Journal of Personality and Social Psychology Copyright 1996by the American PsychologicalAssociation. Inc.
1996, Vol.71, No. 2, 230-244 0022-3514/96/$3.00
Automaticity of Social Behavior: Direct Effects of Trait Construct
and Stereotype Activation on Action
John A. Bargh, Mark Chen, and Lara Burrows
New York University
Previousresearch has shownthat trait conceptsand stereotypesbecomeactiveautomatically in the
presence of relevantbehavior or stereotyped-groupfeatures. Through the use of the same priming
procedures as in previous impression formation research, Experiment l showedthat participants
whose concept of rudeness was primed interrupted the experimenter more quickly and frequently
than did participants primed with polite-relatedstimuli. In Experiment 2, participants forwhom an
elderlystereotype was primed walkedmore slowlydownthe hallwaywhen leavingthe experiment
than did control participants, consistent with the content of that stereotype. In Experiment 3, participants
for whom the African American stereotype was primed subliminally reacted with more
hostilityto a vexatiousrequestofthe experimenter.Implicationsofthis automatic behaviorpriming
effectforself-fulfillingpropheciesare discussed,as iswhethersocialbehavioris necessarilymediated
by consciouschoiceprocesses.
For many years, social psychologists have studied the effects
of priming on the individual's subsequent impressions of others.
Priming refers to the incidental activation of knowledge
structures, such as trait concepts and stereotypes, by the current
situational context. Many studies have shown that the recent
use of a trait construct or stereotype, even in an earlier or unrelated
situation, carries over for a time to exert an unintended,
passive influence on the interpretation of behavior (see Bargh,
1994; Higgins, 1989; Wyer & Srull, 1989, for reviews).
We argue here that such passive, automatic effects of priming
need not be limited to social perception. Recent research has
shown that attitudes and other affective reactions can be triggered
automatically by the mere presence of relevant objects
and events, so that evaluation and emotion join perception in
the realm of direct, unmediated psychological effects of the environment
(see Bargh, 1994, in press, for reviews). But assuming
that behavioral responses to situations are also represented
mentally, as are stereotypes and attitudes, they should also be
capable of becoming automatically activated, by the same principles
that govern the development of automaticity of other
representations.
John A. Bargh,Mark Chen, and Lara Burrows, Department of Psy-
chology,NewYorkUniversity.
This researchwassupported inpart byGrant SBR-9409448from the
National ScienceFoundation to John A. Barghand by a Max Planck
Society (Germany) Research Prize to John A. Bargh and Peter Gollwitzer.
The contributions of Elliott Rosenthal, Gayatri Taneja, and
Jayme Yocum to this research in their capacity as experimenters is
gratefullyacknowledged.We alsothank Peter Gollwitzer,John Skowronski,
Dan Wegner,and especiallyLeonard Berkowitzand Dan Gilbert
fortheir comments and suggestionsconcerningan earlierversionofthis
article.
Correspondenceconcerningthis article should be addressedto John
A. Bargh,Department of Psychology,New YorkUniversity,6 Washington
Place, Seventh Floor,New York,New York 10003.Electronic mail
may be sent viathe Internet to bargh@psych.nyu.edu.
Automaticity in Attitudes and Social Cognition
The extent to which one's own thought and behavior are or
are not under one's own intentional control is a fundamental
existential question (see Posner & Snyder, 1975; Uleman &
Bargh, 1989). Indeed, over the past two decades, researchers in
the area of attitudes and social cognition have documented that
many of the phenomena they study are unintentional or automatic
in nature (for reviews, see Bargh, 1994; Smith, 1994, in
press; Wegner & Bargh, in press). Attitudes are discovered to
become activated automatically on the mere presence of the attitude
object, without conscious intention or awareness (i.e.,
preconsciously; see Bargh, 1989), to then exert their influence
on thought and behavior (Bargh, Chaiken, Govender, & Pratto,
1992; Bargh, Chaiken, Raymond, & Hymes, 1996; Fazio, Sanbonmatsu,
Powell, & Kardes, 1986). The self-concept (Bargh,
1982; Bargh & Tota, 1988; Higgins, 1987; Strauman & Higgins,
1987 ) is shown to become active automatically on the presence
of self-relevant stimuli to affect self-perception and emotions.
Stereotypes become active automatically on the mere presence
of physical features associated with the stereotyped group
(Brewer, 1988; Devine, 1989; Perdue & Gurtman, 1990; Pratto
& Bargh, 1991 ), and categorizing behavior in terms of personality
traits (e.g., Carlston & Skowronski, 1994; Winter & Uleman,
1984) and then making dispositional attributions about
the actor's personality (e.g., Gilbert, 1989; Gilbert, Pelham, &
Krull, 1988) have both been shown to occur automatically to
some extent.
This growing evidence ofautomaticity in social psychological
phenomena notwithstanding, it remains widely assumed that
behavioral responses to the social environment are under conscious
control (see review in Bargh, 1989). These responses
might well be consciously chosen on the basis of automatically
produced perceptions and feelings(especially when the individual
was not aware of the potential for any such nonconscious
influence;see Herr, 1986, and Neuberg, 1988), but the ultimate
behavioral decisions themselves are believed to be made con-
230
AUTOMATIC SOCIAL BEHAVIOR 231
sciously. Devine (1989), for example, argued for a two-stage
model of prejudice in which the perceptual phase is automatic
(i.e., activation of stereotypes by the target person's features),
whereas the second phase of prejudiced behavior is a matter of
conscious choice, driven by one's relevant values. Fiske (1989)
argued that a person could "make the hard choice" and overcome
stereotypic influences on behavior if sufficiently motivated
to do so.
Indeed, the traditional rationale for the study of attitudes and
social cognition is the belief (even faith) that choices of social
behavior are based on the outcome of these processes--thus, it
is social behavior that is the long-term focus of this research
(i.e., "thinking is for doing"; Fiske, 1992). The historic purpose
of attitude research has been that attitudes predicted behavior,
and evidence to the contrary (e.g., LaPiere, 1934; Wicker, 1969)
was cause for alarm, similar to the purpose of personality research
(e.g., Mischel, 1968).
The impetus behind automaticity research is no different.
Research into the automaticity of attitudes was first conducted
because it was hypothesized that attitudes that became active
automatically (preconsciously) in the presence of the attitude
object would be more likely to influence behavior toward the
object than those that depended on intentional conscious retrieval
of the attitude (Fazio et al,, 1986). The presumption behind
studies of automatic influences in social perception, such
as via primed or chronically accessible trait constructs (e.g.,
Bargh & Pietromonaco, 1982; Bargh & Thein, 1985), was that
such preconscious influences would play a stronger than usual
role in subsequent behavior toward the target person, as the perceiver
would not be aware of the interpretive bias and so could
not correct for it (Bargh, 1989). In a similar fashion, research
into the automaticity of stereotyping has been motivated by a
larger concern with the controllability of prejudicial behavior
(Devine, 1989; Fiske, 1989).
Focusing the research spotlight on attitudes and perceptions
as mediators of behavior, in the present view, has obscured the
possibility that behavior need not always be so mediated. Although
it is quite reasonable to assume that attitudes and social
perceptual processes exist in the service ofguiding behavior, this
does not require the assumption that behavioral responses always
require such services.
The Case for Automatic Social Behavior
We propose that social behavior is often triggered automatically
on the mere presence of relevant situational features; this
behavior is unmediated by conscious perceptual or judgmental
processes. We turn next to a discussion of several lines of support,
both theoretical and empirical,, for this hypothesis.
Behavioral Responses Can Be Associated With
Situational Features
Social-behavioral responses are represented mentally just as
are trait concepts and attitudes. Thus, they should be capable
of becoming activated automatically on the mere presence of
relevant features in the environment by the same principles that
produce automatic trait categorization and automatic attitude
activation.
Several theorists have argued that behavioral responses are
activated immediately by the situational context. Lewin's
(1943) notion of the psychological situation considered it to
consist of the totality of the individual's immediate reactions to
the objective, external situation. Mischel (1973) further developed
this concept as part of his social-cognitive model of personality.
He noted that an individual can have all sorts of immediate
reactions to a person or event, not limited to cognitive
or perceptual ones but including (a) expectancies for what was
going to happen next in the situation; (b) subjective evaluations
of what was happening; (c) emotional reactions one has had in
that situation in the past; and, most important to the present
thesis, (d) the behavioral response patterns one has available
within the situation based on one's past experience (see also
Higgins, 1987).
There is no theoretical or conceptual reason why the effects
of preconscious, automatic activation should be limited to perception
and evaluation. Preconscious activation of mental representations
develops from their frequent and consistent activation
in the presence of a given stimulus event in the environment
(Bargh, 1989; Shiffrin & Schneider, 1977). This is the
mechanism behind the automaticity of trait construct activation
given the mere observation of trait-relevant behavior, of attitudes
in the mere presence of the attitude object, and of stereotypes
on the mere presence of a stereotyped group member.
To the extent that an individual repeatedly has the same reaction
to a social stimulus event, the representation of that response
should come eventually to be activated automatically on
the mere occurrence of that event. Thus, if an individual consistently
behaves the same way in response to a situation, that
behavioral response should become automatically associated
with those situational features. In harmony with this hypothesis,
Mischel and Shoda ( 1995; Shoda, Mischel, & Wright, 1994)
have provided several demonstrations of a high degree of consistency
over time in an individual's behavioral responses to the
same situations, when situations are defined in terms of specific,
concrete sets of features.
The Principle of ldeomotor Action
William James held that the mere act of thinking about a
behavior increased the tendency to engage in that behavior; he
called this the principle ofideomotor action: "We may lay it
down for certain that every representation of a movement awakens
in some degree the actual movement which is its object"
( 1890, p. 526). James's notion of awakening here is similar to
modern notions of accessibility, in that the internal (through
ideation ) activation of a representation (i.e., through imagining
the behavior) increases its accessibility of likelihood of activation.
Modern research on construct accessibility has shown that
mental representations can become activated from many
sources, including one's goals, external environmental events,
long-term use, and recent thought. Moreover, all of these possible
sources increase the accessibility or ease of use of that representation
in an interchangeable, additive fashion (Bargh,
Bond, Lombardi, & Tota, 1986; Higgins & King, 1981 ). For
James, imagining or thinking about a behavioral response had
the same kind of priming effect on the likelihood of engaging in
that response.
232 BARGH, CHEN, AND BURROWS
However, Lashley (1951), in a famous discourse on the sequential
organization of behavior, was the first to use the term
priming to describe the preparatory function of thought.
Thinking has the function of preparing the body for action, and
Lashley's given example of this was the production of fluent
speech. To be able to speak words in an understandable, serial
fashion, just as to act in a sequential manner, requires a prior
organization of the representations that are to be used, in the
intended sequence. The function of this preparation allows for
the fluidity of spoken thoughts and ideas and the enactment of
organized movements in the proper order. Thus, like James,
Lashley argued for the necessity of a direct connection between
thought and behavioral representations (N.B., not limited to
those used in speech).
The automaticity of the ideomotor-action effect--that merely
thinking about a behavior makes it more likely to occur, even if
it is unintended--has been demonstrated recently in a series of
studies by Wegner and his colleagues (see Wegner, 1994, for a
review). Wegner's ironicprocess model contends that acts of
intentional control over our thought and behavior involve an
automatic monitoring of the presence of the unwanted state.
When this automatic vigilance notices the to-be-controlled
thought or response tendency, conscious processing can inhibit
it from occurring. In this way, experimental participants can
distract themselves from thinking about white bears (for
example) by consciously thinking about something else. But
this control over unwanted thoughts can occur only when there
is sufficient attentional capacity available for the act of control.
If the person is distracted, or is under attentional load, an ironic
effect is likely to occur: The very thought or behavior one did
not want to happen, does happen.
The irony of this effect is that the likelihood of this occurrence
(under attentional load) is actually greater than if the person
had not tried to stop that response. Thus, in one experiment,
participants under attentional load who are trying not to
make sexist completions to word fragments actually make more
than participants not instructed to try to avoid sexism. According
to the ironic process model, this occurs because the representation
of the unwanted response is more accessible than
usual because the person is watching out for its occurrence and
has to keep it in mind to do so. For present purposes, the importance
of these findings is that the mere act of thinking about
a response, even when the thought involved is meant to help
prevent that response, has the automatic eflbct of increasing the
likelihood of that response. The principle of ideomotor action,
to put it another way, operates in the absence of the person's
intention to engage in that behavior and even when the person
is trying to avoid that behavior.
Ansfield and Wegner (1996) applied the ironic process model
to understanding the classic literature on automatisms, or behaviors
that do not appear to be consciously produced, like the
spinning table in seances, a divining rod, or the movement of
the pendulum in Chevreuii's illusion. In the latter case, the pendulum
held dangling above a table moves--apparently of its
own accord--when the person is told to hold it completely still.
In fact, the pendulum tends to move along the very axis along
which the person is trying to prevent it from moving. Ansfield
and Wegner showed in several experiments that this effect is
produced by the very attempt to prevent the seance table or the
pendulum from moving; participants are of course not aware of
the automatic effect that the thoughts about the to-be-avoided
movement have on their behavior and so cannot control it.
The Perception-Behavior Link
Just as the accessibility or likelihood of use of a concept increases
no matter what the particular the source of that accessibility,the
likelihood of a behavioral response may increase from
thinking about that behavior, regardless of the source of that
thought. Specifically, cognitions about a type of behavior can
come not only from internal sources, as in the above examples,
but also from external sources, such as perceiving that type of
behavior enacted by others.
There is a strong historical precedent for postulating an automatic
link between the representations used to perceive behavior
and those used to engage in that behavior oneself. Imitation,
for example, consists of performing an action that corresponds
in its structure to the perceived action of another person. The
capacity to imitate is present in early childhood (Piaget, 1946)
and even in newborns (Meltzoff & Moore, 1977, 1983). Such
scholars of imitative behavior as Koffka (1935), Piaget (1946),
and Bandura (1977) all have proposed that imitation is made
possible by a common or shared representational system for
perceptual and action codes (see Prinz, 1990, for a review).
Schank and Abelson's (1977) script theory argues that the same
mental structures used to understand and anticipate the sequence
of behavior in social situations also is used to generate
appropriate responses to them. Theories of speech production
have increasingly emphasized the mediational role played by
the representations involved in speech perception (see Dell,
1986; Meyer & Gordon, 1984; Prinz, 1990). Also, Zajonc, Pietromonaco,
and Bargh (1982) showed that people implicitly
mimic the facial expressions of others, such that when this subtle
imitation is prevented (i.e., by having them chew gum while
the faces are presented), memory for the faces is impaired.
In 1984, Berkowitz reformulated his theory of how violence
portrayed in the mass media increased the probability of aggression
in the viewer by invoking James's principle of ideomotor
action. Activation was said to spread in memory from representations
of the violent acts perceived in the media to other
aggressive ideas of the viewer, and this spreading activation occurred
"automatically and without much thinking" (p. 410).
Similar to Mischel's (1973) analysis, Berkowitz (1984) argued
that behavioral responses as well as thoughts and emotions
could all be activated automatically by aggressive stimuli: "The
present conception does not stop with the individual's thoughts
and memories...It holds that externally presented ideas can
activate particular feelings and even specific action tendencies
as well" (p. 410).
An experiment by Carver, Ganellen, Froming, and Chambers
(1983) provided evidence in line with Berkowitz's (1984) ideomotor
action model of the effect of aggressive cues on aggression.
In a first experiment, some participants' concept of hostility
was primed subliminally, following the procedure of Bargh
and Pietromonaco (1982). Then, in what they believed to be an
unrelated second experiment, participants were instructed to
give shocks to another "learner" participant (actually a
confederate) whenever he or she gave an incorrect answer. Com-
AUTOMATIC SOCIAL BEHAVIOR 233
pared to participants who were exposed to neutral priming
stimuli, those presented subliminally with hostility-related
primes gave longer shocks. Carver et al. (1983) accounted for
their results in terms of a behavioral schema for hostility and its
close semantic associativeties to the "interpretive schema" used
to perceive hostility. Because of the degree of semantic feature
overlap between the two representations, the authors argued,
activation will spread automatically from the interpretive to the
behavioral schema, so that perceiving another person's hostilty
increases the likelihood that one will behave in a hostile manner
oneself.
The behavioral schema notion, which is a variant of the ideomotor
action hypothesis, has the desirable ability to account
for how the same priming manipulation can produce effects on
impression formation in one study (Bargh & Pietromonaco,
1982) and behavior in another(Carver et al., 1983, Experiment
2). Because the only difference between the two studies was the
particular dependent measure collected following the priming
manipulation, the inescapable conclusion is that the activation
of the concept of hostility had the simultaneous effects of making
the participant both more likely to perceive hostility in another
person and to behave in a hostile manner him- or herself.
To us, Carver et al.'s ( 1983 ) results are an intriguing clue that
the influence of perception on behavioral tendencies is automatic,
in that it is passive, unintentional, and nonconscious.
Therefore, recent evidence of automatic influences in" social
perception, such as the automatic activation of stereotypes and
priming effects on impression formation (see Bargh, 1994, for
a review), when related to the foregoing discussion, implies that
there may be behavioral consequences of automatic social perception
for the perceiver. For it is precisely when the individual
is not aware of a perceptual process that conscious control over
it is not possible (Bargh, 1989; Strack & Hannover, 1996 ), maximizing
the possibilityof the passive perception-behavior effect.
The Present Experiments
From the various streams of evidence reviewed above, several
principles can be derived concerning the conditions under
which automatic social behavior will be produced. First, behavioral
representations exist and can become activated. They can
become active and accessible when one thinks about that kind
of behavior, either actively or passively. The tendency to behave
in line with the representation is increased when it is activated,
whether the reason for that activation is (a) an intention to prepare
to engage in that behavior (e.g., Lashley), (b) an intention
not to engage in that behavior (e.g., Wegner), (c) merely thinking
about that behavior without an intention to engage in it or
not (e.g., James), or (d ) merely perceiving that kind of behavior
in another person (e.g., Berkowitz).
The present hypothesis is that social behavior should be capable
of automatic activation by the mere presence of features
of the current environment just as are social perceptions and
attitudes. By the mere presence of environmental features, we
mean that the activation of the behavioral tendency and response
must be shown to be preconscious; that is, not dependent
on the person's current conscious intentions (see Bargh, 1989,
in press). By these criteria, none of the research reviewed above
has demonstrated direct, automatic behavioral effects. The
ironic process research has indeed shown automatic behavior in
that it is unintended by the individual and even uncontrollable
when attention is in short supply. These effects are goal dependent
in that they are produced by an act of conscious intention
(see Wegner, 1994, for a similar but more elaborate analysis)
and would not occur without that intention in place.
Moreover, in all of the studies reviewed by Berkowitz (1984)
that were in favor of the perception-behavior link, including
Carver et al.'s ( 1983 ) experiments, participants were given the
explicit, conscious goal to engage in the behavior that was
shown to be affected by the priming manipulation. In Berkowitz
and LePage's (1967) and Carver et al.'s (1983) studies, for example,
participants were instructed to take the role of teacher
and give shocks to a learner. These studies showed that the intentional
behavior could be affected in intensity or duration by
the aggression priming manipulation (the presence of guns or
prior exposure to synonyms of aggression), but they did not
show the behavior to be produced automatically, in the absence
of that explicitlygiven intention.
Thus, although all of these lines of evidence are suggestive
and supportive of the hypothesis of automatic social behavior,
they have not demonstrated it. The three experiments we report
in this article were designed to provide a definitive test of this
hypothesis. In Experiment 1, participants were primed on the
traits ofeitber rudeness or politeness (or neither) with Srull and
Wyer's (1979) scrambled-sentence test. In Experiment 2, we
again used the scrambled-sentence priming manipulation but
used it to activate the participants' stereotype of elderly people.
In neither experiment were participants given any explicit conscious
instructions to act in line with any of the trait dimensions
being primed or measured. In fact, in both experiments the key
dependent behavioral measures were taken at times when participants
believed they were not currently engaged in an experimental
task at all (i.e., in the hallway between parts of the experiment,
or after they thought the experiment was over). In
Experiment 3, a different priming manipulation--photographs
of male African American faces--was used, and it was presented
subliminally. This change was intended to extend the
generality of the present findings to more realistic environmental
stimuli and to effectivelyrule out any possible demand interpretations
of the first two experiments.
Experiment 1: Behavioral Consequences of Trait
Construct Priming
Method
Participants. A total of 34 students at New York University who
wereenrolledin the Introductory Psychologycourse participatedin the
experiment in partial fulfillment of a course research requirement. On
their arrival at the laboratory waiting room they were randomly assigned
to one ofthe three priming conditions.
Materials. The priming manipulation took the form of a "Scrambled
Sentence Test" (Srull & Wyer, 1979), presented to participantsas
a test of language ability. For each of 30 items, participants are to use
the fivewords listed to construct a grammaticallycorrect four-wordsentenceas
quickly as possible. The fivewords presented in a giventest item
are in scrambled order, such as "he it hides finds instantly."
Three versions ofthe scrambled-sentence test were constructed: One
was intended to prime the construct rude,another the constructpolite,
234 BARGH, CHEN, AND BURROWS
and a third was intended to prime neither trait (the neutral priming
condition). For both the rude and the polite priming versions, 15 of the
30 items contained an adjective or verb semantically related to the trait
in question. For the rude priming version, the critical priming stimuli
were aggressively, bold, rude. bother, disturb, intrude, annoyingly, interrupt,
audaciously, brazen, impolitely, infringe, obnoxious, aggravating,
and bluntly (e.g., "they her bother see usually"). For the polite
priming version, the 15 critical stimuli were respect, honor,considerate,
appreciate, patiently, cordially, yield, polite, cautiously, courteous, graciously,
sensitively;discreetly, behaved, and unobtrusively (e.g., "they her
respect see usually"), in the neutral priming version, these t5 words
were replaced by exercising, flawlessly; occasionally rapidly, gleefully,
practiced, optimistically, successfully, normally, send, watches, encourages.
gives, clears, and prepares (e.g., "'they her send see usually").
Procedure. Participants took part in the experiment one at a time.
When they arrived at the central waiting area, they were greeted by the
experimenter, who showed them into the first experimental room. They
were told that the experiment was actually composed of two short studies
having to do generally with language ability. After obtaining their
consent to participate, the first study, ostensibly to do with construction
of grammatical English sentences, was explained to them. The experimenter
handed the participant an envelope that contained one of the
three forms of the scrambled-sentence test. Which of the three versions
(rude, polite, or neutral priming) the participant received had been randomly
determined by another experimenter who did not have any contact
with the participant, with the constraint that approximately equal
numbers of participants received each of the forms. Neither the experimenter
nor the confederate (see below) knew the priming condition to
which a particular participant had been assigned until after the experimental
session was over.
The participant was given the general instructions for the scrambledsentence
test. The experimenter told the participant that when the test
had been completed (most participants took about 5 rain to complete
the 30-item test), he or she should come out into the hallway to find the
experimenter, who would be at a different room around the corner. The
experimenter said that at that time he would give the participant the
next short experimental task to complete.
The experimenter then waited for the participant in the doorway of
another laboratory room, located on the same hallway but around a
corner from the first room. A confederate of the experimenter was
seated in this second room, in sight of the experimenter but not of anyone
coming from the direction of the first room, as she was seated behind
a half-opened door. The confederate posed as another participant
who was apparently having difficulty understanding how to complete a
task. The experimenter and confederate engaged in a conversation, with
the experimenter standing so that his body was open to the direction
from which the participant would be coming down the hall, at about a
45° angle, while still facing the confederate in the room.
When the participant turned the hallway corner and became visible
to the experimenter, without looking at the participant or acknowledging
his or her presence the experimenter made a subtle prearranged sign
to the confederate (i.e., touching his right pant leg). At this the confederate
started a stopwatch. The experimenter and confederate continued
their conversation as the participant approached and while he or she
stood near the experimenter, waiting for the experimenter to acknowledge
his or her presence and give him or her the next experimental task
to complete.
Our dependent measure was the amount of time the participant
would wait until interrupting the conversation between experimenter
and confederate and ask to be given the next experimental task. Until
he or she did so, the conversation continued, with the confederate asking
questions and requesting clarification and just not getting it, and the
experimenter repeating the instructions and clarifying if possible and
also not turning to look at the participant or make eye contact with him
or her. Our hypothesis was that participants in the rude prime condition
would interrupt more quickly than neutral prime condition participants,
and those in the polite prime condition would wait longer to interrupt
than would neutral condition participants.
When the participant began to say anything to the experimenter, such
as "Excuse me," or "Sorry, but...'" the confederate stopped the stopwatch
and recorded the elapsed time. We placed a 10-min limit on how
long the participant would have to wait if he or she did not interrupt at
all, reasoning that by that time he or she was not going to. When the
participant interrupted, or at the end of the 10-min maximum waiting
time, the experimenter showed him or her into a room adjacent to the
one in which the confederate was seated and gave the participant a brief
anagram puzzle task that took no more than 2 min to complete. After
the participant had completed this task, he or she was partially debriefed
and questioned concerning how they thought the first, scrambled-sentence
test might have influenced them in the rest of the experiment.
No participant showed any awareness or suspicion as to the scrambledsentence
test's possible influence on their interruption behavior; nearly
all participants reported either no effect of the first task or that both it
and the anagram task were related to language ability, which was the
cover story.
At this point the experimenter thanked the participant for helping
with the study, and the participant headed toward the elevator to leave
the building. Waiting there was a second confederate, who posed as a
department representative inquiring as to the participant's experience
in psychology experiments. The confederate asked if the participant
would mind completing a "'Survey of Experimental Participants," and
all hut 3 participants agreed to do so. On the survey were six questions
concerning whether the experiment was interesting and whether it was
considered a valid educational experience. The last three items concerned
the experimenter, whether he or she was on time, whether he or
she explained the study and answered questions, and the critical item:
"Was the experimenter courteous and polite to you?" This the participant
responded to on a -3 to +3 scale that ranged from -3 (not at all)
to +3 (very much so). This item served as our check for a potential
alternative interpretation of our results, to be discussed below.
When the participant had completed the survey, he or she was fully
debriefed as to the purpose and hypothesis of the experiment and was
thanked again by the second confederate.
Results
Our primary dependent variable was the number of seconds
the participants waited before interrupting the experimenter.
A one-way analysis of variance (ANOVA) of these data, with
priming condition as the single factor, revealed a significant
main effect, F(2, 33) = 5.76, p = .008. Participants in the rude
priming condition interrupted significantly faster (M = 326 s)
than did participants in the neutral (M = 5 t9 s) or polite (M =
558 s) priming conditions. Within the significant main effect,
simple t tests revealed that the rude prime condition mean was
significantly shorter than each of the other two means (both ps
< .04), which were not reliably different from one another
(t< 1).
Although this result supports our hypothesis that social interaction
behavior can be primed, the time-to-interruption distribution
varied considerably from normality. Fully 21 of the 34
participants did not interrupt at all in the 10 rain available to
them, so that the time variable suffered from a severe ceiling
effect. Thus, we reanalyzed the data in terms of the percentage
of participants in each priming condition who interrupted at all
during the 10-rain period. These percentages are shown in Figure
1.
AUTOMATIC SOCIAL BEHAVIOR 235
Figure 1. Percentage of participants who interrupted the experimenter
within the 10-rain period, by trait priming condition
(Experiment 1).
Because we made a specific directional prediction regarding
the ordering of the three percentages, rather than an omnibus
prediction of any difference among the three, we applied the test
for a linear trend in proportions (Snedecor & Cochran, 1980,
pp. 206-208). The specific hypothesis tested was that the proportion
of participants who interrupted the experimenter
would increase as a function of the trait concept primed, from
politeness through no priming to rudeness. In support of the
hypothesis, the test revealed a reliable linear trend (Z = 2.32, p
---.02, two-tailed). Participants whose concept of rudeness had
been surreptitiously activated in the scrambled-sentence test
subsequently were more likely to interrupt the conversation between
the experimenter and confederate than were the other
participants, and those whose concept of politeness had been
activated were the least likely to interrupt.
Whereas our interpretation of this result is that the priming
manipulation produced a direct, preconscious effect on the participants'
subsequent behavior in a social interaction, an alternative
explanation in terms of the social perceptual effects of
priming must be considered. This alternative centers on how the
participant may have interpreted or perceived the experimenter's
behavior during his conversation with the confederate. If
the experimenter's behavior was ambiguous enough to permit
different impressions of him, depending on the relative accessibility
of the participant's relevant trait constructs (e.g., Higgins
& King, 1981 ), it is possible that rude-primed participants were
more likelyto perceive the experimenter as rude because he was
ignoring them and attending to the confederate, whereas politeprimed
participants were more likely to perceive the experimenter
as polite because he so patiently dealt with the confederate's
questions. The participants' own behavior to the experimenter
(i.e., whether they interrupted him) might have been
based on these differential perceptions (as has been demonstrated
in studies by Neuberg, 1988, and Herr, 1986) instead of
as a direct effect of the priming manipulation.
It is important to note, however, that we did not design or
intend the experimenter's behavior to be ambiguous in this way,
because our intention was to demonstrate an effect of priming
on behavior that was not mediated by differential interpretations
of the experimenter's behavior. In previous category
accessibilityresearch, some studies have sought to examine features
of accessibility other than its effect on interpretation.
Bargh and Thein ( 1985 ), for instance, studied the ability of accessible
constructs to pick up relevant behavioral information
under attentional overload. In this study, the construct-relevant
behaviors were unambiguous and clearly diagnostic of the trait
in question, and accessibility of the trait concept did not influence
impressions of the target (though it did affect ability to
process the information). The top-down effects of accessibility
should influence impressions only when the informational input
is sufficiently ambiguous (i.e, a relatively weak bottom-up
effect; see Higgins, 1989).
To assess whether the priming manipulations had resulted in
differential perceptions of the experimenter's politeness, we examined
the ratings participants made on the "Experimental
Participation Survey." On the critical item having to do with
the experimenter's politeness, which ranged from -3 (not at all
polite) to +3 (very polite), there was no reliable difference in
the ratings made in the three priming conditions, F(2, 28) < 1
(polite prime condition M: = 0.6, rude prime condition: M =
0.4, neutral prime condition: M = 0.8). Apparently, the fact
that the experimenter essentiallyignored participants (focusing
his attention on the confederate and her questions) while they
waited in the hall led all of them to perceive him as equivalently
nonpolite (but not impolite either).
The fact that the behavioral measure showed quite strong
effects of the priming manipulation, whereas the effect on the
judgment measure was nonexistent, argues against the alternative
interpretation of our findings. It was not the case that the
priming manipulation affected consciously made judgments
about the experimenter, which then determined behavioral responses
to him. The results instead point to a direct effect on
behavior that is not mediated by conscious perceptual or judgmental
processes. In fact, the present priming effect on behavior
is much more substantial than the typical priming effect in impression
formation research (cf. Bargh et al., 1986; Srull &
Wyer, 1979). The powerful effects of priming on behavioral relative
to perceptual dependent variables is a topic to which we
return in the General Discussion.
Discussion
In this first experimental test of whether trait construct priming
would produce differences in behavior as it has been shown
to do for perceptual interpretation, participants primed with
rudeness-related stimuli in an ostensibly unrelated first experiment
interrupted a conversation reliably faster and, as a group,
more frequently than did other participants. Those whose concept
of politeness was surreptitiously activated interrupted the
least often. Behavior in social interaction, like social perception
236 BARGH, CHEN, AND BURROWS
and evaluation, apparently can be driven directly by environmental
stimuli--that is, preconsciously and automatically.
The results of Experiment 1 showed that the passive, automatic
activation of a trait concept results in traitlike behavior
by the individual, but there are two ways in which trait concepts
can be activated directly by the environment. One is a direct
activation by the presence in the environment of trait-relevant
behavior; several lines of research show that behavior relevant
to a trait automatically activates that trait concept (Newman &
Uleman, 1989; Srull & Wyer, 1979). Another way in which a
given trait concept can be activated automatically, however, is
by its membership in a larger schema, such as a stereotype. Stereotypes
of social groups consist, in part, of constellations of
interrelated trait concepts (e.g., Brewer, 1988; Devine, 1989;
Fiske & Neuberg, 1990) that become active in an all-or-none
fashion (see Hayes-Roth, 1977) in the presence of the features
of a group member.
Devine's (1989) Experiment 2 demonstrated the all-or-none
activation feature quite clearly. She primed the stereotype of
African Americans (as held by White Americans) subliminally
using a set of stereotype-related stimuli (e.g., musical, athletic)
that did not include any item related to hostility, though hostility
had been shown by pretesting (and much prior research) to
participate in the stereotype. However, Devine (1989) showed
that the concept of hostility had indeed become activated by the
priming manipulation, because primed participants subsequently
rated a target person's ambiguously relevant behavior
as more hostile than did a control group (just as had participants
in Bargh & Pietromonaco's [1982] study who were
primed directly with hostile-related stimuli). The only way this
result could have been obtained was if the concept of hostility
had become active (and thus temporarily more accessible for
use in interpreting the target's behavior) by its participation in
the activated stereotype.
Thus, research has shown that the priming or automatic activation
of both single trait concepts (e.g., Bargh & Pietromonaco,
1982; Srull & Wyer, 1979) and stereotypes (Devine,
1989) influences social perception without the individual being
aware of or intending this influence. The present Experiment 1
has shown that priming a single trait concept influences subsequent
social behavior as well. The deduction that follows naturally
from these two sets of findings is that the priming or automatic
activation of stereotypes should make the perceiver himor
herself more likely to act in accordance with the trait concepts
that participate in that stereotype. We designed Experiments
2 and 3 to test this hypothesis in the context of two distinct
stereotypes: one for elderly people and one for African
Americans.
Experiments 2a and 2b: Behavioral Effects of Activating
the Elderly Stereotype
Method
Overview. The method and procedure for Experiments 2a and 2b
are identical and so are described together. Experiment 2b is a replication
of 2a. The present experiments were designed to study the effect of
activation of the elderly stereotype on behavior. Participants were instructed
to work on a scrambled-sentence task as part of a language
proficiency experiment. The scrambled-sentence task contained words
relevant to the elderly stereotype in the elderly priming condition, but
all references to slowness, which is a quality stereotypically associated
with elderly people, were excluded. The neutral priming condition
scrambled-sentence task contained age-non-specific words in the place
of elderly stereotyped words.
After completing the task, each participant was partially debriefed
and thanked for his or her participation. A second experimenter then
surreptitiously recorded the amount of time the participant took to
walk down the corridor after exitingthe laboratory room. Subsequently,
participants were debriefed once again with a complete explanation of
the purpose of the experiment. The main hypothesis was that participants
who had been primed with the elderly stereotype would walk
more slowly compared to participants who had not been primed with
the stereotype-relevant stimuli.
Participants. Thirty male and female New York University undergraduates
who were enrolled in Introductory Psychology participated
in Experiment 2a, and a different sample of 30 participated in Experiment
2b, to partially fulfill a course requirement. In both experiments,
participants were randomly assigned to either an elderly prime condition
or a neutral prime condition. The experimenter kept himselfblind
to condition by prepackaging the various scrambled-sentence tasks and
picking packets randomly when the participant arrived at the laboratory
waiting area.
Materials. As in Experiment 1, the priming manipulation took the
form of a scrambled-sentence task presented to participants as a test of
language proficiency. We constructed two versions of the scrambledsentence
task: one elderly prime version, which contained words related
to the elderly stereotype, and another, neutral version. For the elderly
prime version, the critical stimuli were worried, Florida, old, lonely,
grey, selfishly, careful, sentimental, wise, stubborn, courteous, bingo,
withdraw,forgetfuL retired, wrinkle, rigid, traditional bitter, obedient,
conservative, knits, dependent, ancient, helpless, gullible, cautious, and
alone. These prime words were obtained from previous research that
examined the components of the elderly stereotype (Brewer, Dull, &
Lui, 1981; Harris & Associates, 1975; McTavish, 1971; Perdue & Gurtman,
1990). In the neutral version, the elderly prime words were replaced
with the words unrelated to the elderly stereotype (e.g., thirsty,
clean, private).
Procedure. Participants took part in the experiment one at a time.
The participant was informed that the purpose of the study was to investigate
language proficiency and that he or she was to complete a
scrambled-sentence task. The task consisted of 30 sets of fiveword combinations.
The participant was instructed to write down a grammatically
correct sentence using only four of the five words given. Participants
were also informed that the task was self-paced. After giving the
instructions, the experimenter left the room so that the participant
could complete the task in privacy.
After the participant completed the task and notified the experimenter,
the experimenter re-entered the lab room and partially debriefed
the participant. He or she was informed that the experiment
was concerned with how individuals use words in various, flexibleways.
Waiting until the participant had gathered all of his or her belongings,
the experimenter told the participant that the elevatorwas down the hall
and thanked him or her for participating.
Using a hidden stopwatch, a confederate of the experimenter, who
was sitting in a chair apparently waiting to talk to a professor in a nearby
office, recorded the amount oftime in seconds that the participant spent
walking a length of the corridor starting from the doorway of the experimental
room and ending at a broad strip of silver carpet tape on the
floor 9.75 m away.
Afterward, the experimenter caught up with the participant near the
elevator and gave the complete debriefing, explaining the experimental
hypotheses verbally as well as giving the participant an accompanying
written version. Participants were also informally asked (prior to this
AUTOMATIC SOCIAL BEHAVIOR 237
Figure2. Meantime (in seconds) to walkdown the hallwayafter the
conclusion of the experiment, by stereotype priming condition, separately
for participants in Experiment 2a and 2b.
final debriefing) whether they thought the scrambled-sentence task
might have affectedthem in any way,and if they had known that the
task contained words relevantto the elderlystereotype. No participant
expressedanyknowledgeofthe relevanceofthewordsinthe scrambledsentencetask
to the elderlystereotype.Furthermore, no participant believedthat
the wordsbad an impact on hisor her behavior.
Results
Experiment 2a. A t test was computed to ascertain the effect
of the priming manipulation on walking speed. Participants in
the elderly priming condition (M = 8.28 s) had a slower walking
speed compared to participants in the neutral priming condition
(M = 7.30 s), t(28 ) = 2.86, p < .0 l, as predicted.
Experiment 2b. In the replication, analyses revealed that
participants in the elderly priming condition (M = 8.20 s) again
had a slower walkingspeed compared to participants in the neutral
priming condition (M = 7.23 s), t(28) = 2.16, p < .05.
Thus, across both studies, passivelyactivatingthe elderlystereotype
resulted in a slower walking speed (see Figure 2 ).
Discussion
The results of the present experiments suggest that exposing
individualsto a series of words linked to a particular stereotype
influences behavior nonconsciously. How the activated stereotype
influencesbehavior depends on the content of the activated
stereotype itself, not the stimulus words actually presented. Because
there were no allusions to time or speed in the stimulus
materials, the results of the study suggest that the elderly priming
stimuli activated the elderly stereotype in memory, and participants
subsequently acted in ways consistent with that activated
stereotype.
Awareness Check Study
The crucial factor in concludingthat these results show automatic
effects on behavior derives from the perceiver's lack of
awareness of the influence of the words. Previous research (see
review in Bargh, 1992) has indicated that it is not whether the
primes are presented supraliminally or subliminally, but
whether the individualis aware of the potential influenceof the
prime that is critical; diametrically opposite effects on judgments
are obtained if the participant is aware versus not aware
of a possible influence by the priming stimuli (see Lombardi,
Higgins, & Bargh, 1987; Strack & Hannover, 1996). We conducted
a subsequent study to explicitly test whether the participants
were aware of the potential influence of the scrambledsentence
task. Our conclusionsin terms of automatic social behavior
depend on the participants' not being aware of this
influence.
Method. Nineteen male and femaleundergraduate students at New
YorkUniversityparticipated in the experiment to partially fulfillcourse
credit. On arrivalat the laboratory waitingroom, participants wererandomlyassignedto
eitherthe elderlystereotypepriming condition or the
neutral priming condition.
Participants took part in the experiment one at a time. They were
informed that the purpose ofthe study wasto investigatelanguageproficiencyand
that they wouldcomplete a scrambled-sentencetask. Participants
wererandomlyadministered eitherthe versionofthe task containing
words relevant to the elderlystereotype or the neutral version
containing no stereotype-relevantwords.Immediatelyaftercompletion
ofthe task, participants wereaskedto complete a versionofthe contingency
awareness funnel debriefing, modeled after Page (1969). This
contingencyawarenessdebriefingcontained items concerningthe purpose
of the study,whether the participant had suspected that the purpose
of;the experiment was different from what the experimenter had
explained,whetherthe wordshad any relation to each other,what possiblewaysthe
wordscould have influencedtheir behavior,whether the
participants could predict the direction of an influence if the experimenter
had intended one, what the words in the scrambled-sentence
task could haverelated to (ifanything), and ifthe participant had suspectedor
had noticedanyrelationbetweenthe scrambled-sentencetask
and the concept of age. Afterward, the experimenter explained the
hypothesesto the participants and thanked them fortheir help.
Resultsanddiscussion. Inspection of the responses revealed
that only l of the 19 participants showed any awareness of a
relationship between the stimulus words and the elderly stereotype.
However, even this participant could not predict in what
form or direction their behavior might have been influenced
had such an influence occurred. Thus, it appears safe to conclude
that the effect of the elderly priming manipulation on
walking speed occurred nonconsciously.
Does Mood Mediate the Effect ofPriming
on Walking Speed?
One alternative explanation that can be offered for the effect
of the elderly-stereotype-related stimuli on walking speed is
that, in general, the words relating to the elderly stereotype are
more likely than control words to induce in participants a sad
mood, which might then be the reason they walked more slowly.
Method. A total of 33 undergraduate men and women from New
YorkUniversityparticipated in the experiment in order to partially fulfillcourse
credit. On arrival at the experimental waitingarea, partici-
238 BARGH, CHEN, AND BURROWS
pants were randomly assigned to either the elderly or the neutral priming
conditions.
Each participant completed the version of the scrambled-sentence
task corresponding to their assigned priming condition: either the prime
version containing elderly-stereotype-relevant words, or a neutral version
without the elderly-stereotype-relevant words. Immediately after
they completed the task, a modified version of the Affect-Arousal Scale
(Salovey & Birnbaum, 1989) was administered to each participant.
This questionnaire contained eight bipolar items differentiatingfeelings
of emotion and arousal on scales ranging from -8 to +8. The emotion
dimensions were bad-good, disappointed-satisfied, sad-happy, and
displeased-pleased.The arousal dimensions were calm-excited,tiredenergetic,down-elated,and
sedate-aroused.Participants responded to
each dimension in terms of how they felt at that moment. After completion
of the scales, participants were debriefed and thanked for their
cooperation.
Results. We conducted a multivariate analysis of variance
(MANOVA) using priming condition as the independent variable
and the average of the four emotion scales and the average
of the four arousal scales as the dependent measures. Analyses
revealed that the main effect of priming condition was not significant
across the two dependent measures, F( 1, 31 ) < 1. Furthermore,
there was no interaction between priming condition
and the two dependent measures of emotion and arousal, F( 1,
31 ) < 1. If anything, participants in the elderly priming condition
were in a more positive mood (M = 1.7) than control participants
(M = 0.3) and were more aroused or energetic as well
(Ms = -0.5 and -1.2, respectively). Thus, participants who
were exposed to the elderly stereotype stimuli were not more
likely to be sad or less aroused compared to participants who
were not exposed to the stereotype-related stimuli. The alternative
explanation for our findings in terms of a mediating effect
of mood caused by the elderly stereotype primes appears
untenable.
Experiment 3: Behavioral Effects of the African
American Stereotype
The results of Experiment 2 showed that the automatic activation
of the elderly stereotype has direct and nonconscious
effects on behavior in line with the content of the stereotype.
Experiment 3 was intended to assess the generality of these results
to an entirely different stereotype--that for African
Americans. As discussed above, Devine (1989) already showed
that this stereotype becomes active automatically to influence
perceptions of a target's hostility. Therefore, according to the
present model, this automatic activation should also produce a
tendency toward hostile behavior in the perceiver.
Another important change in procedure from Experiment 2
was that the priming stimuli were presented subliminally, completely
ruling out experimenter demand or other explanations
of our results in terms of conscious, strategic processes.
Method
Overview. The present experiment was designed to study the effect
of activation of an African American stereotype on social behavior. Participants
were instructed to work on a computerized visual task that
pretesting had shown was considered to be very boring and tedious.
Immediately before each trial, the computer flashed a subliminal picture
of a young African American male face or a picture of a young
Caucasian male face. On the 130th trial, the computer alerted the participant
of an ostensible data-saving failure and also informed the participant
that he or she would have to do the entire computer task again.
A hidden video camera was placed in the lab room to capture participants'
facial reactions to the ostensible computer error and news that
the task would have to be redone. Also, the experimenter rated the hostility
of the participant's reaction. The hypothesis was that participants
presented subliminally with African American faces would react with
greater hostility to the computer error, compared to participants primed
with Caucasian faces.
Participants, Participants were 41 non-African-American undergraduate
students from New York University who participated in the
experiment to partially fulfill course credit.
Apparatus. A Gateway 486 computer with a VGA color monitor
was used to administer the priming manipulation. A Visual Basic program
was created that allowed the experimenter to be blind to priming
conditions, gave participants directions about an odd-even task, and
administered the experimental trials. When the experimenter entered
the participant's ID number into the computer, the computer randomly
assigned the participant to a condition, keeping the participant's condition
from the experimenter's knowledge. During the experimental
phase, each participant was subliminally exposed to black-and-white
photographs of African American or Caucasian faces. On a given trial,
the face photograph was followed by two different pattern masks in
rapid succession. The first pattern mask was composed of a black-andwhite
pattern of diagonal cross-hatches. The second mask, presented
immediately after the first, was conceptually similar but not identical to
the target picture, which was composed of from 4 to 20 colored circles
on a gray background. Pretesting had shown that the masking procedure
was effective in that individuals were not aware of the presence of
the face photographs.
The presentation speed of each subliminal picture (faces and masks)
was bounded by the computer hardware. Specifically, the 76 Hz monitor
screen refresh rate resulted in a minimum presentation time of 13 ms
and a maximum of 26 ms. Each target picture was then presented for 3
s before it disappeared and the participant was asked to make the odd
or even response.
Participants were videotaped throughout the odd-even task by means
of a hidden video camera. The video camera allowed a clear view of
facial expressions while the participant completed the computer task
and was exposed to the ostensible computer error.
Procedure. Each participant completed the experiment individually.
When brought into the experimental room, he or she was informed
that the experimenter was interested in how individuals use different
aspects of a picture to make quick judgments. Pictures of 4-25 different
colored circles would appear inside a small box drawn on the screen for
a few seconds at a time. For each picture, the participant was to make a
decision as to whether the picture contained an odd or an even number
or circles. Although each target picture was on for only 2 or 3 s, the
participant was asked to make the most accurate judgment possible.
The experimenter left the room after the participant completed the
practice trials and when the experimenter was sure the participant understood
the directions.
At the end of the 130th trial, the computer program beeped and displayed
an error message stating "F11 error: failure saving data." After
the experimenter clicked on the screen response button "OK" the computer
displayed another message: "You must start the program over
again," The participant was then instructed to get the experimenter. The
experimenter came back into the room and fiddled with the computer,
finally announcing "I'm sorry, but you'll have to do the experiment over
again." After more fiddling, the experimenter concluded "Actually, it
looks like the computer did save your data. You don't have to do it over
again." The critical dependent measures were the participant's videotaped
reactions to the news that he or she might have to do the task
AUTOMATIC SOCIAL BEHAVIOR 239
overagain and the experimenter's rating of the participant's degree of
hostility.
Subsequently,participants wereasked to fillout two questionnaires:
the Racial Ambivalence Scale (Katz & Hass, 1988) and the Modern
Racism Scale(McConahay, 1986). Participants wereinformed that the
scales werejust part of a pilot pretest in preparation for some future
experiments and werenot linkedto the previousexperiment. Whenthe
experimenter leftthe room to allowthe participant to fillout the surveys
in private, the experimenter recorded her own ratings of the participant's
irritability, hostility, anger,and uncooperativenesson 10-point
unipolarscalesbased on her interaction with the participant.
Finally, participants were probed for suspicion and then debriefed.
Participants wereasked if they had seenanything unusual in the computer
pictures, if they thought the presence of any pictures could have
changed their behavior in any way,and to describe how.Only 2 of the
participants actually reported seeingfacesin the computer task. However,those
participants could not identifywhether the faceswereCaucasian
or African American, and the participants could not guessthat
the study wasabout race in anyway.Furthermore, no participants were
suspiciousthat completing the two race scaleshad anything to do with
the present experiment. Participants wereinformed that they had been
videotaped to obtain facialreactions to the computer error.
Twocoders who wereblindto experimental conditionsand hypotheses
rated all the videotaped facialexpressionsfrom the 41 participants
on a 5-point unipolar scaleofhostilityrangingfrom0 (notat allhostile)
to 10 (extremely hostile). The correlation between the two raters was
.64, and discrepanciesbetweenthe two sets of ratings wereresolvedby
averagingthe two setsinto an overallhostilityindex.
Figure 3. Meanhostility ratings of the participants' reactions to the
computer error, by stereotypeprimingcondition, separately for the experimenters"and
the blindcoders"ratings(Experiment 3).
Results
We conducted a MANOVA using an average of the experimenter's
hostility ratings across the four trait scales and an average
of the two videotape coders' hostility ratings as dependent
measures, and priming condition as the independent measure.
The MANOVA revealed a significant effect of priming condition
across both indexes of hostility, F( 1, 39) = 6.95, p < .05,
such that participants primed with photographs of African
American faces behaved in a more hostile fashion (M = 2.79)
compared to participants primed with Caucasian faces (M =
2.13; see Figure 3). Furthermore, codings of hostility did not
significantly differ by source of rating (experimenter vs. blind
coders), F( 1, 39) < 1, and there were no interactions between
rating source and priming condition, F( 1, 39) < 1.
To examine the possibility that our effects could have been
moderated by attitudes toward African Americans, we calculated
correlations between the averaged hostility measure across
rater type and the participant's level of racism as measured by
the Racial Ambivalence Scale and the Modern Racism Scale
for each priming condition separately. None of the correlations
between hostility and racism as measured by either scale were
significantfor either priming condition (all rs < .25, ps > .35 ).
Thus, participants who were low in racist attitudes toward African
Americans were just as likely to behave in a hostile manner
as participants who were high in racist attitudes, regardless of
priming condition.
This finding corresponds to that of Devine (1989, Experiment
2), who found that the automatic effect of the African
American stereotype on social perception did not vary as a
function of level of consciously expressed prejudice, as measured
by the Modern Racism Scale. More recently, Fazio, Jackson,
Dunton, and Williams (1995) also found no moderation
by Modern Racism scores of their obtained effect of racial attitudes
on behavior (but see Lepore & Brown, 1996).
General Discussion
Across three experiments, the activation of a trait construct
or a stereotype in one context resulted in behavior consistent
with it in a subsequent unrelated context. The participants were
not aware of the influenceor potential influenceof the priming
events on their behavior. The same priming techniques that
have been shown in prior research to influence impression formation
produce similar effects when the dependent measure is
switched to social behavior.
One explanation for the present results is in terms of situationally
linked behavior; that is, that a person has behavioral
responses (e.g., assertiveness, anger, patience) that are associativelylinked
to particular situations(e.g., being made to wait by
another person, losing one's work because of another's
mistake). That people have automatic behavioral as well as perceptual
responses to the social environment is congruent with
the proposal by Lewin (1943), Mischel (1973), Berkowitz
(1984), Higgins (1987), and others that immediate psychological
reactions to the environment are not only cognitive or
affective in nature but also include motivationaland behavioral
responses. The present results show that these behavioral responses
become automatically linked to representations of social
situations just as previous research has found perceptual
trait constructs, stereotypes, and attitudes to become automatically
activated.
Although the present results are consistent with the hypothesis
that behavioral representationsare linked directly to the representations
of social situations, another theoretical perspective
240 BARGH, CHEN, AND BURROWS
consistent with the present findings is that of an automatic perception-behavior
link (e.g., Berkowitz, 1984; Carver et al.,
1983). Because our studies used the same kind of trait and stereotype
priming manipulations that have been shown by prior
research to affect impression formation (e.g., Devine, 1989;
Srull & Wyer, 1979), we now know the identical priming manipulations
result simultaneously in automatic perceptual and
behavioral effects.
Thus, our results are in harmony with those of Carver et al.
(1983), who advanced a "behavioral schema" model as an explanation
for modeling effects. According to the behavioral
schema model, the perceptual and actional representations of
the same type of behavior share many features in common and
thus develop strong connections. As a result, if one has just perceived
another person acting in a generous or an aggressive way,
for example, one's behavioral schema for generosity or aggression
is activated and accessible, and so one is more likely to behave
that way oneself in subsequent situations for which generosity
or aggression is a relevant response.
Under What Conditions Will Behavior Be Automatic?
To many, we presume, the present findings conjure up images
of the old subliminal advertising debate. In the stock example,
frames of Pepsi bottles or boxes of popcorn were flashed in the
middle of a movie and were supposed to create a mad rush to
the concession counter. That kind of ploy did not and does not
work (see Kihlstrom, Barnhardt, & Tataryn, 1992; Moore,
1982). Why, then, were we able to prime behavior nonconsciously
in the present studies?
For one thing, our priming stimuli directly activated trait
concepts, either directly as in Experiment 1 or by virtue of their
inclusion in all-or-none stereotypic representations, as in Experiments
2 and 3. As described above, these representations
contain the knowledge of what it means to act in the traitlike
manner, as well as the mechanisms for producing that kind of
behavior (see Carver et al., 1983; Prinz, 1990). We have gone
directly to the behavior representation, in other words; but a
picture of a bottle of Pepsi would directly activate only the
"Pepsi" representation, or the "soft drink" or "cola" representation.
Unlike the present priming manipulations, it takes an
additional step for "Pepsi" activation to spread to a behavioral
representation such as "to drink" or "to buy."
For the sake of argument, however, let us assume that because
drinking it is what one usually does with a bottle of Pepsi, there
exists an automatic link between the Pepsi representation and
the "drink" behavior representation, so that the picture of a
bottle of Pepsi does activate the "drink" behavior representation.
Why, then, do people not act on that automatic behavior
activation? Consideration of this question highlights another
difference between the present experimental situations and the
movie theater: the ease or relevance of the automatic behavior
to the person's current situational goals.
We constructed the "interruption" situation so that both of
the primed behaviors--rudeness/assertiveness or politeness/
patience--were relevant responses to it. Similar to walking
down the hall after the session in Experiment 2 was ostensibly
over, one can walk to the elevator as quickly as one wishes. In
Experiment 3, the hostility that we assume was activated as part
of the primed stereotype was one of the appropriate responses
to the computer error and seeming loss of the last 10 minutes'
work. In short, the primed behavior was relevant and appropriate
for the experimental situation into which we placed the participant.
In accessibilitylogic, the representation of that behavior
was applicable to the situational information (Higgins, in
press).
Not so with the movie situation. One is there to watch the
movie, and getting up and leaving one's seat conflicts with that.
It also conflicts with another goal inherent in the theater situation,
that of not disturbing the others in the audience also trying
to watch the film. The subliminal message would affect behavior
only for those in the audience for whom getting up to buy a
soda or popcorn is a behavior they have associated with that
situation, that is, a behavioral representation linked to the situational
features. For those of us who buy concessions only on
the way in to our seats, and not after the movie starts, there is
nothing there to prime. Automatic activation can occur only if
the individual has that behavioral representation available in
the first place.
Thus, automatic social behavior may occur only if the behavioral
representation that is activated is already associated with
that situation by the individual. In the present experiments, it is
likely that all primed behaviors were in the participants' repertoire
for those situations. Lewin's (e.g., 1943) field theory holds
that although you might be able to affect a person's behavior by
making some motivations more salient than others, you cannot
give the person a motivation that he or she does not already have
and make him or her do something for which he or she has no
motive base (Cartwright, 1959). It is doubtful, for example,
that the participants in Experiment 2 left our building to go buy
condos in Florida.
Relative Strength ofPriming Effects on
Behaviors VersusImpressions
The limited range of behavioral options that participants had
in our experimental situations may also account for the strength
of the findings, especially in Experiment 1 in which participants'
responses appeared polarized toward either the rude or
the polite option. But in general, across the three experiments,
obtaining priming differences on these behavioral measures was
considerably easier than in our laboratory's previous research
on social judgment measures (e.g., Bargh et al., 1986; Bargh,
Lombardi, & Higgins, 1988; Bargh & Pietromonaco, 1982).
This may be surprising news to those who might assume that
because judgments and perceptions mediate behavior, if anything
the impression effects should be stronger.
Although ultimately one cannot really compare the strength
of priming effects on behavioral versus judgmental measures,
because there are too many differences between the experimental
situations to allow for a meaningful comparison, we can
point to some differences that are likely to produce greater or
lesser priming effects. The major difference, we believe, is that
public behavior such as we measured here is more constrained
than are subjective judgments along personality scales; there is
less variability, and so effects are easier to detect.
AUTOMATIC SOCIAL BEHAVIOR 241
Does Social Perception Always Mediate Social Behavior?
However, the greater variabilityand freedom that people have
in perceiving others' behavior also suggests that social perception
will be less of a mediating influence on behavior than many
might assume. Perceptions may play a role, but they are one of
several influences, and it is clear from the present results that
behaviors can be triggered directly without being mediated by
impressions or judgments of the person with whom one is interacting.
In Experiment 1, impressions of the experimenter
were not affected by the priming manipulation, whereas behavior
strongly was.
One might argue that our impression measure lacked enough
sensitivity to detect the presence of the mediation. In response
to that we would first point out that the means on the impression
measure were in a different direction than the behavior
measure. But more' important, to those who hold to the mediational
model, it must be noted that the evidence for such conscious
mediation of behavior by attributions or judgments is
historically weak, that when behaviors and mediators are measured
in the same study it is the behavioral measure that shows
the clear, predicted difference and the mediator that shows weak
or nonexistent differences, just as in the present Experiment 1
(see Bargh, in press; Bern, 1972, pp. 50-51). As Bern (1972)
put it, although it seems intuitively plausible that changes in
perceptions will be reflected in changes in behavior:
There seems to be only one snag. It appears not to be true. It is not
that the behavioral effectssometimes failto occuras predicted;that
kind ofnegativeevidence rarelyembarrassesanyone. It isthat they
occur more easily, more strongly, more reliably, and more persuasivelythan
the attribution changes that are, theoretically, supposed
to be mediatingthem. (p. 50)
Bern supported this statement by reviewing several classic
studies in which measures were taken both of the behavior and
the presumed mediator (e.g., Valins & Ray, 1967; Zimbardo,
Cohen, Weisenberg, Dworkin, & Firestone, 1969); in all of
them, well-designed and conducted as they were, the behavioral
measure showed clear predicted effects, and the evidence of the
"mediator" was nonexistent. In others (e.g., Darley & Latane~
1968), the assumed mediator (e.g., diffusion of responsibility)
was not measured at all but was inferred from the behavioral
measure (e.g., likelihood of helping).
By no means are we arguing here that social perceptions and
judgments do not mediate behavior. Our point is merely that
they need not and, moreover, we should hold the assumption of
mediation to the same standards of proof and evidence to which
we hold the hypothesis of direct, nonmediated behavior instead
of assuming mediation by default. Reviews of the evidence of
such mediation, in fact, find it in short supply (Bargh, in press).
Can Automatic Behavior Be Controlled?
If, as the ideomotor principle has it, the mere thought of behaving
in a certain way increases the tendency to so behave,
and if, as the perception-behavior link notion has it, merely
perceiving the behavior of others (in the media as well as real
life) activates tendencies to behave in the same way, then automatic
behavioral impulses are occurring with great frequency
and regularity. Add to this those situations commonly enough
experienced so that we have behavioral responses strongly associated
with them, and the probability of automatically triggered
behavior becomes even greater.
The question is not, therefore, how often such automatic behavioral
effects occur, but whether and how often they are controlled
or overridden by some conscious intention and purpose.
Control over automatic influences requires three things: (a)
awareness of the influence or at least the possibility of the influence,
(b) motivation to exert the control, and (c) enough attentional
capacity (or lack of distractions) at the time to engage
in the control process (see Bargh, 1989; Wegner, 1994). Awareness
of the automatic effect is necessary for the motivation to be
engaged, and for the motivation to operate to control the automatic
impulse it must be supported by sufficient processing capacity.
Given that control requires all three of these features to
be in place, it is not difficult to see that there are many realworld
circumstances in which not all three are present. Even
with the best of underlying intentions, one cannot control an
influence if one is not aware of its operation, or at least its potential
for operating (viz. Devine, 1989, Experiment 2). Moreover,
even if one is aware of the influence it is possible to slip up
if one is not paying enough attention, as in ironic failures of
control (Wegner, 1994) and action slips (e.g., Norman, 1981 ).
One notable feature of the present demonstrations is that
many of the behaviors automatically triggeredurudeness, slowness,
and hostilityuare negative and so run counter to norms
for socially appropriate behavior. That these effects occurred
despite the general situational norms against them underscores
the strength of the automatic behavior effect. At the same time,
one can imagine situations in which consequences for negative
behavior are sizable enough that motivations would be strong
enough to overcome them--as when the person making you
cool your heels in the hallway is your boss or supervisor. Nonetheless,
we believe the individual's lack of awareness of the
source of the automatic behavior impulses usually translates
into a lack of monitoring or attempt to control them--as well
as a tendency to misattribute them to possible (and justifiable)
causes of which the individual is aware (e.g., Nisbett & Wilson,
1977)--which will also increase the likelihood that the activation
of automatic behavior responses will find expression.
Implicationsfor Stereotype Confirmation
and Empathic Reactions
One important message to be taken from the results of Experiments
2 and 3 is that there may be an automatic, nonconscious
basis for self-fulfilling prophecy effects (e.g., Snyder,
Tanke, & Berscheid, 1977). If the automatic activation of a stereotype
by the physical features (including speech accent, skin
color, gender, and age-related features) of another person causes
the perceiver him- or herself to behave in line with the stereotype
first, the perceiver's own initial behavior to the target could
well produce similar behavior in the stereotyped individual. In
the case of the African American stereotype, for example, the
nonconscious influence of the activated stereotype could cause
the perceiver to behave in a hostile manner to African Americans
he or she encounters and produce behavioral confirmation
of the stereotype (i.e., a hostile response in reaction). It is ira-
242 BARGH, CHEN, AND BURROWS
portant to note that the perceiver would be unaware of the role
his or her own initial hostility had played in this sequence, because
the effect of the stereotype on behavior was nonconscious--the
perceiver would have no conscious experience of
choosing that mode of behavior. Thus, the perceiver's subjective,
phenomenal experience, and hence memory of the event,
would be of the stereotyped group member's unprovoked initial
hostility.
Another important implication is the automatic creation of
empathic reactions. When in the normal course of interacting
with another person, one perceives his or her friendliness or
honesty or anger, and the same kind of behavior in response is
made more likely merely by the ideomotor function, nonconsciously
producing empathic behavioral and emotional reactions
(see Hodges & Wegner, in press). The perception-behavior
link may be an important ingredient in the "glue" that binds
two (or more) interaction partners, keeps them on the same
wavelength, and helps to bring each partner a sense of validation
by others of their experience.
This is not to say that an automatic effect of perception on
behavior is necessarily entirely beneficial for social interaction.
Dodge and Crick's (1990) research on the social cognition of
aggressive children shows that they are both more likely to perceive
another child's action as aggressive in nature and to respond
themselves with aggression. Although Dodge's account
of these and related findingsis in terms of a consciousdeliberate
choice of aggression based on a perceived aggressive provocation
by the other, it also is possible, on the basis of the present
results, that conscious choice may play less of a role in producing
the aggression. It may be that the mere accessibility of the
mental category of aggression increases the likelihood of aggressive
behavior, via the passive perception-behavior link. Moreover,
the automatic effect of priming on behavior that we have
documented is further reason to believe that aggression in the
mass media does produce aggressive tendencies in the viewer
(e.g., Berkowitz, 1984)--perhaps even more insidiously and
pervasivelythan previously believed.
Conclusions
Our central message is that social behavior can be triggered
automaticallyby features of the environment.In Experiment 1,
the same trait-priming manipulationsthat have exerted a nonconscious
influenceover social perceptual processes in previous
research were shown to produce traitlike behavior as well. Experiments
2 and 3 showed that traitlike behavior is also produced
via automatic stereotype activation if that trait participates
in the stereotype. The major implications of the findings
are, first, the apparent degree to which social behavior occurs
unintentionallyand without conscious involvementin the production
of that behavior. Second, the findings point to the possibilitythat
the automatic activation of one's stereotypes of social
groups, by the mere presence of group features (e.g., African
American faces in Experiment 3), can cause one to behave
in line with that stereotype without realizing it (e.g., with hostility
of facial expression or tone of voice). By this first strike,
therefore, one may elicit that very type of behavior in response.
But because one is not aware of one's own role in provoking it,
one may attribute it to the stereotyped group member (and,
hence, the group).
As a first demonstration of automatic social behavior, the limits
and parameters of these effects remain to be established.
However, an important theoretical point can already be made:
Social behavior is like any other psychological reaction to a social
situation, capable of occurring in the absence of any conscious
involvementor intervention.The implications for many
social psychological phenomena--among them conformity,
emotional and behavioral contagion, empathy, imitation and
modeling, and the behavioral confirmation of stereotypes--
would appear to be considerable.
References
Ansfield,M., & Wegner,D. M. (1996). The feelingof doing. In E M.
Gollwitzer& J. A. Bargh(Eds.), The psychology ofaction (pp. 482-
506). NewYork:Guilford Press.
Bandura, A.( 1977). Sociallearning theory,EnglewoodCliffs,NJ: Prentice
Hall.
Bargh, J. A. (1982). Attention and automaticity in the processingof
self-relevantinformation. Journal ofPersonality and Social Psychology,
49, 1040-1053.
Bargh, J. A. (1989). Conditional automaticity: Varietiesof automatic
influencein socialperception and cognition. In J. S. Uleman &J. A.
Bargh (Eds.), Unintended thought (pp. 3-51 ). New York:Guilford
Press.
Bargh, J. A. (1992). Does subliminality matter to social psychology?
Being aware of the stimulus versus aware of its influence. In R. E
Bornstein & T. Pittman (Eds.), Perception without awareness (pp.
236-255 ). NewYork:Guilford Press.
Bargh,J. A'.(1994). The Four Horsemenofautomaticity. In R. S.Wyer
&T. K. Srull (Eds.), Handbook ofsocial cognition (pp. 1-40). Hillsdale,
NJ: Erlbaum.
Bargh,J. A. (in press). The automaticity ofeverydaylife.In R. S.Wyer
(Ed.), Advances in socialcognition (Vol.10).Hillsdale,NJ: Erlbaum.
Bargh,J. A., Bond,R. N., Lombardi, W.J., &Tota, M. E. (1986). The
additivenature ofchronic and temporary sourcesofconstruct accessibility.Journal
ofPersonalityand Social Psychology,50, 869-878.
Bargh,J. A., Chaiken,S., Govender,R., &Pratto, F.(1992). The generalityofthe
automatic attitude activationeffect.Journal ofPersonality
and Social Psychology,62, 893-912.
Bargh,J. A.,Chaiken,S., Raymond,P., &Hymes,C. (1996). The automatic
evaluation effect: Unconditionally automatic attitude activation
with a pronunciation task. Journal of Experimental Social Psychology,
32, 104-128.
Bargh, J. A., Lombardi, W., & Higgins,E. T. (1988). Automaticityin
Person × Situation effectson person perception: It'sjust a matter of
time. Journal ofPersonalityand Social Psychology,55, 599-605.
Bargh, J. A., & Pietromonaeo, P. (1982). Automaticinformation processingand
socialperception:The influenceoftrait information presented
outside of conscious awareness on impression formation.
Journal ofPersonalityand Social Psychology,43, 437-449.
Bargh,J. A., & Thein, R. D. (1985). Individualconstruct accessibility,
person memory,and the recall-judgment link: The caseof information
overload. Journal of Personality and Social Psychology, 49.
1129-1146.
Bargh,J, A., & Tota, M. E. (1988), Context-dependentautomatic processingin
depression:Accessibilityofnegativeconstructswith regard
to selfbut not others. Journal of Personality and Social Psychology;
54. 925-939.
Bem, D. J. (1972). Self-perceptiontheory. In L. Berkowitz(Ed.), Advances
in experimental social psychology (Vol. 6, pp. 1-62). New
York:AcademicPress.
AUTOMATIC SOCIAL BEHAVIOR 243
Berkowitz, L. (1984). Some effects of thoughts on anti- and prosocial
influences of media events: A cognitive-neoassociation analysis. PsychologicalBulletin,
95, 410-427.
Berkowitz, L., & LePage, A. (1967). Weapons as aggression-eliciting
stimuli. Journal ofPersonality and Social Psychology, 7, 202-207.
Brewer, M. B. (1988). A dual process model of impression formation.
In T. K. Srull & R. S. Wyer,Jr. (Eds.), Advances in social cognition
(Vol. l, pp. 1-36). Hillsdale, NJ: Erlbaum.
Brewer, M. B., Dull, V., & Lui, L. ( 1981). Perceptions of the elderly:
Stereotypes as prototypes. Journal ofPersonality and Social Psychology,
41, 656-670.
Carlston, D. E., & Skowronski, J. J. (1994). Savings in the relearning of
trait information as evidence for spontaneous inference generation.
Journal qflPersonalityand Social Psychology,66, 840-856.
Cartwright, D. (1959). A field theoretical conception of power. In D.
Cartwright (Ed.), Studies in socialpower(pp. 183-220). Ann Arbor,
MI: Institute for Social Research.
Carver, C. S., Ganellen, R. J., Froming, W. J., & Chambers, W. (1983).
Modeling: An analysis in terms of category accessibility. Journal of
Experimental Social Psychology, 19, 403-421.
Darley, J. M., & Latane~ B. ( 1968). Bystander intervention in emergencies:
Diffusion of responsibility. Journal of Personality and Social
Psychology,8, 377-383.
Dell, G. S. ( 1986 ). A spreading activation theory of retrieval in sentence
production. PsychologicalReview,,93, 283-321.
Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and
controlled components. Journal of Personality and Social Psychology,
56, 5-18.
Dodge, K. A., & Crick, N. R, (1990). Social information-processing
bases of aggressive behavior in children. Personality and Social PsychologyBulletin,
16. 8-22.
Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995).
Variability in automatic activation as an unobtrusive measure of racial
attitudes: A bona fide pipeline? Journal ofPersonality andSocial
Psychology,69, 1013-1027.
Fazio, R. H., Sanbonmatsu, D. M., Powell, M. C., & Kardes, E R.
(1986). On the automatic activation of attitudes. Journal ofPersonality
and Social Psychology,50, 229-238.
Fiske, S. T. (1989). Examining the role of intent: Toward understanding
its role in stereotyping and prejudice. In J. S. Uleman & J. A. Bargh
(Eds.), Unintended thought (pp. 253-283). New York: Guilford
Press.
Fiske, S. T. (1992). Thinking is for doing: Portraits of social cognition
from daguerrotype to laserphoto. Journal of Personality and Social
Psychology,63, 877-889.
Fiske, S. T., & Neuberg, S. L. (1990). A continuum of impression formation,
from category-based to individuating processes: Influences
of information and motivation on attention and interpretation. In
M. P. Z,anna (Ed.), Advances in experimental socialpsychology (Vol.
23, pp. 1-74). New York: Academic Press.
Gilbert, D. T. (1989). Thinking lightly about others. In J. S. Uleman &
J. A. Bargh (Eds.), Unintended thought (pp. 189-211 ). New York:
Guilford Press.
Gilbert, D, T., Pelham, B. W., & Krull, D. S. ( 1988 ). On cognitivebusyness:
When person perceivers meet persons perceived. Journal ofPersonality
and Social Psychology,54, 733-739.
Harris, L., & Associates. (1975). The myth and reality of aging in
America. Washington, DC: National Council on Aging.
Hayes-Roth, B. (1977). Evolution of cognitivestructure and processes.
PsychologicalReview,,84, 260-278.
Herr, P. M. (1986). Consequences of priming: Judgment and behavior.
Journal ofPersonality and Social Psychology,51, 1106-1115.
Higgins, E. T. ( 1987). Self-discrepancy:A theory relating selfand affect.
PsychologicalReview, 94, 319-340.
Higgins, E. T. (1989). Knowledge accessibility and activation: Subjectivity
and suffering from unconscious sources. In J. S. Uleman & J. A.
Bargh (Eds.), Unintended thought (pp. 75-123 ). New York:Guilford
Press.
Higgins, E. T. (in press). Knowledge activation: Accessibility, applicability,
and salience. In E. T. Higgins & A. W. Kruglanski (Eds.), Social
psychology.-Handbook ofbasicprinciples. New York: Guilford Press.
Higgins, E. T., & King, G. A. ( 1981). Accessibility of social constructs:
Information-processing consequences of individual and contextual
variability. In N. Cantor & J. Kihlstrom (Eds.), Personality, cognition,
and social interaction (pp. 69-122 ), Hillsdale, NJ: Erlbaum.
Hodges, S., & Wegner, D. M. (in press). Automaticity and control of
empathy. In W. J. lckes (Ed.), Empathic accuracy New York:Guilford
Press.
James, W. (1890). Principles ofpsychology, New York: Holt.
Katz, I., & Hass, R. G. (1988). Racial ambivalence and American value
conflict: Correlational and priming studies of dual cognitive structures.
Journal ofPersonality and Social Psychology,55, 893-905.
Kihlstrom, J. E, Barnhardt, T. M., & Tataryn, D. J. (1992). Implicit
perception. In R. E Bornstein & T. S. Pittman (Eds.), Perception
without awareness (pp. 17-54). New York: Guilford Press.
Koffka, K. (1935). Principles of Gestalt psychology, New York: Harcourt,
Brace.
LaPiere, R. T. (1934). Attitudes vs. actions. SocialForces, 13, 230-237.
Lashley, K. S. ( 1951). The problem of serial order in behavior. In L. A.
Jeffress (Ed.), Cerebral mechanisms in behavior: The Hixon symposium
(pp. 112-136 ). New York: Wiley.
Lepore, L., & Brown, R. (1996). Category and stereotype activation:Is
prejudice inevitable?Manuscript submitted for publication.
Lewin, K. (1943). Defining the "field at a given time." Psychological
Review,,50, 292-310.
Lombardi, W. J., Higgins, E. T., & Bargh, J. A. (1987). The role of
consciousness in priming effects on categorization. Personality and
Social PsychologyBulletin, 13, 411-429.
McConahay, J. B. (1986). Modern racism, ambivalence, and the modem
racism scale. In J. E Dovidio & S. L. Gaertner (Eds.), Prejudice,
discrimination, and racism (pp. 91-125 ). Orlando, FL: Academic
Press.
McTavish, D.G. ( 1971). Perceptions of old people: A reviewof research
methodologies and findings. The Gerontologist, 11, 90-102.
Meltzoff, A. N., & Moore, M, K. (1977). Imitation of facial and manual
gestures by human neonates. Science, 198, 75-78.
Meltzoff, A. N., & Moore, M. K. ( 1983). Newborn infants imitate adult
facial gestures. Child Development, 54, 702-709.
Meyer, D. E., & Gordon, P. C. (1984). Dependencies between rapid
speech perception and production: Evidence for a shared sensorimotor
voicing mechanism. In H. Bouma & D. G. Bouwhuis (Eds.), Attention
andperformance X(pp. 365-377 ). Hillsdale, NJ: Erlbaum.
Miscbel, W. ( 1968). Personality and assessment. New York: Wiley.
Mischel, W. ( 1973). Toward a cognitive social learning reconceptualization
of personality. PsychologicalReview,80, 252-283.
Mischel, W., & Shoda, Y. ( 1995 ). A cognitive-affectivesystem theory
of personality: Reconceptualizing situations, dispositions, dynamics,
and invariance in personality structure. Psychological Review, 102,
246-268.
Moore, T. E. (1982). Subliminal advertising: What you see is what you
get. Journal ofMarketing, 46, 38-47.
Neuberg, S. L. (1988). Behavioral implications of information presented
outside of conscious awareness: The effect of subliminal presentation
of trait information on behavior in the Prisoner's Dilemma
Game. Social Cognition, 6, 207-230.
Newman, L. S., & Uleman, J. S. (1989). Spontaneous trait inference.
In J. S. Uleman & J. A. Bargh (Eds.), Unintended thought (pp. 155-
188). New York:Guilford Press.
244 BARGH, CHEN, AND BURROWS
Nisbett, R. E., & Wilson, T. D. ( 1977 ). Telling more than we can know:
Verbal reports on mental processes. PsychologicalReview, 84, 231-
259.
Norman, D. A. (1981 ). Categorization of action slips. Psychological
Review,,88, 1-15.
Page, M. M. (1969). Social psychology of a classical conditioning of
attitudes experiment. Journal of Personalityand Social Psychology,
11, 177-186.
Perdue, C. W., & Gurtman, M. B. (1990). Evidence for the automaticity
of ageism. Journal of Experimental Social Psychology,26, 199-
216.
Piaget, J. (1946). Laformation du symbole chez l'enfant [Symbol-formation
of the child]. Paris: Delachaux & Niestle7
Posner, M. 1., & Snyder, C. R. R. ( 1975 ). Attention and cognitive control.
In R. L. Solso (Ed.), Inlbrmationprocessingand cognition:The
Loyola symposium (pp. 55-85 ). Hillsdale, NJ: Erlbaum.
Pratto, F., & Bargh, J. A. (1991). Stereotyping based on apparently
individuating information: Trait and global components of sex stereotypes
under attention overload. Journal of Experimental Social
Psychology,27, 26-47.
Prinz, W. (1990). A common coding approach to perception and action.
In O. Neumann & W. Prinz (Eds.), Relationships betweenperceptionand
action(pp. 167-201 ). Berlin: Springer-Verlag.
Salovey, P., & Birnbaum, D. ( 1989 ). Influence of mood on health-relevant
cognitions. Journal of Personality and Social PsTchology, 57.
539-551.
Schank, R. C., & Abelson, R. P. (1977). Scripts, plans, goals and understanding.
Hillsdale, NJ: Erlbaum.
Shiffrin, R. M., & Schneider, W. (1977). Controlled and automatic human
information processing: II. Perceptual learning, automatic attending,
and a general theory. PsychologicalReview,84. 127-190.
Shoda, Y., Mischel, W., & Wright, J. C. (1994). lntra-individual stability
in the organization and patterning of behavior: Incorporating psychological
situations into the idiographic analysis of personality.
JournalofPersonalityand SocialPsychology,67, 674-687.
Smith, E. R. (1994). Procedural knowledge and processing strategies
in social cognition. In R. S. Wyer & T. K. Srull (Eds.), Handbook of
social cognition (2nd ed., Vol. 1, pp. 99-152). Hillsdale, NJ:
Erlbaum.
Smith, E. R. (in press). Mental representation and memory. In D. T.
Gilbert, S. T. Fiske, & G. Lindzey (Eds.), Handbook of socialpsychology(4th
ed.). New York: McGraw-Hill.
Snedecor, G. W., & Cochran, W. G. (1980). Statistical methods'(7th
ed.). Ames: Iowa State University Press.
Snyder, M., Tanke, E. D., & Berscheid, E. ( 1977 ). Social perception and
interpersonal behavior: On the self-fulfilling nature of social stereotypes.
JournalofPersonalityand Social Psychology,35, 656-666.
Srull, T. K., & Wyer, R. S., Jr. (1979). The role of category accessibility
in the interpretation of information about persons: Some determinants
and implications. JournalofPersonalityandSocialPsycholog);
37, 1660-1672.
Strack, F., & Hannover, B. ( 1996 ). Awareness of influence as a precondition
for implementing correctional goals. In P. M. Gollwitzer &
J. A. Bargh (Eds.), The psychology of action (pp. 579-596). New
York: Guilford Press.
Strauman, T. J., & Higgins, E. T. ( 1987 ). Automatic activation of selfdiscrepancies
and emotional syndromes: When cognitive structures
influence affect. Journal of Personality and Social Psychology, 53.
1004-1014.
Uleman, J. S., & Bargh, J. A. (Eds.). (1989). Unintendedthought.New
York: Guilford Press.
Valins, S., & Ray, A. A. (1967). Effects of cognitive desensitization on
avoidance behavior. Journal ofPersonalityand Social Psychology, 7.
345-350.
Wegner, D. M. (1994). Ironic processes of mental control. PsychologicalReview,
101,34-52.
Wegner, D. M., & Bargh, J. A. (in press). Control and automaticity in
social life. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), Handbook
ofsocialpsychology(4th ed.). New York: McGraw-Hill.
Wicker, A. W. ( 1969 ). Attitude versus actions: The relationship of verbal
and overt behavioral responses to attitude objects. Journal ofSocialIssues.
25, 41-78.
Winter, L., & Uleman, J. S. (1984). When are social judgments made?
Evidence for the spontaneousness of trait inferences. Journal ofPersonalityand
SocialPsychology,47, 237-252.
Wyer, R. S., Jr., & Srull, T. K. (1989). Memory and cognition in its
socialcontext. Hillsdale, NJ: Erlbaum.
Zajonc, R. B., Pietromonaco, P., & Bargh, J. A. (1982). Independence
and interaction of affect and cognition. In M. S. Clark & S. T. Fiske
(Eds.), Affect and cognition: The seventeenthannual CarnegieSymposium
on Cognition(pp. 211-228 ). Hillsdale, NJ: Erlbaum.
Zimbardo, P. G., Cohen, A., Weisenberg, M., Dworkin, L., & Firestone,
I. (1969). The control of experimental pain. In P. G. Zimbardo
(Ed.), The cognitivecontrolof motivation (pp. 100-125 ). Glenview,
IL: Scott, Foresman.
Received August 2, 1995
Revision received February 14, 1996
Accepted March 3, 1996 ˇ