Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
SOCIAL NEUROSCIENCE, 2008, 3 (2), 164177
EEG components of spontaneous trait inferences
Marijke Van Duynslaeger
Vrije Universiteit Brussel, Brussels, Belgium
Caroline Sterken
Vrije Universiteit Brussel, Brussels, Belgium
Frank Van Overwalle
Vrije Universiteit Brussel, Brussels, Belgium
Edwin Verstraeten
Swansea University, Swansea, UK
Can event-related electro-encephalogram (EEG) responses provide support for the occurrence of
spontaneous trait inferences (STI)? Participants read sentences describing the behavior of a target
person from which a strong trait could be inferred. The last word of each sentence determined the
consistency with the trait induced during an introductory paragraph. In comparison with sentences that
were consistent with the implied trait, when the sentences were inconsistent, an event-related P300
waveform was observed at parietal scalp regions (Pz). This dependency on behavioral consistency
indicates that trait inferences were made spontaneously. Memory measures taken after the presentation
of the stimulus material involved sentence completion and trait-cued recall, and also supported the
occurrence of spontaneously inferred traits associated with the actor. Interestingly, increased memory for
consistent relative to inconsistent behaviors at the trait-cued recall task was significantly correlated with
the P300, which supports this latter measure as a valid neural correlate of spontaneous trait inferences.
INTRODUCTION
Imagine the following situation.
Ann is talking to her boyfriend, Thomas, about
her new hobby, bungee jumping. Thomas raises
his eyebrows when Ann says she’s going to
jump from the Eiffel Tower tomorrow and he
says: ‘‘But Ann, yesterday you didn’t even dare
to climb the ladder to pick cherries!’’
While reading the first sentence we might
spontaneously infer that Ann is an ‘‘adventurous’’
person. However, as we read further it seems like
she didn’t even dare to climb the ladder to pick
cherries, which violates our initially formed expectation and impression about Ann. In this
article, we explore whether such inconsistencies
with spontaneous trait inferences (STI) result in
different neural responses compared to information that is consistent with the inferred trait. If
Correspondence should be addressed to: Frank Van Overwalle, Department of Psychology, Vrije Universiteit Brussel, Pleinlaan
2, Brussels, B-1050, Belgium. E-mail: [email protected]
This research was supported by an OZR Grant of the Vrije Universiteit Brussel to Edwin Verstraeten. We are very grateful to
Bruce Bartholow for providing his experimental stimulus material. Caroline Sterken is now at the Neuropsychological Lab,
Department of Medicine, Catholic University of Leuven, Belgium.
# 2008 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
www.psypress.com/socialneuroscience
DOI:10.1080/17470910801907226
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
EEG AND SPONTANEOUS TRAIT INFERENCES
our brain responds differently to behavioral
statements that either confirm or violate previously established trait-based expectancies, this
neural response can be used as a novel measure to
index that a STI has actually been made previously. Moreover, research into such neural
responses is critical to understanding the interaction between top-down expectancy-driven processes and bottom-up processing of novel
information in impression formation, and as
such can shed more light on the underlying brain
processes involved in spontaneous inferences
about other persons.
STI can be defined as impressions that are
formed about a target person without intention or
awareness (Uleman, 1999; Uleman, Blader, &
Todorov, 2005; Uleman, Newman, & Moskowitz,
1996). They are relatively automatic (Bargh,
1989) in the sense that they require little mental
effort, are difficult to suppress, and are hard
to interfere with, although minimal mental capacity can reduce STI for counterstereotypical
behaviors (Wigboldus, Sherman, Franzese &
van Knippenberg, 2004) and some goals can
reduce STI, such as deciding whether or not the
communicated information is false (Skowronski,
Carlston, Mae & Crawford, 1998). In contrast,
intentional trait inferences (ITI) are made with
the explicit intention to form an impression about
the target. Hence, ITI are under the control of the
perceiver and require some amount of mental
capacity to be performed.
Earlier research and measures of STI
In previous research on STI, various psychological measures have been used to demonstrate the
occurrence of STI. One of the earliest paradigms
was cued recall, developed by Winter and Uleman
(1984). When people make STI while observing
behaviors, the inferred traits are assumed to be
stored in memory together with the behavioral
information from which they are inferred. As a
result, these traits are effective retrieval cues for
the behavioral information. Winter and Uleman
(1984) found that trait cues are stronger aids to
recall than semantic cues that were a priori
related with the actor or other sentence parts, or
when no cue is given (non-cued recall). A
disadvantage of the cued recall measure is that
there is a considerable time delay between the
assumed spontaneous inference process and
the memory measures to detect the inferences,
165
so that it is not very clear whether either
spontaneous encoding or more strategic (i.e.,
intentional) retrieval processes are responsible
for the enhanced memory. For instance, participants may use the trait cue to build a gist meaning
of the behavior or may actively search for
behavioral associates, which then lead them to
the correct sentences. Another shortcoming is
that it is unclear whether the implied trait reflects
an impression of the actor or only an interpretation of the behavior (Carlston & Skowronski,
1994; Van Overwalle, Drenth & Marsman, 1999).
This latter limitation was largely overcome in
more recent memory tasks, such as relearning
(Carlston & Skowronski, 1994) and false recognition (Todorov & Uleman, 2002), which measure
the link between a photo of the actor and the
implied trait, using more implicit measures (i.e.,
facilitation or speeding up while relearning the
material and recognition errors respectively).
However, these measures share with trait-cued
recall the disadvantage that there is a considerable delay between encoding and retrieval,
although the likelihood of possible strategic
processes is typically minimized by presenting a
large set of behaviors. This delay disadvantage
is largely overcome by online memory techniques, such as the probe recognition task (e.g.,
McKoon & Ratcliff, 1986; Van Overwalle et al.,
1999), which measures recognition of words
immediately after reading of each behavioral
sentence. Unfortunately, such online tasks cannot
be employed in the present research, because
they interfere with the measurement of an EEG
during ongoing spontaneous social processes. In
sum, recent research using more refined implicit
techniques confirmed that trait-cued recall is
indeed a valid measure of STI, and that strategic
retrieval during a delay can be avoided by
presenting an abundance of stimulus material.
However, trait-cued recall clearly does not provide evidence for a traitactor link. As we will
describe shortly, we developed an extension to
this recall task to overcome this limitation.
In addition to STI research, it has been shown
that behavior that is inconsistent with one’s
expectations (e.g., impressions, stereotypes) about
a person is often recalled better than consistent
behavior (Hastie, 1980; Hastie & Kumar, 1979;
Stangor & McMillan, 1992; Srull & Wyer, 1989).
It is generally assumed that such unexpected
information receives more cognitive processing
and is therefore recalled better than expected
behavior. To some extent, recall of inconsistent
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
166
VAN DUYNSLAEGER ET AL.
behaviors can be considered a measure of (spontaneous) person inferences established previously,
if perhaps not in terms of specific traits, at least in
terms of one’s general valenced impression of a
person (e.g., positive or negative). If inconsistencies are not tied to an impression or expectation
about the person, then no inconsistency resolution processes emerge. For instance, when an
impression or stereotype is formed on a group of
loose individuals rather than a single individual,
there seems to be decreased rather than enhanced
memory for stereotype-inconsistent behaviors
(for a review, see Fyock & Stangor, 1994).
Although largely neglected in STI research,
earlier work on person perception has documented that impression judgments are more influenced by an actor’s negative behaviors than
positive behaviors (Cacioppo, Gardner, &
Berntson, 1999; Ybarra, 2002). Thus, negative
behaviors that disconfirm a positive trait expectancy are more effective at changing impressions
than positive disconfirming behaviors. This asymmetry has been most often explained by the
differential diagnosticity of negative versus positive social behaviors (Reeder & Brewer, 1979;
Skowronski & Carlston, 1989). When one is
making inferences about honesty and kindness,
negative behaviors are more diagnostic because
moral actors are constrained to display only
moral behaviors (e.g., an honest person never
cheats), while immoral persons are free to
exhibit either moral or immoral behaviors
(Lupfer, Weeks & Dupuis, 2000; Reeder, 1997;
Reeder & Fulks, 1980; Reeder & Spores, 1983;
Skowronski & Carlston, 1987). An additional
question of this research is whether this
positivenegative asymmetry will also be revealed
for EEG components of STI.
Event-related potentials (ERP)
The important innovation of EEG measures is
that they allow measurement of the neural
correlates of STI. As far as we are aware, EEG
measures have not been used previously for
detecting STI, although they have been applied
under intentional trait instructions by Bartholow
and his colleagues (Bartholow, Dickter & Sestir,
2006; Bartholow, Fabiani, Gratton & Bettencourt,
2001; Bartholow, Pearson, Gratton, & Fabiani,
2003). In these studies, ERP were recorded
while participants read behavioral statements
that either confirmed or violated previously
established target-based expectancies (i.e., person
impression) in order to track the neural activity
associated with inconsistency resolution and to
examine how this activity relates to later recall
(Bartholow et al., 2003, 2006).
The P300 and evaluative inconsistency
ERP are electronic waves measured at different scalp locations in response to discrete events
and are assumed to reflect information processing
in the brain after the occurrence of the event.
Previous research documented an increased positivity in ERP amplitude*known as P300 or also
as late positive potential*beginning at about
300400 ms after the encoding of discrepant
information and often continuing till 600 or
1000 ms poststimulus (for a review, see Nieuwenhuis, Ashton-Jones, & Cohen, 2005). For brevity,
we use P300 throughout this manuscript to refer
to this late positive shift in amplitude. It has been
documented that there is a relation between the
P300 and the processing of anomalous, inconsistent, or infrequent stimuli presented in a context
of otherwise normal or frequent information, as
long as this information is relevant for the task.
The amplitude of the P300 increases as a function
of the amount of discrepancy between the stimulus and the preceding context, and correlates with
later recall of the discrepant stimuli, especially
when elaborate rehearsal strategies are minimized (Andreassi, 2000; Fabiani, Karis, &
Donchin, 1986; Fabiani & Donchin, 1995). These
findings have led to the view that the P300 is an
index of online updating of working memory after
inconsistency detection, and that its amplitude
and timing reflect the amount and duration of
information processing in the brain. These characteristics make ERP ideally suited for exploring
the neural correlates of inconsistency resolution,
and provide an advantage over the use of memory
measures alone to infer STI in studies like this.
The P300 and person perception
In social research, Cacioppo and coworkers
(Cacioppo, Crites, Berntson, & Coles, 1993;
Cacioppo, Crites, Gardner, & Berntson, 1994)
found that an evaluative inconsistency between a
trait word and previously presented trait words
(e.g., a negative trait after a sequence of positive
traits) elicited a large P300 between approximately 500 and 1000 ms at the central and parietal
scalp. Note that the trait was not inferred here,
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
EEG AND SPONTANEOUS TRAIT INFERENCES
but simply given. Hence, of more interest is work
by Bartholow and colleagues (Bartholow et al.,
2001, 2003), who instructed their participants to
form impressions about actors engaged in a series
of behaviors depicted in short sentences. These
behaviors either implied the same trait (traitconsistent sentences) or violated the implied trait
(trait-inconsistent sentences). The inconsistency
of the sentence was evident only after reading the
last, critical word of the sentence, which allowed
measure of ERP potentials that were time-locked
on the last word.
Bartholow et al. (2001, 2003) found greater
P300 activation at 300800 ms after presenting the
critical word for trait-inconsistent sentences as
opposed to trait-consistent sentences. Because a
P300 indicates the detection of a violation of an
expectancy generated by the previous stimulus
sequence history (Nieuwenhuis et al., 2005), the
enhanced P300 amplitude for the trait-inconsistent sentences can be interpreted as indicating
that traits had been inferred earlier. Interestingly,
Bartholow et al. (2003) found this P300 effect
only after discrepant negative behaviors following
a positive trait expectancy, consistent with earlier
research on the higher diagnosticity of negative social behaviors (Reeder & Brewer, 1979;
Skowronski & Carlston, 1989; Ybarra, 2002) and
stronger ERP after negative inconsistencies embedded in series of positive trait words (for a
review, see Cacioppo et al., 1999). The increase in
P300 was paralleled by enhanced memory performance for trait-inconsistent sentences in comparison with trait-consistent sentences on a sentence
completion task. As noted earlier, a memory
advantage for inconsistent information in person
perception is often explained in terms of deeper
processing and greater cognitive activity required
to reconcile the inconsistent information with an
already formed person impression (Hastie, 1980;
Hastie & Kumar, 1979; Srull & Wyer, 1989;
Stangor & McMillan, 1992).
Scalp locations of the P300
Research with brain-imaging techniques has
suggested that the P300 originates from working
memory tasks in the middle frontal and parietal
lobes (McCarthy, Luby, Gore, & Goldman-Rakic,
1997). There is also evidence suggesting that the
temporo-parietal junction is a widespread key
neural generator of P300 (see Nieuwenhuis
et al., 2005). Similarly, recent functional magnetic
resonance imaging (fMRI) evidence suggests that
167
two brain areas involved in the understanding and
attribution of mental states (i.e., goals and traits)
of others are the temporo-parietal junction and
the medial frontal cortex (see Frith & Frith, 2001;
Harris, Todorov, & Fiske, 2005), although recent
research seems to suggest that the latter is
more essential in attributing traits (see Mitchell,
Banaji, & Macrae, 2005; Todorov, Gobbini,
Evans, & Haxby, 2006). A recent study by
Mitchell, Cloutier, Banaji, and Macrae (2006)
compared intentional with spontaneous (i.e.,
memory) instructions while participants were
scanned using fMRI, but these instructions were
alternated between trials so that it is unlikely that
the ‘‘spontaneous’’ trait inferences were made
without any awareness and intention. Although
the spatial resolution of EEG waves is generally
very poor so that they are difficult to compare
with fMRI data, this nevertheless suggests that
event-related EEG responses in response to STI
are most likely to be found in parietal scalp
regions.
Present research and hypotheses
To study the ERP correlates of STI about others,
we modified Bartholow et al.’s (2001, 2003)
inconsistent behavioral information paradigm
for spontaneous inferences, by instructing our
participants to read the stimulus material carefully, without mentioning anything about person
traits or impressions. Apart from some additional
modifications discussed shortly, the other basic
aspects of the paradigm were largely left unaltered. Specifically, advanced information was
provided to spontaneously build up a trait expectation about an actor, and then several sentences depicting the behavior of the actor were
shown. The last word in each sentence provided
information that was consistent, inconsistent, or
irrelevant with respect to the trait. With respect to
the main goal of this article, we expected to
observe a P300 waveform given behaviors that
violated the previously inferred trait, especially if
the discrepancy involved negative behaviors that
violated a positive trait (cf., positivenegative
asymmetry). We measured the ERP at midline
scalp locations (i.e., between the two hemispheres) at the frontal, central, and parietal lobes.
On the basis of earlier EEG work especially by
Bartholow et al. (2001, 2003), we expect that the
P300 appears within a 300600 ms window mainly
at the parietal scalp site, and most likely after
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
168
VAN DUYNSLAEGER ET AL.
negative trait discrepancies (Bartholow et al.,
2003).
An additional modification to Bartholow
et al.’s (2001, 2003) paradigm and typical STI
research is that we investigated whether inconsistency detection was sensitive not only to
evaluative discrepancies of the behavior, but
also to descriptive discrepancies (Stangor &
McMillan, 1992). All the traits implied in
Bartholow et al.’s research involved moral behaviors, which capture only one dimension of
personality. To explore the breath of inconsistency resolution and the scope of STI with respect
to the descriptive domain of personality, we also
included behavioral inconsistencies with respect
to another dimension of personality, namely,
competence. If STI are sensitive to such descriptive discrepancies, they should also elicit a P300.
Another modification is that we took several
memory measures of STI as behavioral validation
of the ERP measures, including sentence completion and cued recall. These memory tasks were
taken after the presentation of all stimulus
material so that they did not interfere with the
ongoing EEG measures involving inconsistency
detection. Moreover, these memory measures
take a somewhat different interpretation than in
earlier STI research because of some essential
differences with the present paradigm.
The sentence completion task, borrowed from
Bartholow et al. (2001, 2003), consists of completing the last, trait-implying word of the original
sentence. Based on prior research documenting
better memory for inconsistent information
(Hastie, 1980; Hastie & Kumar, 1979; Srull &
Wyer, 1989; Stangor & McMillan, 1992), we
expect better sentence completion scores for
trait-inconsistent behaviors as opposed to traitconsistent behaviors. This task has not been used
earlier in STI studies, because each behavior in
these studies typically involved a different actor
so that trait inconsistencies were absent in the
material. In contrast, in the present paradigm, the
behaviors are not independent but form a collection performed by the same actor. Hence, increased memory in the sentence completion task
suggests a violation with an impression formed
earlier about the actor, although this impression
might reflect a general person evaluation rather
than a specific trait.
To firmly establish that also traits were inferred, we added a memory task that rests on
traits, the trait-cued recall task originally developed by Winter and Uleman (1984). We also
included recall cued by antonym traits, that is, the
opposite traits implied by the trait-inconsistent
behaviors. In line with earlier research, we
expected better trait-cued recall for consistent
behaviors that were implied by the trait, as
opposed to trait-inconsistent sentences. Conversely, we expected enhanced antonym-cued recall
for inconsistent behaviors implied by the antonym traits, as opposed to trait-consistent behaviors. This enhanced recall cued by the trait or its
antonym suggests that trait interpretations were
made about consistent and discrepant behaviors
respectively. More importantly, to firmly establish
in this study that an association was developed
between the trait and the actor, trait-cued recall
should be higher than antonym-cued recall,
because the actortrait association serves as an
additional retrieval cue for trait-cued recall of
consistent behaviors, but not for antonym-cued
recall because the antonym is not associated with
the actor (but only with the discrepant behavior).
Finally, if the ERP reflects inconsistency processing during spontaneous processing, we predict
a significant correlation between the P300 and
sentence completion (as it tracks the inconsistent
behaviors) or trait-cued recall (as it tracks the
traits these behaviors are inconsistent with).
Although both memory tasks may validate the
P300 as measure of spontaneous inference, a
correlation with trait-cued recall provides more
direct validity for our hypothesis that traits are
inferred rather than, for instance, general person
impressions. Moreover, because inconsistencies
are infrequent, we do not expect our participants
to form firm traits about them, so that a strong
correlation with antonym-cued recall is unlikely.
To recapitulate, we predict better memory for
sentence completion and antonym-cued recall for
inconsistent sentences and better trait-cued recall
for consistent sentences. This may seem contradictory at first sight. However, one must take into
account the different modes of retrieval. Retrieval during the sentence completion task is aselective, and thus benefits from general enhanced
memory due to inconsistency detection. In contrast, during cued recall, retrieval occurs selectively with the aid of (the trait or antonym as) a
cue, and therefore enhanced recall is expected
only when that trait or its antonym was actually
inferred while reading the behavioral sentences.
In addition, we expect at least one of the memory
tasks to correlate with ERP.
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
EEG AND SPONTANEOUS TRAIT INFERENCES
METHOD
Participants
Participants were 26 students at the Vrije Universiteit Brussel (VUB). There were 20 women
and 6 men, with ages varying between 19 and 27,
and an average of 22.091.9. All participants were
recruited via a university-wide electronic mailing
system for all university students. In exchange for
their participation, they were paid 15 euros. Of
the participants, 18 received a typical STI instruction to read the sentences (13 women and 5 men,
average age 22.491.9). Another 8 participants
also received this instruction, but were additionally instructed to look for inconsistencies
(7 women and 1 man, average age 2191.9).
This latter instruction was added halfway through
the experiment in the hope of making the
detecting P300 patterns more likely. However,
this added instruction had no effect and did not
reveal any significant differences with the read
instruction. Participants were initially assigned to
the read instruction only, and about halfway
through the experiment they were randomly
assigned to the read and inconsistency instruction
conditions.
Stimulus material
The design and stimulus material were borrowed
from Bartholow et al. (2001, 2003), with some
important modifications. Participants read 20
introductory paragraphs that described the general behavior of a fictitious target person and
from which a strong trait could be inferred. The
paragraphs involved 10 positive and 10 negative
moral traits, and each paragraph was shown for
30 s on the computer screen. To avoid association with a familiar and/or existing name,
fictitious ‘‘Star Trek’’-like names were used
(Bartholow et al., 2001, 2003). For example,
this paragraph described the general behavior
of target person ‘‘Tolvan’’: Tolvan smiles at
everyone on the way to work. Whenever it snows,
Tolvan shovels her elderly neighbor’s walk. Tolvan always stops to help when she sees someone
with car trouble. Tolvan’s coworkers are all
quite fond of her. This paragraph implies that
‘‘Tolvan’’ is a friendly person. After each paragraph, a series of 12 behavioral sentences was
169
presented, each consisting of 6 words shown in
the center of the computer screen. Every word
was presented for 300 ms, followed by a 350 ms
blank (Osterhout, Bersick, & McLaughlin, 1997).
The last word of each sentence was the critical
one, because it determined the degree of consistency with the previously inferred trait: traitconsistent (TC), trait-inconsistent (TI), competence-inconsistent (CI) and irrelevant (IRR).
TC-sentences describe moral behaviors that are
consistent with the inferred trait (for example
‘‘Tolvan gave her sister a hug’’). TI-sentences
(for example ‘‘Tolvan dared the stranger to
fight’’) are evaluatively inconsistent with the
inferred trait (here ‘‘friendly’’) but describe the
same personality content. The CI-sentences describe competence-related behaviors (for example ‘‘Tolvan obtained for math an F’’) that are
inconsistent with the inferred trait in regard to
both valence and descriptive content. The IRRsentences describe neutral behaviors (for example ‘‘Tolvan gave her mother a bottle’’).
After each introductory paragraph, a series of
12 behavioral sentences was presented. These
consisted of 4 filler sentences, always ending
with TC-behavior, followed by the 8 experimental
sentences, consisting of 2 TC-, 2 TI-, 2 CI- and 2
IRR-behaviors presented in a random order. All
the material was borrowed from Bartholow et al.
(2001, 2003) and translated from English into
Dutch while keeping the same number of words
(which sometimes required us to develop different sentences implying the same trait), except
for the CI-sentences, which were all newly developed in Dutch. All Dutch sentences were pilot
tested (n199) to check if they reflected the
expected degree of consistency with the inferred
trait, and results showed that the rank order of
the ratings (on a 010 scale) from TC-, TI-, CI- to
IRR-sentences was correct for all traits. A similar
pilot test (n83) was performed for the novel CIsentences to check whether they reflected the
intended low or high competence, and the results
revealed that CI-sentences reflecting high competence received higher ratings (M8.0 on a
010 scale) than those reflecting low competence
(M3.7). Eight antonym cues were selected
from the 10 original trait cues, and an additional
pilot test (n35) confirmed that the related TIsentences strongly reflected these antonym traits
(M7.9 on a 010 scale).
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
170
VAN DUYNSLAEGER ET AL.
Procedure
After the participant was seated, the electrodes
for the EEG and electooculogram (EOG) were
placed at the correct locations. The instructions
were presented on a computer screen. The
participants were informed that they would read
stories about several persons and that each story
would start with a paragraph followed by different sentences about it. Because participants could
not read at their own pace as in previous STI
research, but were shown each word at a fixed
pace of 300 ms, they were also told to pay as much
attention as possible to each word, because they
would get questions about it afterwards (for
similar instructions, see Todorov & Uleman,
2002). It was also emphasized to move and eyeblink as little as possible to limit artifacts in the
EEG (Stern, Ray, & Quigley, 2001). For the read
version of the STI instruction, participants were
told once more to read as attentively as possible.
For the inconsistency version of the STI instruction, participants were asked to be alert for
inconsistencies in the presented information.
During reading the EEG was recorded. Afterwards the electrodes for the EEG and EOG were
removed. Next, the participants were given the
cued recall and the sentence completion task in
the same order for all participants. In the cued
recall task, participants had to write as many
behavioral sentences as possible with the aid of
words that consisted of the implied and antonym
traits. There were trait cues for all 20 series of
behavioral sentences, while these same traits
served as antonym cues for 16 of these 20 series.
In the sentence completion task, participants
were presented with incomplete TC-, TI- and
CI-sentences and had to complete the last word.
Electrophysiological registration and
analysis
The EEG was recorded at midline frontal,
central, and parietal regions (Fz, Cz, and Pz)
according to the international 1020 electrode
system, using electrodes fixed in a stretch head
cap (Activeshield) from Advanced Neuro Technology. The linked electrodes at the left and right
mastoids served as the offline reference. A
ground electrode was located along the midline
between the Fz and Cz electrodes. Vertical
and horizontal eye movements were recorded
bipolarly via EOGs using electrodes placed above
and below the left eye and 1 cm external to the
outer canthus of each eye, respectively. Impedance was kept below 10 kV. The EEG was
recorded continuously during the whole experiment, with a digitizing rate of 256 Hz. Stimulus
presentation, recording and analysis were done
with the hardware (Cognitrace) and software
(Eevoke, Eemagine, and ASA) from Advanced
Neuro Technology. The stimulus material was
presented on screen and directly time-locked on
the EEG recordings by the EEVoke software.
The raw EEG data were filtered by a 0.03 to
30 Hz band-pass, horizontal and vertical EOG
artifacts were corrected using the SOBI algorithm
(Joyce, Gorodnitsky, & Kutas, 2004), and remaining artifacts beyond 75 and 75 mV were removed before analysis. For ERP, the relevant
EEG sequences were averaged per participants,
channel, and condition. Each relevant sequence
began 250 ms prior to the presentation of the last
critical word in each sentence (prestimulus baseline) and lasted to 1125 ms after the presentation
of the critical word, leaving 77% artifact-free
sequences. A grand average was then calculated
on these individual ERP data across all participants.
RESULTS
Electrophysiological data
To analyze the time course of positive and
negative peaks in the ERP amplitude, we divided
the ERP data of each participant in sequential
time intervals (50300, 300450, 450600, and
6001000 ms) in much the same manner as
Bartholow et al. (2001).
We first conducted a preliminary check to
ascertain whether the additional instruction for
some participants (i.e., to look for inconsistencies) had any effect on our ERP data. Therefore,
the largest positive and negative peaks (or maximum and minimum amplitudes respectively) in
each interval were identified and statistically
analyzed for each of the three channels by means
of a repeated measures analysis of variance
(ANOVA) with Trait Context (positive, negative), Consistency (TC, TI, CI, IRR), and Interval
(50300, 300450, 450600, 6001000 ms) as
within-participants factors and Instruction (read,
inconsistency) as a between-participants factor.
The ANOVA revealed for none of the channels
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
EEG AND SPONTANEOUS TRAIT INFERENCES
and peaks any significant main or interaction
effect with Instruction, so that we collapsed all
further analyses along the Instruction factor. Of
interest, the analysis also revealed a consistent
main effect of Interval for the positive peaks in all
channels, F(3, 69)4.197.73, pB.01.
Figure 1 shows the ERP grand averages
(positive amplitudes are directed downward)
revealing that negative TI behaviors violating a
positive moral trait generally elicits a larger
positive P300-like waveform than TC behaviors.
To test our specific hypothesis, we conducted
planned contrasts using t-tests for evaluative (TC
vs. TI) inconsistencies separately for positive and
negative behaviors on the peaks in the 300600
interval. This analysis confirmed that there were
greater positive peak amplitudes at the Pz scalp
region in the 450600 ms interval for negative TI
171
behaviors (M3.96 mV) in comparison with
positive TC behaviors (M2.92 mV), t(24)
2.50, pB.05, which lasted into the next 600
1000 ms interval, t(24)2.30, pB.05 (M5.09
and 3.19 mV respectively). The same effect for the
Cz location approached significance in the earlier
300450 ms interval (p.064; M2.63 and
1.65 mV respectively). No other effects for positive peaks and none of the effects for negative
peaks (e.g., N400) were significant in the expected
direction.
For explorative purposes, we also analyzed
descriptive (TC vs. CI) inconsistencies. Figure 2
depicts the ERP grand averages showing that
negative competence-related CI violations elicit a
brief positive waveform. Planned comparisons
using t-tests revealed a significant effect in the
300450 ms interval at the Fz site indicating
Figure 1. Effects of negative evaluative trait-inconsistency on grand-averaged ERP waveforms at midline parietal scalp sites
showing a P300 positive deflection. Dark lines denote TI; light lines denote TC. A positive amplitude is shown downward. The small
arrows show the approximate onset of the P300.
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
172
VAN DUYNSLAEGER ET AL.
Figure 2. Effects of negative descriptive trait-inconsistency on grand-averaged ERP waveforms at midline parietal scalp sites
showing a P300 positive deflection. Dark lines denote CI; light lines denote TC. A positive amplitude is shown downward. The small
arrows show the approximate onset of the P300.
that their were greater positive peak amplitudes
after inconsistent competence-related behaviors
(M3.01 mV) than after consistent moral behaviors (M2.12 mV), t(24)2.13, pB.05, and also
at the Cz site which only approached significance
(p.058, M2.57 and 1.65 mV respectively).
However, none of these effects survived a Bonferroni correction taking into account the number
of channels and comparisons between consistency
conditions (32). No other effects were significant.
In addition, we analyzed the latencies of the
positive and negative peaks comparing consistent
behaviors with moral (TI) or competence-related
(CI) inconsistencies following positive or negative
traits, and found only occasional differences
beyond the .05 a-level (i.e., 6 out of 96 possible
comparisons), which did not survive a Bonferroni
correction.
These findings seem to support our hypothesis
that trait inferences were made spontaneously,
especially for immoral behaviors that violate a
positive trait expectancy. Thus, negative violations of moral trait inferences prompted an
inference of evaluative inconsistency as indexed
by P300, while positive inconsistencies with immoral trait inferences generated little brain activity. This is consistent with earlier research on the
asymmetry between negative and positive moral
behaviors. Although ERP differences were found
for competence-related behaviors suggesting that
these EEG-measures were also sensitive at indexing descriptive inconsistencies, these results
should be taken with some caution because they
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
EEG AND SPONTANEOUS TRAIT INFERENCES
are exploratory and did not survive a Bonferroni
correction. All the ERP deviations start at about
the same latency in all channels and conditions.
Memory measures
The responses on the cued recall and sentence
completion tasks were scored on the basis of
verbatim accuracy of the sentence (without the
actor’s name) although synonyms were allowed.
As a preliminary check on whether the read vs.
consistency instruction made any difference in the
memory outcomes, a repeated measures ANOVA
similar to the one above was conducted on the
recall and sentence completion scores. However,
because many mean recall scores were zero (i.e.,
for all participants), a complete ANOVA was
impossible due to lack of individual variation. We
could only statistically analyze the difference
between trait-cued recall for TC-sentences and
antonym recall for TI-sentences. These analyses
did not reveal any significant interaction effects
of the Instruction factor with the other factors,
p.60, so that we again collapsed the data across
instruction condition. The results are listed in
Table 1.
Cued recall
This memory measure was taken to verify the
hypothesis that trait inferences were made, and if
so, that these traits were linked to the actor. This
latter hypothesis requires higher trait-cued recall
for TC-sentences in comparison with antonymcued recall for TI-sentences. Because of the many
zero scores in cued recall, we tested for each
Consistency condition whether the mean recall
was statistically significant from zero, using a
single sample t-test. As can be seen in Table 1,
trait cues elicited reliable recall of TC-sentences,
while TI-sentences did not differ from zero.
173
Conversely, antonym cues elicited reliable recall
of TI-sentences, while TC-sentences did not differ
from zero. Of more importance was the comparison between trait-cued recall of TC behaviors
and antonym-cued recall of TI behaviors. In line
with our expectations, this difference was highly
significant, t(24)8.11, pB.001. This suggests
that the implied trait was not only associated
with consistent behaviors but also with the actor,
and that this latter association provided a memory advantage for trait-cued recall in comparison
with the antonym trait that was associated with
inconsistent behaviors only. No other effects were
significant.
Sentence completion
This memory measure was included to test the
hypothesis that an impression was formed about
the actor, and requires higher recall for TIsentences than for TC-sentences. None of the
mean scores was zero (see Table 1). Hence, we
conducted a conventional ANOVA with Consistency (TC, TI and CI) as within-participants
factor. In analogy with Bartholow et al. (2001,
2003), the IRR condition was not included in the
sentence completion task. There was a main
effect of Consistency, F(2, 46)39.91, pB.001.
As predicted, planned comparisons showed that
TI-sentences were remembered better then TCsentences, F(1, 23)17.22, pB.001, confirming
that an impression was built about the actor.
Sentence completion was best for CI-sentences,
and significantly more so than TC-sentences,
F(1, 23)70.75, pB.001, and TI-sentences
F(1, 23)24.19, pB.001. However, the high
memory for CI-sentences is probably due to a
methodological limitation, as in fact there was
less variation across the CI behaviors, in that
they often described the same grade obtained so
that participants could easily guess the correct
answer.
TABLE 1
Proportion correct recall and sentence completion as a function of trait consistency
Trait consistency of the behavior
Memory measure
Trait-cued recall
Antonym-cued recall
Sentence completion
Consistent
Trait-Inconsistent
Competence-Inconsistent
Irrelevant
17.3***
0.0
5.1***
0.0
3.1*
8.0***
0.8*
1.0*
14.9***
0.3
0.0
Notes: Tests indicate whether the proportion is statistically different from zero, using a single-sample t-test. *pB.05; ***pB.001.
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
174
VAN DUYNSLAEGER ET AL.
Correlations between ERP and memory
measures
As noted earlier, the processing of negative
trait-inconsistent behaviors was associated with
an increase of P300 positivity which was significant at the Pz scalp location. In order to validate
this effect as a potential indicator of inconsistency
resolution during spontaneous inferences, for
each participant, we computed a difference
score between the TC and TI conditions on the
highest positive amplitude in the 450600 and
6001000 ms intervals separately for positive and
negative violations. Likewise, we compute a
similar TC TI difference score for trait-cued
recall, antonym-cued recall and sentences completion. We then computed a Pearson correlation
between these two sets of difference scores. We
found a significant correlation between the TC TI difference scores of the P300 for negative
violations and trait recall (r.44 and .37 for the
450600 and 6001000 ms intervals respectively,
pB.05, one-sided), suggesting that the P300 is
associated with increased memory for trait impressions about the actor. This strongly supports
the idea that the P300 is indicative of a spontaneously built trait impression about the actor. We
found no significant correlation with sentence
completion. Although such correlation might
perhaps provide stronger evidence for the role
of trait violations on the P300, as noted earlier,
trait-cued recall tracks the process of trait consolidation and therefore provides stronger validity to our hypothesis that the P300 is indicative of
spontaneous trait inferences. However, the memory of antonym cues is presumably too shallow
(see Table 1) and unrelated to the actor so that
they do not reveal a reliable correlation with the
P300.
DISCUSSION
Our major question was whether ERP responses
support the occurrence of inconsistency resolution during STI. In support of our hypothesis, we
found a significant P300 increase for trait-inconsistent behaviors that violated a trait expectancy,
and this effect appeared for negative immoral
behaviors that violated a positive trait of high
morality. These results are obtained, despite the
large number of behavioral sentences, the impoverished nature of the behavior information, and
the poor memory for the information. This is very
much analogous to earlier ITI research by
Bartholow et al. (2001, 2003), that revealed very
similar P300 results for controlled trait impressions. Also as hypothesized, the P300 component
was found at the parietal scalp site, and this
location is consistent with earlier brain imaging
research showing that memory updating involves
the parietal lobe (McCarthy et al., 1997) and that
spontaneous understanding of the goals and traits
of others often involves the temporo-parietal
junction (Frith & Frith, 2001; Harris, Todorov &
Fiske, 2005). Taken together, our results indicate
that people’s neural responses to the presented
behavioral information are different depending
on whether or not the information is consistent
with a trait spontaneously inferred during earlier
encoding of behavioral information. This implies
that participants actually inferred the trait before
detecting the inconsistency, without instruction or
intention to do so. The fact that our P300 results
are analogous to earlier ITI research (Bartholow
et al., 2001, 2003) supports the conclusion that in
the present paradigm, under STI instructions,
traits were inferred about the actor.
To cross-validate our ERP measures, we also
adopted memory tasks traditionally used in research on STI and inconsistency resolution. In
the sentence completion task, we found the
predicted memory advantage for inconsistent
information, which is in line with earlier work
on person impression formation (Hastie, 1980;
Hastie & Kumar, 1979; Stangor & McMillan,
1992; Srull & Wyer, 1989) and extends this
measure to research on spontaneous inferences.
This memory advantage implies that a spontaneous impression was formed and associated with
the actor, although it leaves open whether this
impression involves a specific trait. However, the
existence of actor-trait associations was further
supported by enhanced recall for consistent
behaviors cued by the implied trait in comparison
with inconsistent behaviors cued by antonym
traits, suggesting that these actor-trait associations
aided retrieval of consistent behaviors.
Another important validation of our neural
findings is that trait-cued recall of spontaneous
inferences was correlated with an increased P300
amplitude. This is consistent with earlier research
showing that the amplitude of the P300 is
associated with later recall, especially when
elaborate rehearsal strategies are minimized
(Allen, Iacono, & Danielson, 1992; Andreassi,
2000; Fabiani et al., 1986; Fabiani & Donchin,
1995). Although the present correlations are not
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
EEG AND SPONTANEOUS TRAIT INFERENCES
very high, it should be noted that trait recall is an
incidental and thus indirect side-effect of spontaneous inferences, and therefore we should expect
these correlations to be moderate at best. Taken
together with the enhanced trait-cued recall, the
present correlation of trait recall strongly suggests
that the P300 is an index of specific trait
inferences made spontaneously. This suggests
that the present ERP results provide neurological
evidence supporting the occurrence of STI. Future research is obviously needed to corroborate
this conclusion and to explore whether the P300 is
also a valid measure of violations during other
spontaneous social processes.
This research presented a large number of
behaviors (up to 200 in the sentences, not including the introductory paragraphs) to minimize
strategic retrieval strategies, in line with recent
STI research using memory tasks (e.g., Todorov &
Uleman, 2002). Future research might also consider the use of these alternative implicit memory
measures to validate actortrait associations
for STI and that can be applied after the
presentation of all stimulus material to avoid
interrupting EEG measurements. Promising alternatives used in earlier research appear to be false
recognition (Todorov & Uleman, 2002) and relearning (Carlston & Skowronski, 1994), which
measure the link between a photo of the actor
and the implied trait. The false recognition task
(Todorov & Uleman, 2002) involves an explicit
trait recognition based on the actor photo, and
therefore also requires a large set of behaviors
(including distractors) to minimize controlled
retrieval. The relearning task (Carlston &
Skowronski, 1994) is probably the best available
measure of implicit memory, and involves the
relearning of phototrait pairs that is facilitated
by the trait spontaneously inferred earlier when
reading about the behaviors. However, it is
somewhat less practical because it requires at
least an extra 30 min to administer. This is a long
time, given that the electrode preparation and
main experimental task in the present study take
more than an hour. Moreover, in both memory
tasks, the inclusion of visual information may
complicate and confound the EEG measurement
of a person inference.
This study, although the first to report EEG
correlates of STI, obviously has its limitations.
For one thing, although we found a reliable
correlation between the ERP and trait-cued
memory, there is a possibility that the P300
indexes violations of behavior valence, rather
175
than violations of specific inferred traits. However, this is unlikely. The P300 is sensitive to
many inconsistencies and is not restricted to
changes in valence (Nieuwenhuis et al., 2005). If
the P300 varied in function of behavioral valence
only in the present study, then it would have to
occur for all inconsistencies, including positive
inconsistencies and inconsistencies to competence-related behaviors. However, this did not
happen and the P300 appeared only in that
condition where we predicted it to emerge most
likely, that is, for negative violations of a moral
trait. This points to the importance of the inferred
trait in the inconsistency resolution process.
Future studies using trait inferences and inconsistencies within the same valence might provide
more evidence on the role of valence in trait
violations. Another point is that we did not
measure EEG responses at scalp locations other
than the midline. This may have limited the
sensitivity of the ERP to positive discrepancies,
although one would still expect these to be
smaller than negative discrepancies (cf. Reeder
& Brewer, 1979; Skowronski & Carlston, 1989;
Ybarra, 2002). Similarly, these locations might
have reduced the sensitivity to the competenceinconsistent behaviors, although earlier research
identified semantic discrepancies (e.g., N400) also
at midline locations (e.g., Fabiani & Donchin,
1995). Moreover, given that we did not measure
ITI on the same material and in the same lab, it
seems necessary to replicate the present research
with the intentional trait instructions used originally by Bartholow et al. (2001, 2003). This would
allow us to compare ERP measures during
spontaneous and intentional inference processes.
This would not only broaden our insight in the
cognitive difference underlying spontaneous and
intentional processes, but would also offer a
privileged way to compare different EEG indicators of social cognitive functioning.
Although ERP data are excellent for giving us
information about the time course and order of
mental operations, they are less precise to draw
conclusions about the location of the neural
activity. Social processes are very complex and
the brain regions involved are sometimes very
hard to locate. This makes the processes underlying ITI and STI difficult to locate. Future
research could make use of the combination of
brain-imaging techniques for a better location
and ERP measures for studying the time course
of social inference. Neural measures not only
provide an unique window to the brain processes
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
176
VAN DUYNSLAEGER ET AL.
during social cognitive functioning, they may also
help to resolve questions that were hereto impossible to answer with traditional measures, such
as the interaction between spontaneous and
intentional processes (e.g., Mitchell et al., 2006;
Todorov et al., 2006). Traditional research employed different tasks for intentional measures
(e.g., trait ratings) and for spontaneous measures
(e.g., implicit memory measures). Neural techniques surpass this boundary and allow the study of
what is unique and common to spontaneous and
intentional social processes.
Manuscript received 1 December 2006
Manuscript accepted 29 November 2007
First published online 1 February 2008
REFERENCES
Allen, J., Iacono, W. G., & Danielson, K. D. (1992). The
identification of concealed memories using the
vent-related potential and implicit behavioral measures: A methodology for prediction in the face
of individual differences. Psychophysiology, 29,
504522.
Andreassi, J. L. (2000). Psychophysiology: Human
behavior and physiological response. Mahwah, NJ:
Lawrence Erlbaum.
Bargh, J. A. (1989). Conditional automaticity: Varieties
of automatic influence in social perception and
cognition. In J. S. Uleman & J. A. Bargh (Eds.),
Unintended thought (pp. 351). New York: Guilford
Press.
Bartholow, B. D., Dickter, C. L., & Sestir, M. A. (2006).
Interpersonal relations and group processes Stereotype activation and control of race bias:
Cognitive control of inhibition and its impairment
by alcohol. Journal of Personality and Social Psychology, 90, 272286.
Bartholow, B. D., Fabiani, M., Gratton, G., & Bettencourt, B. A. (2001). A psychophysiological examination of cognitive processing of and affective
responses to social expectancy violations. Psychological Science, 12, 197204.
Bartholow, B. D., Pearson, M. A., Gratton, G., &
Fabiani, M. (2003). Effects of alcohol on person
perception: A social cognitive neuroscience approach. Journal of Personality and Social Psychology, 85, 627638.
Cacioppo, J. T., Crites, S. L., Berntson, G. G., & Coles,
M. G. H. (1993). If attitudes affect how stimuli are
processed, should they not affect the event-related
brain potential? Psychological Science, 4, 108112.
Cacioppo, J. T., Crites, S. L., Gardner, W. L., &
Berntson, G. G. (1994). Bioelectrical echoes from
evaluative categorizations: I. A late positive brain
potential that varies as a function of trait negativity
and extremity. Journal of Personality and Social
Psychology, 67, 115124.
Cacioppo, J. T., Gardner, W. L., & Berntson, G. G.
(1999). The affect system has parallel and integrative processing components: Form follows function.
Journal of Personality and Social psychology, 76,
839855.
Carlston, D. E., & Skowronski, J. J. (1994). Savings in
the relearning of trait information as evidence for
spontaneous inference generation. Journal of Personality and Social Psychology, 66, 840856.
Fabiani, M., & Donchin, E. (1995). Encoding processes
and memory organization: A model of the von
Restorff effect. Journal of Experimental Psychology:
Learning, Memory and Cognition, 21, 224240.
Fabiani, M., Karis, D., & Donchin, E. (1986). P300 and
recall in an incidental memory paradigm. Psychophysiology, 23, 298308.
Frith, U., & Frith, C. (2001). The biological basis of
social interaction. Current Directions in Psychological Science, 10, 151155.
Fyock, J., & Stangor, C. (1994). The role of memory
biases in stereotype maintenance. British Journal of
Social Psychology, 33, 331343.
Harris, L. T., Todorov, A., & Fiske, S. T. (2005).
Attributions on the brain: Neuro-imaging dispositional inferences, beyond theory of mind. NeuroImage, 28, 763769.
Hastie, R. (1980). Memory for behavioral information
that confirms or contradicts a personality impression. In R. Hastie, T. Ostrom, E. Ebbesen, R. Wyer,
D. Hamilton, & D. Carlston (Eds.), Person memory:
The cognitive basis of social perception. Hillsdale,
NJ: Lawrence Erlbaum.
Hastie, R., & Kumar, P. (1979). Person memory:
Personality traits as organizing principles in memory
for behaviors. Journal of Personality and Social
Psychology, 37, 2538.
Joyce, C. A., Gorodnitsky, I. F., & Kutas, M. (2004).
Automatic removal of eye movement and blink
artifacts from EEG data using blind component
separation. Psychophysiology, 41, 313325.
Lupfer, M. B., Weeks, M., & Dupuis, S. (2000). How
pervasive is the negativity bias in judgments based
on character appraisal? Personality and Social
Psychology Bulletin, 26, 13531360.
McCarthy, G., Luby, M., Gore, J., & Goldman-Rakic, P.
(1997). Infrequent events transiently activate human
prefrontal and parietal cortex as measured by
functional MRI. Journal of Neurophysiology, 77,
16301634.
McKoon, G., & Ratcliff, R. (1986). Inferences about
predictable events. Journal of Experimental Psychology: Learning, Memory, and Cognition, 12, 8291.
Mitchell, J. P., Banaji, M. R., & Macrae, C. N. (2005).
The link between social cognition and self-referential thought in the medial prefrontal cortex. Journal
of Cognitive Neuroscience, 17, 13061315.
Mitchell, J. P., Cloutier, J., Banaji, M. R., & Macrae, C.
N. (2006). Medial prefrontal dissociations during
processing of trait diagnostic and nondiagnostic
person information. SCAN, 1, 4955.
Nieuwenhuis, S., Ashton-Jones, G., & Cohen, J. D.
(2005). Decision making, the P3, and the locus
coeruleusmorepinephrine system. Psychological
Bulletin, 131, 510531.
Downloaded By: [Overwalle, Frank Van] At: 11:55 19 May 2008
EEG AND SPONTANEOUS TRAIT INFERENCES
Osterhout, L., Bersick, M., & McLaughlin, J. (1997).
Brain potentials reflect violations of gender stereotypes. Memory & Cognition, 25, 273285.
Reeder, G. D. (1997). Dispositional inferences of
ability: Content and process. Journal of Experimental Social Psychology, 33, 171189.
Reeder, G. D., & Brewer, M. B. (1979). A schematic
model of dispositional attribution in interpersonal
perception. Psychological Review, 86, 6179.
Reeder, G. D., & Fulks, J. L. (1980). When actions
speak louder than words: Implicational schemata
and the attribution of ability. Journal of Experimental Social Psychology, 16, 3346.
Reeder, G. D., & Spores, J. M. (1983). The attribution
of morality. Journal of Personality and Social Psychology, 44, 736745.
Skowronski, J. J., & Carlston, D. E. (1987). Social
judgment and social memory: The role of cue
diagnosticity in negativity, positivity and extremity
biases. Journal of Personality and Social Psychology,
52, 689699.
Skowronski, J. J., & Carlston, D. E. (1989). Negativity
and extremity biases in impression formation: A
review of explanations. Psychological Bulletin, 105,
131142.
Skowronski, J. J., Carlston, D. E., Mae, L., & Crawford,
M. T. (1998). Spontaneous trait transference: Communicators take on the qualities they describe in
others. Journal of Personality and Social Psychology,
74, 837848.
Srull, T. K., & Wyer, R. S. (1989). Person memory and
judgment. Psychological Review, 96, 5883.
Stangor, C., & McMillan, D. (1992). Memory for
expectancy-congruent and expectancy-incongruent
information: A review of the social and social
developmental literatures. Psychological Bulletin,
111, 4261.
Stern, R. M., Ray, W. R., & Quigley, K. S. (2001).
Psychophysiological recording (2nd ed.). New York:
Oxford University Press.
177
Todorov, A., Gobbini, I. M., Evans, K. K., & Haxby, J.
V. (2006). Spontaneous retrieval of affective person
knowledge in face perception. Neuropsychologia, in
press.
Todorov, A., & Uleman, J. S. (2002). Spontaneous trait
inferences are bound to actors’ faces: Evidence from
a false recognition paradigm. Journal of Personality
and Social Psychology, 83, 10511064.
Uleman, J. S. (1999). Spontaneous versus intentional
inferences in impression formation. In S. Chaiken &
Y. Trope (Eds.), Dual-process theories in social
psychology (pp. 141160). New York: Guilford
Press.
Uleman, J. S., Blader, S. L., & Todorov, A. (2005).
Implicit impressions. In R. R. Hassin, J. S. Uleman,
& J. A. Bargh (Eds.), The new unconscious (pp. 362
392). New York: Oxford University Press.
Uleman, J. S., Newman, L. S., & Moskowitz, G. B.
(1996). People as flexible interpreters: Evidence and
issues from spontaneous trait inference. In M. P.
Zanna (Ed.), Advances in experimental social psychology (Vol. 28, pp. 211279). San Diego, CA:
Academic Press.
Van Overwalle, F., Drenth, T., & Marsman, G. (1999).
Spontaneous trait inferences: Are they linked to the
actor or to the action? Personality and Social
Psychology Bulletin, 25, 450462.
Wigboldus, D. H. J., Sherman, J. W., Franzese, H. L., &
van Knippenberg, A. (2004). Capacity and comprehension: Spontaneous stereotyping under cognitive
load. Social Cognition, 22, 292309.
Winter, L., & Uleman, J. S. (1984). When are social
judgments made? Evidence for the spontaneousness
of trait inferences. Journal of Personality and Social
Psychology, 47, 237252.
Ybarra, O. (2002). Naı̈ve causal understanding of
valenced behaviors and its implications for social
information processing. Psychological Review, 128,
421441.