which to
assess response patterns in the faking conditions.
The initial HLM analysis tested the presence of the inverted-U The r terms corresponding to error were either fixed at zero or
pattern under the honest responding instructions. The Level 1 estimated based on chi-square tests in a preliminary analysis.
model represented the average response times for each participant These tests of the variance components were significant (p < .05)
as predicted by the selected response options, modeled to include for the intercept for each NEO subscale, the linear component for
the linear and quadratic relations between response option and re- the agreeableness subscale, and the quadratic component for
sponse time: extraversion, openness, agreeableness, and conscientiousness sub-
scales. The significant r terms suggest that there is variability
RT ri ¼ p0i þ p1i ROri þ p2i RO2ri þ eri
across individuals in terms of the size of these parameters (e.g.,
168 M.K. Shoss, M.J Strube / Organizational Behavior and Human Decision Processes 116 (2011) 163–171

Table 3
Honest-responding fixed effects estimation (with robust standard errors).

Scale Fixed effect Coefficient Standard error t df p


Neuroticism Intercept, p0i 3624.36 92.51 39.18 59 .001
Linear, p1i 62.16 27.15 2.29 271 .023
Quadratic, p2i 137.41 22.98 5.98 271 .001
Extraversion Intercept, p0i 3484.36 98.69 35.31 59 .001
Linear, p1i 44.28 26.11 1.70 261 .091
Quadratic, p2i 148.06 28.26 5.24 59 .001
Openness Intercept, p0i 4081.31 98.93 41.25 59 .001
Linear, p1i 37.79 40.18 0.94 59 .351
Quadratic, p2i 181.07 34.38 5.27 59 .001
Agreeableness Intercept, p0i 3516.84 92.19 38.15 59 .001
Linear, p1i 25.57 36.38 0.70 59 .485
Quadratic, p2i 87.75 30.09 2.92 59 .006
Conscientiousness Intercept, p0i 3687.11 97.21 37.93 59 .001
Linear, p1i 94.24 32.41 2.91 264 .004
Quadratic, p2i 89.55 33.30 2.69 59 .010

4400
Honest Responding
4200 Good Impression
4000 Specific Impression (Extravert)

3800 Specific Impression (Introvert)

3600
Response Time (ms)

3400

3200

3000

2800

2600

2400

2200

2000

1800
1 2 3 4 5
Response Option

Fig. 1. Response times as a function of response option for the extraversion dimension of the NEO PI-R under honest response, fake-good response, and fake-specific response
instructions.

in the overall response times or in the shape of the response time significantly as a function of the desire to make a generally good
distributions). Attempts to find individual difference moderators impression. Examination of response times under the ‘‘fake good’’
(NEO-PI-R dimension scores) were largely unsuccessful, however.1 instructions required a three-level model. As in the previous anal-
This suggests that there may be individual differences not measured ysis, the Level 1 model corresponded to response times per re-
in this study (e.g., alertness) that may cause systematic variability in sponse options, but unlike the previous analysis, required an
response times. additional subscript, s, indicating the nesting of response times un-
der the set of instructions (honest or good impression) under
Fake-good response time analysis which they were collected.

In the previous analysis, only the response times for the honest- RT rsi ¼ p0si þ p1si ROrsi þ p2si RO2rsi þ ersi
responding condition were examined. We next included both the
The Level 2 model examined the variability in the Level 1
honest and fake-good conditions in order to assess whether the
parameters and included a variable to account for the instructions
distribution of response times across response options varied
under which responses were made (honest versus good impres-
1
sion). Therefore, the p coefficients now include variability across
In an attempt to account for individual variability in the honest responding, fake-
conditions.
good, and fake-specific analyses, follow-up analyses were conducted in which
individual difference measures (NEO-PI-R dimension scores from the honest response
condition) were entered as predictors in the highest level model. Very few significant p0si ¼ b00s þ b01s Instructionsi þ r0si
effects emerged and no consistent patterns were found. Thus, it did not appear that p1si ¼ b10s þ b11s Instructionsi þ r1si
the personality traits included in this study were associated with differential abilities
to fake.
p2si ¼ b20s þ b21s Instructionsi þ r2si
M.K. Shoss, M.J Strube / Organizational Behavior and Human Decision Processes 116 (2011) 163–171 169

The estimates of the fixed coefficients are presented in Table 4. response instruction condition is compared to the honest response
Because the instruction condition was dummy coded with 0 as- condition (c211) and when it is compared to the fake-good condi-
signed to the honest responding condition, the significance of tion (c221). Both parameters were significant [c211, t(177) = 2.80,
the b20 coefficient corresponds to the curvature of the distribution p = .006; c221, t(177) = 2.11, p = .036]. These effects are displayed
of response times for the honest response condition. As indicated graphically in Fig. 1. The fake-specific public relations instruction
by the non-significant b21 coefficients, response distributions for resulted in a significantly more distinct inverted-U distribution
the neuroticism, extraversion, and openness scales displayed a than the fake-specific accountant instruction, t(118) = 2.56,
significant inverted-U shape in the fake-good condition that was p = .012. This between-subjects difference only occurred in the
not significantly different from the honest-responding inverted- specific instruction conditions. The c010 coefficient tests whether
U. Furthermore, the nonsignificant b01 coefficients for all but the honest responding condition had overall response times that
the openness scale indicate that overall response times in the were different from the public relations condition. This coefficient
fake-good condition were not distinguishable from the honest was significant [t(58) = 2.16, p = .035] and indicated that honest
responding condition. Taken together, these results are most con- responding was significantly slower. The c001 coefficient tests
sistent with the adopted schema model. Agreeableness and con- whether the accountant condition had overall response times that
scientiousness were the only two scales for which the curvature were different from the public relations condition. This coefficient
was different between the honest-responding and fake-good con- was significant [t(58) = 2.04, p = .046] and indicated that the
ditions, and in both cases the inverted-U pattern was more pro- accountant condition produced the fastest response times overall.
nounced in the fake-good condition. Likewise, the only overall By implication, this condition was also faster than honest respond-
response time difference occurred for openness and was faster ing. These results support the predictions made by the adopted
for the fake-good condition. These findings are consistent with schema model that there would be no difference in the quadratic
the adopted schema model under the assumption that the fake- effect between the honest and fake-good conditions, that the
good instruction called to mind a rather specific image to be fab- fake-specific public relations condition would exhibit the most
ricated with regard to these dimensions (at least more specific profound inverted-U relative to the other conditions, and that the
than the participants’ own self-concepts).2 These results are cer- specific impression conditions would produce faster responding
tainly inconsistent with the self-schema model and the semantic- than the honest response condition.3
exercise model.

Fake-specific response time analysis Discussion

Extraversion response times for the specific instruction condi- The purpose of this study was to evaluate the validity of three
tions were also examined in a three-level model that required a competing models of impression-managed responding to a person-
modification at Levels 2 and 3. At Level 2, an additional parameter ality test—the self-schema, adopted-schema, and semantic-exer-
was necessary to account for all three instruction conditions which cise models. Distinguishing between and testing these different
were dummy coded (specific: 0, 0; honest: 1, 0; fake-good: 0, 1): models is crucial for determining whether response times, specifi-
cally the inverted-U response time pattern, could be used to iden-
tify fakers. We assessed response time patterns for responses to the
p0si ¼ b00s þ b01s Instruction 1si þ b02s Instruction 2si þ r0si
NEO PI-R when participants were asked to respond honestly, make
p1si ¼ b10s þ b11s Instruction 1si þ b12s Instruction 2si þ r1si a good impression, and make a specific impression. The latter con-
p2si ¼ b20s þ b21s Instruction 1si þ b22s Instruction 2si þ r2si dition was included to test nuanced predictions made by these pre-
viously hypothesized models of impression-managed responding.
These b coefficients were modeled in a Level 3 model where the The inverted-U response time pattern found for each of the
fake-specific condition was dummy coded (public relations = 0; scales in the honest response condition replicates previous studies
accountant = 1). Recall that the fake-specific condition was a be- (Akrami et al., 2007; Kuiper, 1981). In addition, it provides the
tween-subjects manipulation where half of the participants were baseline for the analysis of response time differences between con-
instructed to respond as if applying for a job in public relations ditions. Our results support the predictions made by the adopted-
and the other half were instructed to respond as if applying for a schema model—that all conditions would produce an inverted-U
job in accounting: response time pattern, the inverted-U would be more pronounced
for the fake-specific public relations condition, and response times
b00s ¼ c000 þ c001 Specifici þ u00i would be faster overall for the fake-specific conditions. Thus, it ap-
pears that individuals, when faking, access a schema of an ideal
b01s ¼ c010 þ c011 Specifici þ u01i
respondent. This schema may be more or less clear depending on
b02s ¼ c020 þ c021 Specifici þ u02i whether one is trying to portray a particular impression or a more
b10s ¼ c100 þ c101 Specifici þ u10i general good impression.
b11s ¼ c110 þ c111 Specifici þ u11i The differences in the scale scores between conditions suggest
b12s ¼ c120 þ c121 Specifici þ u12i what the schemas used in faking might entail. The individual scale
scores for neuroticism, extraversion, agreeableness, and conscien-
b20s ¼ c200 þ c201 Specifici þ u20i
tiousness differed significantly when individuals responded
b21s ¼ c210 þ c211 Specifici þ u21i honestly compared to when they responded to make a good
b22s ¼ c220 þ c221 Specifici þ u22i impression, as expected. The differences in extraversion scores
between the conditions were particularly interesting. Supporting
The key parameter tests are for c010, c001, c211 and c221. The lat- prior research findings (Bagby & Marshall, 2003), extraversion
ter two indicate whether the inverted-U pattern is different be- scores were higher when participants were responding to make a
tween the two specific instruction conditions when the specific good impression than when they were responding honestly. Those

2 3
The Level 3 model estimated individual variability in the Level 2 parameters. No The Level 3 model also estimated individual variability in the Level 2 parameters.
interpretable effects were found. No interpretable effects were found.
170 M.K. Shoss, M.J Strube / Organizational Behavior and Human Decision Processes 116 (2011) 163–171

Table 4
Analysis of Level-2 model (instruction as a predictor) fixed effects estimation (with robust standard errors).

Fixed effect for: Coefficient SE t df p


Neuroticism Intercept, p0i b00 (intercept) 3625.38 92.87 39.04 59 .001
b01 (condition) 141.09 109.27 1.29 118 .199
Linear, p1i b10 (intercept) 61.09 27.13 2.25 118 .026
b11 (condition) 181.34 65.10 2.79 118 .007
Quadratic, p2i b20 (intercept) 134.92 22.86 5.90 118 .001
b21 (condition) 48.89 48.90 1.00 118 .320
Extraversion Intercept, p0i b00 (intercept) 3484.35 98.52 35.37 59 .001
b01 (condition) 119.94 116.75 1.03 118 .307
Linear, p1i b10 (intercept) 42.37 26.74 1.59 512 .113
b11 (condition) 112.47 47.89 2.35 512 .019
Quadratic, p2i b20 (intercept) 148.06 27.96 5.30 118 .001
b21 (condition) 33.79 41.62 0.81 118 .419
Openness Intercept, p0i b00 (intercept) 4080.90 99.07 41.19 59 .001
b01 (condition) 406.43 91.01 4.47 118 .001
Linear, p1i b10 (intercept) 40.68 40.43 1.01 118 .317
b11 (condition) 71.66 42.07 1.70 118 .091
Quadratic, p2i b20 (intercept) 179.03 34.54 5.18 118 .001
b21 (condition) 20.04 42.19 0.48 118 .635
Agreeableness Intercept, p0i b00 (intercept) 3513.86 91.76 38.29 59 .001
b01 (condition) 143.10 95.41 1.50 118 .136
Linear, p1i b10 (intercept) 20.12 36.18 0.56 59 .580
b11 (condition) 85.51 50.15 1.71 118 .090
Quadratic, p2i b20 (intercept) 86.85 30.05 2.89 59 .006
b21 (condition) 90.84 42.08 2.16 118 .033
Conscientiousness Intercept, p0i b00 (intercept) 3688.12 96.96 38.04 59 .001
b01 (condition) 153.91 111.30 1.38 118 .169
Linear, p1i b10 (intercept) 85.45 32.49 2.63 118 .010
b11 (condition) 31.97 64.66 0.49 118 .621
Quadratic, p2i b20 (intercept) 88.71 33.62 2.64 118 .010
b21 (condition) 109.91 53.79 2.04 118 .043

in the fake-specific-public relations condition inflated their extra- The prior literature on faking has suggested that a way to min-
version scores above their fake-good scores. Those in the fake-spe- imize the extent of faking would be to have individuals complete
cific accountant condition, however, reported essentially the same the personality assessment with as few specifics details of the
level of extraversion as they did when they were responding hon- job made known as possible (Vasilopoulos et al., 2000). This study
estly. This mirrors the results from the response-time analysis, and suggests that doing so would not result in lower levels of faking. In
suggests it may be more difficult to fake introversion than extra- fact, for the conscientiousness scale, it could result in a higher level
version; that is, relatively more thought may need to go into select- of response distortion. It is also interesting to note that the stan-
ing extreme responses. Perhaps it is difficult to fake introversion dard deviations for the scale scores decreased in the fake good con-
because the general desire to make a good impression in any job dition. This suggests that individuals shared a consolidated view of
interview calls for a more outgoing and friendly demeanor. There- not only what attributes are considered positive, but also at what
fore, the pressure to make a good impression and to appear to be levels.
an ideal applicant for an accountant position may cancel each Practically, in addition to suggesting that response time distri-
other, resulting in baseline (honest responding) level scores and butions, such as the inverted-U, would be ineffective for detecting
a flatter response time pattern than that for the fake-specific extra- faking, this study also casts doubt on the validity of response times
version dimension. However, because this schema is likely less for detecting faking in general. As previously noted, response times
complex than individuals’ self-schemas and schemas for a general can depend on a number of different factors including attention
good impression, response times overall were faster. These results, paid to each item, reading speed, time to search for information,
however, suggest that individuals are sensitive to stereotypes asso- schema clarity, familiarity with questions, test anxiety, and so
ciated with different positions and may reference these stereo- forth. Moreover, in practical settings it is nearly impossible to ac-
types when engaging in impression-managed responding (Mahar quire a baseline average response rate with which to compare, gi-
et al., 2006). ven the wide range of factors that can influence an individual’s
The analysis of the response time distributions for the impres- average response rate. Thus, it seems that researchers and practi-
sion-managed responses demonstrates that the inverted-U re- tioners interested in detecting and reducing faking would do well
sponse time effect cannot be used to discern those who are to focus on other strategies.
faking from those who are responding honestly. In fact, using an
individual’s response time distribution could lead to false predic- Limitations
tions on both the agreeableness and conscientiousness scales,
where the inverted-U effect was more pronounced in the fake- This study has a number of limitations that warrant additional
good condition than in the honest response condition. This sug- investigation. First, the fake-specific manipulation was not a direct
gests that a good-impression, in general, acts like a category and manipulation of extraversion. The manipulation used depended on
individuals can access this category readily when responding. This individuals’ stereotypes of these two positions, which are
also suggests that the self-category is not privileged in any way, an influenced by prior experience interacting with individuals in
implication consistent with the adopted-schema model of impres- these professions. Despite the fact that the strength of the manip-
sion-managed responding. ulation depended on the participants’ schemas, the results suggest
M.K. Shoss, M.J Strube / Organizational Behavior and Human Decision Processes 116 (2011) 163–171 171

that, on average, participants associated public relations with Griffin, B., & Hesketh, B. (2004). Why openness to experience is not a good predictor
of job performance. International Journal of Selection and Assessment, 12,
extraversion and accounting with introversion. Therefore, this
243–251.
study does appear to provide a manipulation, albeit imperfect, of Griffin, B., Hesketh, B., & Grayson, D. (2004). Applicants faking good: Evidence of
the extraversion dimension. In addition, the participants in our item bias in the NEO PI-R. Personality and Individual Differences, 36, 1545–1558.
sample were not actual applicants. It would be interesting to Griffith, R., & McDaniel, M. (2006). The nature of deception and applicant faking
behavior. In R. L. Griffith & M. H. Peterson (Eds.), A closer examination of
examine response time distributions in samples for which the rel- applicant faking behavior (pp. 1–20). Greenwich, Conn.: Information Age Publ.
evant schemas are especially salient and the application outcome Holden, R. R. (1998). Detecting fakers on a personnel test: Response latencies versus
more consequential. a standard validity scale. Journal of Social Behavior & Personality, 13, 387–398.
Holden, R. R., Fekken, G. C., & Cotton, D. H. (1991). Assessing psychopathology using
Second, due to the within-subjects design, participants in this structured test-item response latencies. Psychological Assessment, 3, 111–118.
study responded to the personality items three times. All partici- Holden, R., & Hibbs, N. (1995). Incremental validity of response latencies for
pants responded to the items in the honest responding condition detecting fakers on a personality test. Journal of Research in Personality, 29,
362–372.
first and to the other two conditions in random order. Therefore, Holden, R. R., & Kroner, D. G. (1992). Relative efficacy of differential response
participants’ responses to the later conditions may have been latencies for detecting faking on a self-report measure of psychopathology.
biased by their previous responses. However, if responses were Psychological Assessment, 4, 170–173.
Holtgraves, T. (2004). Social desirability and self-reports: Testing models of socially
influenced by prior responses then information should be easily desirable responding. Personality and Social Psychology Bulletin, 30, 161–172.
accessible because it was just retrieved. Therefore, we would ex- Hough, L., Oswald, F., & Ployhart, R. (2001). Determinants, detection and amelio-
pect that the distribution of response times would be flat. The fact ration of adverse impact in personnel selection procedures: Issues, evidence
and lessons learned. International Journal of Selection and Assessment, 9,
that we found an inverted-U distribution despite any potential
152–194.
practice effects suggests that our findings are robust. Hsu, L. M., Santelli, J., & Hsu, J. (1989). Faking detection validity and incremental
Additionally, given that the number of participants in this study validity of response latencies to MMPI subtle and obvious items. Journal of
is somewhat small, albeit similar to that typically used in other Personality Assessment, 53, 278–295.
Komar, S., Brown, D., Komar, J., & Robie, C. (2008). Faking and the validity of
within-subjects designs (e.g., Browne, Curley, & Benson, 1999), conscientiousness: A Monte Carlo investigation. Journal of Applied Psychology,
the power to detect quadratic effects may be reduced. Thus, this 93, 140–154.
made the statistical analysis relatively conservative. The fact that Kuiper, N. A. (1981). Convergent evidence for the self as a prototype: The ‘‘inverted-
U RT effect’’ for self and other judgments. Personality and Social Psychology
the inverted-U was found for all conditions again suggests that this Bulletin, 7, 438–443.
effect is particularly robust. Mahar, D., Coburn, B., Griffin, N., Hemeter, F., Potappel, C., Turton, M., et al. (2006).
In conclusion, this study suggests that individuals access a sche- Stereotyping as a response strategy when faking personality questionnaires.
Personality and Individual Differences, 40, 1375–1386.
ma of an ideal respondent when responding to give a specific McDaniel, M. A., & Timm, T. (1990). Lying takes time: Predicting deception in biodata
impression as well as to give general good impression. As clearly using response latency. In H. Wing (Chair), Alternative predictors. Symposium
defined exemplars are easier to retrieve, response times across presented at the 98th Annual conference of the American Psychological
Association, Boston, MA.
response options have an inverted-U shape. This indicates that McFarland, L., Ryan, A., & Ellis, A. (2002). Item placement on a personality measure:
response time distributions, such as the inverted-U, are not effec- Effects on faking behavior and test measurement properties. Journal of
tive in detecting faking. We encourage more research on the cogni- Personality Assessment, 78, 348–369.
Mesmer-Magnus, J., & Viswesvaran, C. (2006). Assessing response distortion in
tive processes involved in responding to questionnaires as well as
personality tests: A review of research designs and analytic strategies. In R. L.
ways to mitigate or detect faking. Griffith & M. H. Peterson (Eds.), A closer examination of applicant faking behavior
(pp. 21–42). Greenwich, Conn.: Information Age Publ.
Mueller, J. H., Thompson, W. B., & Dugan, K. (1986). Trait distinctiveness and
References accessibility in the self-schema. Personality and Social Psychology Bulletin, 12,
81–89.
Mueller-Hanson, R., Heggestad, E. D., & Thornton, G. C. (2003). Faking and selection:
Akrami, N., Hedlund, L. E., & Ekehammar, B. (2007). Personality scale response Considering the use of personality from select-in and select-out perspectives.
latencies as self-schema indicators: The inverted-U effect revisited. Personality Journal of Applied Psychology, 88, 348–355.
and Individual Differences, 43, 611–618. Ones, D. S., Dilchert, S., Viswesvaran, C., & Judge, T. A. (2007). In support of
Bagby, R. M., & Marshall, M. B. (2003). Positive impression management and its personality assessment in organizational settings. Personnel Psychology, 60,
influence on the revised NEO personality inventory: A comparison of analog and 995–1027.
differential prevalence group designs. Psychological Assessment, 15, 333–339. Osborne, Jason. W. (2000). Advantages of hierarchical linear modeling. Practical
Birkeland, S. A., Manson, T. M., Kisamore, J. L., Brannick, M. T., & Smith, M. A. (2006). Assessment, Research & Evaluation, 7, 1–3.
A meta-analytic investigation of job applicant faking on personality measures. Popham, S. M., & Holden, R. R. (1990). Assessing MMPI constructs through the
International Journal of Selection and Assessment, 14, 317–335. measurement of response latencies. Journal of Personality Assessment, 54,
Browne, G., Curley, S., & Benson, P. (1999). The effects of subject-defined categories 469–478.
on judgmental accuracy in confidence assessment tasks. Organizational Behavior Vasilopoulos, N. L., Reilly, R. R., & Leaman, J. A. (2000). The influence of job
and Human Decision Processes, 80, 134–154. familiarity and impression management on self-report measure scale scores
Casey, M. M., & Tryon, W. W. (2001). Validating a double-press method for and response latencies. Journal of Applied Psychology, 85, 50–64.
computer administration of personality inventory items. Psychological Viswesvaran, C., & Ones, D. S. (1999). Meta-analyses of fakability estimates:
Assessment, 13, 521–530. Implications for personality measurement. Educational and Psychological
Costa, P. T., Jr., & McCrae, R. R. (1992). NEO PI-R professional manual. Odessa, FL: Measurement, 59, 197–210.
Psychological Assessment Resources, Inc.

Вам также может понравиться