Journal of Personality and Social Psychology Copyright 1996 by the American Psychological Association. Inc. 1996, Vol. 71, No. 3,464-478 0022-3514/96/$3.00 Automatic Activation of Impression Formation and Memorization Goals: Nonconscious Goal Priming Reproduces Effects of Explicit Task Instructions Tanya L. Chartrand and John A. Bargh New York University According to the auto-motive model (J. A. Bargh, 1990), intentions and goals are represented men- tally and, as representations, should be capable of nonconscious activation by the environmental context (i.e., "priming"). To test this hypothesis, the authors replicated 2 well-known experiments that had demonstrated differential effects of varying the information-processing goal (impression formation or memorization) on processing the identical behavioral information. However, instead of giving participants the goals via explicit instructions, as had been done in the original studies, the authors primed the impression formation or memorization goal. In both cases, the original pattern of results was reproduced. The findings thus support the hypothesis that the effect of activated goals is the same whether the activation is nonconscious or through an act of will. One's current intentions and goals affect not only what one considers important enough to pay attention to, but also how one uses, interprets, and subsequently remembers that informa- tion. Although that is a noncontroversial statement today, it was a radical departure from the dominant view of perception when Bruner and Postman (1948) originally proposed it. To claim that motivation influences perception was a major break with the then-dominant view that perception and judgment were en- tirely stimulus-driven (Stevens, 1951 ). The result of their claim that needs and motivations influence perception was the "New Look"--a flood of studies demonstrating that an individual's goals greatly influence which information the individual at- tends to and perceives in the environment, as well as how he or she interprets and remembers that information (Allport, 1955 Bruner, 1951, 1957). Jones and Thibaut (1958) subsequently introduced this idea to the domain of social perception, describ- ing the influence that various potential interaction goals might have on selective attention to and use of information about one's interaction partner. After a period in which motivational and cognitive accounts of phenomena were viewed as mutually exclusive and compet- ing instead of complementary and interdependent (see Goll- witzer & Bargh, 1996 Sorrentino & Higgins, 1986b), there re- cently has been a significant advance in theory and research on the motivation-cognition interface (e.g., Chaiken, Giner-Soro- lla, &Chen, 1996 Goilwitzer & Moskowitz, 1996 Higgins & This research was supported in part by Grant SBR-9409448 from the National Science Foundation. We would like to thank Annette Lee Chai, Serena Chen, Roger Giner-Sorolla, and Mark Chen for their help- ful comments. Correspondence concerning this article should be addressed to Tanya L. Chartrand or John A. Bargh, Department of Psychology, New York University, 6 Washington Place, Seventh Floor, New York, New York 10003. Electronic mail may be sent via the Internet to tanyac@ xp.psych.nyu.edu or to bargh@xp.psych.nyu.edu. Sorrentino, 1990 Hilton & Darley, 1991 Sorrentino & Hig- gins, 1986a). It is now widely acknowledged that the intentions and goals people have while interacting with each other exert a powerful influence over the ways they attend to, perceive, judge, and remember that information. There has also been a comple- mentary increase in our understanding of the cognitive bases of motivation. As to the effect of goals on cognition, there is now abundant evidence that the particular goal a perceiver brings to a social interaction greatly affects the perceiver's processing of that in- formation--that is, the way he or she organizes it in memory, recalls it, and forms it into impressions and judgments. For ex- ample, the goal to form an impression of a target person leads to a greater degree of thematic organization of presented behav- ioral information about the target than does the goal to memo- rize the information (Hamilton, Katz, & Leirer, 1980b). Im- pression formation processes have been shown to differ as a function of whether or not one's important outcomes depend on the other person (Erber & Fiske, 1984 Fiske & Neuberg, 1990 Neuberg & Fiske, 1987) or whether or not one anticipates future interactions with the target (Devine, Sedikides, & Fuhr- man, 1989), with less stereotypic and more individuating judg- ments formed under outcome dependency and anticipated-in- teraction conditions. The range of social information processes, from attention allocation, encoding, memory organization, and retrieval to higher order judgment, have all been shown to vary as a function of the information-processing goals operative in a given situation (for reviews, see Gollwitzer & Moskowitz, 1996 Srull & Wyer, 1986 Wyer & Srull, 1986, 1989). Yet whereas models have recognized the important deter- mining nature of the current goal for the outcome of informa- tion processing and judgment, until recently there has been lit- tle research on the cognitive determinants of the goal itself. In experimental situations the various processing goals usually are given to participants explicitly through instructions (see Bargh, 1990), leaving open the question of what determines which goal 464

AUTOMATIC ACTIVATION OF PROCESSING GOALS 465 the individual will pursue when left to his or her own devices. Recently, however, three different sources of goals have been identified and discussed. The first source of goals is structural: the relative power differential between people interacting (Brewer, 1982 Fiske, 1993 Kipnis, 1976). To the extent that one party in an interac- tion has power over another, the relatively powerless person is motivated to process information about the relatively powerful person effortfully and accurately, because his or her important outcomes depend on what that person does. The relatively pow- erful person, on the other hand, does not have his or her out- comes in the hands of the other, and faces no consequences for an inaccurate impression, and so is more likely to use less effortful processing strategies, such as stereotyping, in forming an impression of the other. A second avenue of research on sources of goals has taken an individual-difference tack. Several recent research programs exemplify this approach. First, Cialdini, Trost, and Newsom ( 1995 ) developed a scale that identifies those who do versus do not have a preference for consistency among their beliefs, atti- tudes, actions, and judgments. Cialdini et al. conducted a study in which only persons who did have this preference--roughly half of the participants studied--showed the classic cognitive consistency effects in three standard paradigms: balance, disso- nance, and "foot in the door." Thus, people apparently differ in the extent to which maintaining consistency in their cognitions is an important cognitive goal. Relatedly, Chaiken and her colleagues (Chen, Shechter, & Chaiken, 1996) demonstrated individual differences in the goals pursued in processing persuasive messages. High self- monitors, for example, are much more likely than low self-mon- itors to pursue a goal of having their interaction partner like them (termed an "impression-motivated goal") and so position their attitude to be in line with what they believe to be the part- net's attitude. Yet another example of the individual-difference approach was provided by Jarvis and Petty (1996), who differ- entiated people who do versus do not have a "need" or prefer- ence to evaluate as good or bad the people, objects, events, and other information they encounter in the environment. The focus of the present experiments was on a third way in which the goal-source issue has been addressed: the auto-motive hypothesis (Bargh, 1990 Bargh & Gollwitzer, 1994). This model holds that although many of the goals an individual pur- sues are the result of conscious deliberation and choice, con- scious choice is not necessary for goal activation and operation. First, the model assumes that goals and intentions are repre- sented in memory in the same way that social attitudes, con- structs, stereotypes, and schemas are. Second, because con- structs and stereotypes are capable of being automatically activated by relevant environmental stimuli (for reviews, see Bargh, 1994 Greenwald & Banaji, 1995), goal representations should have this capability as well, given the same conditions that lead to the development ofautomaticity in the other forms of representation. That is, with repeated and consistent choice (i.e., activation) of a particular goal in a certain social situation over time, the representation of that goal may become directly and automatically linked in memory to the representation of that situation (Bargh, 1984, 1990 Hebb, 1948 Posner, 1978). As a result, situational features in the environment can auto- matically trigger goals chronically associated with those fea- tures. The auto-motive model further holds that the automati- cally activated goal, in turn, activates plans to achieve the goal and that these plans then operate interactively with the available goal-relevant information in the environment (Norman & Shallice, 1986 Schank & Abelson, 1977 Wilensky, 1983). Ac- cording to the model, the entire sequence of goal activation and operation can occur without the individual's intention or aware- ness (Bargh & Barndollar, 1996). Bargh, Gollwitzer, and their colleagues demonstrated in sev- eral recent experiments that social-behavioral goals can be au- tomatically activated or "primed" (for reviews, see Bargh, in press Bargh & Barndollar, 1996 Bargh & Gollwitzer, 1994). These researchers used priming or temporary accessibility to simulate the effects of chronic goal accessibility. Thus, goals made more accessible through priming were expected to behave as did goals made more chronically accessible through frequent and consistent use. Previous research on impression formation has shown that, indeed, the qualitative effects of primed and chronically accessible trait constructs are identical (Bargh, Bond, Lombardi, & Tota, 1986 Bargh, Lombardi, & Higgins, 1988). In Bargh, Gollwitzer, and Barndollar's (1996 Experi- ments l and 2) research, participants primed via a "language test" in an ostensibly unrelated first experiment behaved in line with the primed goal in a second experiment in which that goal could be pursued. Thus, for example, participants whose achievement goal had been nonconsciously primed attained higher scores on a word search puzzle than did control group participants. Given this support for the auto-motive principle in the be- havioral realm, it is reasonable to expect that cognitive, infor- mation-processing goals can also be activated nonconsciously via priming techniques. According to the auto-motive model, how a goal representation becomes activated--whether con- sciously or nonconsciously--has no effect on whether it oper- ates and produces its effects. Thus, if we nonconsciously primed the social information-processing goals that previous research- ers have given to their experimental participants via explicit in- structions, we would expect to find the same qualitative effects. T h e P r i m i n g o f Cognitive G o a l s Although the hypothesis that cognitive goals can be automat- ically activated has not yet been tested directly, several recent experiments provide encouraging preliminary support for this prediction. These studies are similar in that the processing goals that were (consciously) pursued in one task were shown to carry over to affect cognition during a subsequent, ostensibly unre- lated task, even though participants did not consciously choose the carried-over goal in the second task. Gollwitzer, Heckhausen, and Steller (1990) explicitly in- structed participants in the first phase of their study to adopt either a deliberative or an implemental mind-set while thinking about a personal problem. The deliberative mind-set involved thinking about alternative approaches to solving the problem, whereas the implemental mind-set involved considering spe- cific actions they could take to solve the problem. The second phase of the experiment, which the participants believed to be unrelated to the first, consisted of completing a fairy tale after

466 CHARTRAND AND BARGH being given just the opening sentences one such tale involved a king who had to go away to war but was concerned about leaving his daughter, the princess, at the castle unprotected. Partici- pants who had a deliberative mind-set in the first phase were more likely to write continuations of the fairy tale in terms of all the possibilities the king was thinking about, whereas those who had previously been in an implemental mind-set wrote continuations characterized by actions the king took to solve the problem. Chen et al. (1996) also used the unrelated-first-experiment priming technique to activate one of two motivations in their participants: to make a good impression on another person or to h01d accurate beliefs. Participants were instructed to imagine themselves in a situation in which they were concerned either with making an accurate assessment of that situation or with making a good impression on someone else. In an ostensibly unrelated second experiment, participants were told that they would be discussing their opinions on a specific issue with an- other participant, who was described as holding either a pro or con position on the issue. All participants were then given the same essay about the topic, which contained arguments on both sides, and asked to indicate their own position on the issue. As predicted, the final attitude of those who had recently imagined themselves being concerned with ingratiating themselves with another person was more in line with that of the (fictitious) other participant than was the final attitude expressed by par- ticipants in the accuracy motivation priming condition. Bator and Cialdini (1996) used a priming procedure to acti- vate participants' consistency motivation, that is, the degree to which they were motivated to hold consistent cognitions. In the first task, participants read an essay in which the writer either did or did not express a value for consistency between words and deeds. In the allegedly unrelated experiment that followed, participants wrote a counterattitudinal essay under a free- choice or no-choice condition. In this standard dissonance par- adigm, the usual finding is that attitudes in the free-choice group are more in line with the counterattitudinal essay than attitudes in the no-choice group. However, Bator and Cialdini found this effect only for participants whose consistency moti- vation had been primed by the first task. As noted, a feature of the goal priming manipulations in these studies that is important to the auto-motive hypothesis is that the goals were consciously and actively engaged in during the priming task. Participants pursued a goal themselves, imag- ined themselves pursuing it, or read about another person who pursued it. In all these cases, therefore, the goal representation had been recently and explicitly used. The auto-motive hypoth- esis, on the other hand, posits that goals can be activated by environmental stimuli unconditionally, without previous or current involvement of conscious intention and choice. Thus, an exact and conservative test of the auto-motive hypothesis in the case of cognitive goals waits on a demonstration that goals can be primed passively and nonconsciously, and then produce the same effects as when they are pursued consciously and deliberately. P u r p o s e o f the Present R e s e a r c h The purpose of the present research, then, was to test experi- mentally the notion that cognitive goals can be preconsciously activated by environmental features. As in studies demonstrat- ing the preconscious activation of behavioral goals, we used priming as a proxy for features of the social environment. To the extent that our priming of goals produces the same effects that are found when the same goals are consciously induced (i.e., explicitly via experimental instructions) in previous stud- ies, we would have evidence that goals can become active preconsciously. Specifically, we conducted two experiments in which we tested the hypothesis that well-established findings about the effects of goals on impression formation can be replicated when the cognitive goals are not consciously and explicitly given to participants, but rather are primed nonconsciously. By "non- conscious" we mean that the individual is not aware of having or working toward these goals they are activated by a means other than conscious choice. This can be done by presenting priming stimuli either to participants' conscious awareness (i.e., supraliminal priming) or to their preconscious (i.e., sub- liminal priming), as long as participants are not aware of the potential influence of the priming stimuli on their subsequent information processing.~ In the present experiments, we used both supraliminal and subliminal priming techniques. The two paradigms thai we selected for replication are among the most historically important in social cognition research (Smith, in press). In Experiment 1, we used the Scrambled Sen- tence Test supraliminal priming technique (Srull & Wyer, 1979) to prime either an impression formation or a memoriza- tion goal just before replicating Hamilton et al.'s (1980b) semi- nal impression-set versus memory-set study. In Experiment 2, we used the subliminal priming technique of Bargh and Pietro- monaco ( 1982 see also Bargh et al., 1986) to prime an impres- sion formation goal (or no goal) before replicating the classic person memory study by Hastie and Kumar (1979). E x p e r i m e n t 1 Hamilton et al. (1980b) provided one of the earliest demon- strations of the impact that operative processing goals have on perceivers' memory for others. In their study, participants read a series of sentence predicates describing various behaviors of a target person the behaviors were chosen to represent four dis- tinct personality trait categories. Before this task, the experi- menter induced one of two processing sets (i.e., cognitive goals) in participants by instructing them either to remember as much of the information as possible (a memory set) or to form an impression of the target person described by the various behav- iors (an impression set). After performing a brief filler task, all participants were asked to recall as many of the behavioral descriptions as possible. Counterintuitively at the time, partici- pants whose goal had been to form an impression of the target person actually recalled more of the behavioral descriptions Indeed, in several research domains, such as trait construct accessi- bility and stereotyping, the same priming effects have been found re- gardless of whether the technique used was supraliminal or subliminal what matters for the occurrence of priming effects is not conscious awareness oftbe priming stimuli but awareness of the possible influence of those stimuli on subsequent processing (for a review, see Bargh, 1992).