resisting change: information-seeking and stereotype change

27
European Journal of Social Psychology, Vol. 26, 799-825 (1996) Resisting change: information-seekingand stereotype change LUCY JOHNSTON Department of Psychology, University of Canterbury, New Zealand Abstract Despite recent laboratory successes in demonstrating stereotype change in response to disconfirming information, stereotypes remain resistant to change or modification. The reported research employed an information gathering methodology in which perceivers could control the amount and nature of the information they received about members of a stereotyped group prior to evaluating the group on a number of stereotype-relevant characteristics. Perceivers showed a stereotype-preservation bias in their information gathering (Experiments I and 2) and, consequently, showed no modification of existing stereotypic beliefs. Experiment 3 manipulated the salient processing goals under which perceivers guthered information and.found that, under certain conditions, the stereotype preservation bias could be overcome and stereotypes moderated. INTRODUCTION Recent laboratory studies have successfully demonstrated the moderation of stereotypes in response to behavioural information, including stereotype disconfirming information, about target group members (e.g. Gurwitz & Dodge, 1977; Hewstone, Johnston & Aird, 1992; Johnston & Hewstone, 1992; Johnston, Hewstone, Pendry & Frankish, 1994; Weber & Crocker, 1983). These studies fail, however, to capture an important aspect of everyday information processing. Social perceivers are, by nature, active information gatherers rather than passive information receivers (Yzerbyt & Leyens, 1991). As such, they have the ability to control the nature and amount of behavioural information received from a given target. Findings from the hypothesis testing literature, in both social and non-social domains, demonstrate that giving perceivers the ability to decide to select or reject information produces a bias toward hypothesis matching information and, as a Addressee for correspondence: Dr Lucy Johnston, Department of Psychology, University of Canterbury, Private Bag 4800, Christchurch, New Zealand, e-mail: psyc380(u)psyc.canterbury.ac.n~. CCC 0046-2772/96/050799-27 01996 by John Wiley & Sons, Ltd. Received 18 April 1995 Accepted 1 February 1996

Upload: lucy-johnston

Post on 06-Jun-2016

213 views

Category:

Documents


2 download

TRANSCRIPT

European Journal of Social Psychology, Vol. 26, 799-825 (1996)

Resisting change: information-seeking and stereotype change

LUCY JOHNSTON Department of Psychology, University of Canterbury, New Zealand

Abstract

Despite recent laboratory successes in demonstrating stereotype change in response to disconfirming information, stereotypes remain resistant to change or modification. The reported research employed an information gathering methodology in which perceivers could control the amount and nature of the information they received about members of a stereotyped group prior to evaluating the group on a number of stereotype-relevant characteristics. Perceivers showed a stereotype-preservation bias in their information gathering (Experiments I and 2 ) and, consequently, showed no modification of existing stereotypic beliefs. Experiment 3 manipulated the salient processing goals under which perceivers guthered information and.found that, under certain conditions, the stereotype preservation bias could be overcome and stereotypes moderated.

INTRODUCTION

Recent laboratory studies have successfully demonstrated the moderation of stereotypes in response to behavioural information, including stereotype disconfirming information, about target group members (e.g. Gurwitz & Dodge, 1977; Hewstone, Johnston & Aird, 1992; Johnston & Hewstone, 1992; Johnston, Hewstone, Pendry & Frankish, 1994; Weber & Crocker, 1983). These studies fail, however, to capture an important aspect of everyday information processing. Social perceivers are, by nature, active information gatherers rather than passive information receivers (Yzerbyt & Leyens, 1991). As such, they have the ability to control the nature and amount of behavioural information received from a given target. Findings from the hypothesis testing literature, in both social and non-social domains, demonstrate that giving perceivers the ability to decide to select or reject information produces a bias toward hypothesis matching information and, as a

Addressee for correspondence: Dr Lucy Johnston, Department of Psychology, University of Canterbury, Private Bag 4800, Christchurch, New Zealand, e-mail: psyc380(u)psyc.canterbury.ac.n~.

CCC 0046-2772/96/050799-27 01996 by John Wiley & Sons, Ltd.

Received 18 April 1995 Accepted 1 February 1996

800 L. Johnston

consequence, an increased confidence in the validity of that hypothesis (e.g. Beyth- Marom & Fischhoff, 1983; Darley & Gross, 1983; Doherty, Mynatt, Tweney & Schiavo, 1979; Johnston & Macrae, 1994; Skov & Sherman, 1986; Snyder & Swann, 1978; Wason, 1960, 1968; Yzerbyt & Leyens, 1991). Applied to the stereotype domain, information-seeking may result in a bias toward stereotype-matching information and a maintenance of pre-existing stereotypic beliefs (Johnston & Macrae, 1994). Such an information-seeking strategy is problematic if it supports the maintenance of inaccurate or pejorative social stereotypes. Stereotype change researchers need to understand the role of information search and its impact on stereotype moderation. The reported research introduces a new method, a bulletin board approach which makes information-seeking an interactive process for the perceiver, to investigate information search in a stereotyping domain and to consider its implications for stereotype change.

Using a question-asking paradigm, Johnston and Macrae (1994) showed subjects to select more questions whose features matched their existing stereotype of a target group (i.e. referred to stereotypic characteristics) than questions whose features did not match the stereotype (i.e. referred to counter-stereotypic or stereotype neutral characteristics) to ask members of the stereotyped group. That is, they showed biased information-seeking. In addition, these subjects showed no moderation of their stereotype-based judgments of the group, in contrast to subjects who had been forced to ask all the questions, including those that did not match the stereotype of the target group. Although offering support for the basic tenet that active social perceivers will show biased information-seeking and a maintenance of stereotypic beliefs, this study has a number of limitations which restrict the generalizability of its findings. The study fails to identify the processes underlying the confirmatory bias. It used a question-asking paradigm which may be inappropriate (Evett, Devine, Hirt & Price, 1994). Finally, the study only provided perceivers with stereotype-consistent information and so the impact of discrepant information on information-seeking could not be assessed. The present research goes beyond this, investigating the processes underlying information-seeking in a stereotype domain, the impact of information-seeking on stereotype change and how information-seeking biases can be overcome.

Process underlying the information-seeking bias

Two major factors influencing information-seeking have been identified, the diagnosticity and the confirmatory nature of the information. A diagnostic strategy argues that perceivers select the most diagnostic, or useful, information in testing their hypotheses (e.g. Bassok & Trope, 1984; Trope & Bassok, 1982, 1983; Trope, Bassok & Alon, 1984). A confirmatory strategy, on the other hand, suggests that perceivers are biased toward information which confirms their hypothesis (e.g. Snyder, 1981; Snyder & Campbell, 1980; Snyder, Campbell & Preston, 1982; Snyder & Cantor, 1979; Snyder & Swann, 1978). There has been much debate about which strategy dominates information-seeking, although recent literature has suggested that the two strategies may operate simultaneously (e.g. Devine, Hirt & Gehrke, 1990; Skov & Sherman, 1986). Johnston and Macrae (1994) confounded the confirmatory nature of their questions with diagnosticity; the stereotype-matching

Information seeking and stereotype change 80 1

questions were judged more diagnostic than the stereotype-mismatching questions, so it is unclear whether diagnosticity or the nature of the information was driving question choice.

Diagnosticity may not have a large impact on information-seeking about stereotyped groups. The biased hypothesis testing model (Pyszczynski & Greenberg, 1987) proposes that information gatherers selectively seek information which is expected to logically support their desired conclusions or underlying motivations. Given the cognitive expedience (e.g. Macrae, Milne & Bodenhausen, 1994) and social identity benefits (e.g. Abrams & Hogg, 1985) offered by stereotypes, individuals are motivated to maintain their stereotypic beliefs (see Hamilton & Sherman, 1994). As such, then, they are likely to attempt to select stereotype consistent information, despite stereotype-disconfirming information being, at least under certain conditions, more informative than confirming behaviour (e.g. Darley, Fleming, Hilton & Swann, 1988). Higgins and Bargh (1987) argued that information seekers would normally select diagnostic over non-diagnostic information but would show a preference for hypothesis-matching information when the hypothesis under consideration was the only available alternative and was believed to be true. Many stereotypes are examples of hypotheses which perceivers believe to be true. Similarly, Pyszczynski and Greenberg (1987) suggested that when the outcome of an hypothesis-testing sequence has strong affective implications, as may be the case for stereotyped groups, then any preference for diagnostic information may be overridden. The current research considers the impact of the confirmatory status of information above and beyond any possible impact of diagnosticity. The stereotypic and counter-stereotypic traits are matched for importance to the target group and for each trait the available questions are matched for diagnostic value. Any systematic biases seen in perceivers’ information searches cannot, then, be a result of a diagnosticity bias.

The nature of the information-seeking bias

The term ‘confirmation strategy’ is somewhat of a misnomer when applied to question-asking techniques. It is the nature of the answer which determines whether the information is confirmatory or disconfirmatory (Evett et al., 1994; Higgins & Bargh, 1987). By itself asking certain questions is not a biased information-gathering technique. Hypothesis testing is a multistage process, involving both asking questions and responding to the information provided. The questions perceivers ask set up expectancies. Regardless of the type of question asked, perceivers expect to receive confirmatory information (Evett et al., 1994), suggesting that question choice alone is not sufficient to identify a confirmatory bias. However, asking certain questions can increase the likelihood of the answer providing confirmatory information. If perceivers adopt strategies in their question asking which increase the likelihood of receiving confirmatory answers, then they can be said to be adopting confirmatory biases. Asking about topics especially likely to be present if the hypothesis under test is true is more likely to yield confirming information than asking about features unlikely to be present. Such a bias is known as an hypothesis (or stereotype) preservation bias. Slowiaczek, Klayman, Sherman and Skov (1992) showed hypothesis testers to prefer questions about features that are extremely

802 L. Johnston

probable under the hypothesis. These questions also yielded hypothesis-favouring answers more frequently than hypothesis-opposing answers. Similarly, the focus of spontaneous questions is often consistent with the hypothesis being tested (Zuckerman, Knee, Hodgins & Miyake, 1995). Choosing to ask about stereotypic characteristics is, therefore, likely to yield more stereotype-consistent information than asking about counter-stereotypic topics.

A stereotype preservation strategy involves a bias toward questions about stereotypic characteristics (Johnston & Macrae, 1994). Adopting a stereotype preservation strategy increases the likelihood of receiving stereotype-consistent information. It may also decrease the likelihood of stereotype change. In previous, information-giving, research stereotype moderation in response to disconfirming information has been as a result of an increase in the extent to which counter stereotypic traits are perceived as characteristic of the target group and not due to any decrease in the perceived characteristicness of the stereotypic traits (e.g. Johnston & Hewstone, 1992; Locke, MacLeod & Walker, 1994). A change in stereotype-based judgments of a group is only likely, therefore, if perceivers gather information about counter stereotypic characteristics, contrary to a stereotype preservation bias.

Biases toward certain types of question may also be driven by a desire for cognitive expedience. A bias toward questions to which a ‘yes’ answer would confirm the hypothesis (a hypothesis-matching question) under investigation may reflect both the frequently occurring acquiescence response set (e.g. Ray, 1983) and also the greater difficulty of processing negative than positive information (Fazio, Sherman & Herr, 1982; Jenkins & Sainsbury, 1969, 1970; Newman, Wolff & Hearst, 1980). Such a bias is known as a positive test strategy.

Whether perceivers are adopting a stereotype preservation or a positive test strategy has important implications for the overcoming of such biases and subsequently for stereotype change. Previous research has confounded hypothesis preservation and positive test strategies. Consider, for example, testing the hypothesis that Jack is an extrovert. A preference for the question ‘Does Jack enjoy wild parties’ over ‘Does Jack like to spend his evenings alone?’ could be the result of an hypothesis preservation (the first question refers to hypothesis matching, extroverted, characteristics) or a positive test (a ‘yes’ response to the first question but a ‘no’ response to the second question is hypothesis consistent) strategy. The bulletin board technique used in the present research can identify whether perceivers are using either, neither or both of these types of confirmatory biases. Perceivers first select a trait topic (stereotypic, counter-stereotypic or stereotype-neutral) about which they want additional information. They then choose between two questions (one stereotype-matching and one stereotype-mismatching) related to that trait topic. A stereotype preservation strategy will be reflected in a bias toward selecting stereotypic trait topics about which to ask questions, but there is no bias in the type of question subsequently asked. A positive test strategy, on the other hand, will be reflected in a bias toward stereotype-matching questions, but no bias in the trait topics selected. This technique will be described in more detail in the procedure section of Experiment 1 and is represented in Figure 1.

In addition to identifying the types of question asked i t is equally important to consider perceivers’ responses to different types of information, especially to stereotype-inconsistent information, encountered during information search. Pyszczynski and Greenberg (198 1) predicted that the discovery of expectancy-

Injovmation seeking and stereotype change 803

inconsistent information would be more likely than the discovery of expectancy- consistent information to trigger a search for further expectancy-relevant information. Perceivers should, then, show a stronger preference for stereotype- consistent information after receiving a stereotype-inconsistent rather than a stereotype-consistent answer. Similarly, Evett et al. (1 994) showed perceivers to resist stereotype-inconsistent information and to still expect confirmatory information from subsequent questions. In contrast, then, to Johnston and Macrae (1 994), the present research includes stereotype-inconsistent answers in response to some questions. Importantly, this includes answers to both questions about stereotypic topics and stereotype-mismatching questions.

Factors influencing information-seeking

In their recent article on trait hypothesis testing Zuckerman et al. (1995) advised that future research should consider identifying the conditions under which biases in information-seeking are most likely to occur. The present research manipulates both cognitive (time pressure - Experiment 1) and motivational (processing goals - Experiment 3) factors to investigate exactly that. In particular, the present research was concerned with increasing perceivers’ consideration of counter-stereotypic characteristics.

Perceivers are particularly susceptible to stereotypical thinking when cognitive resources are scarce. Numerous experiments have demonstrated the utility of stereotype activation and application in demanding task environments (e.g. Bodenhausen, 1990; Bodenhausen & Lichtenstein, 1987; Bodenhausen & Wyer, 1985; Gilbert & Hixon, 1991; Macrae, Hewstone & Griffiths, 1993; Pratto & Bargh, 1991). It is predicted that this pattern will be repeated with information-seeking. Perceivers are predicted to show a stronger information-seeking bias under conditions of high time pressure (Experiment 1).

Snyder et al. (1 982) illustrated the difficulty of inducing perceivers to consider counter-stereotypic characteristics. They employed four different interventions attempting to convey the potential utility of learning about ways in which people do not fit stereotypes, encouraging perceivers to ask questions about counter- stereotypic characteristics when testing the accuracy of those stereotypes. None of these interventions resulted in an increase in questions relating to counter-stereotypic characteristics of the target groups. A more fruitful intervention strategy may be to attack the motivation underlying the information-seeking bias. If information search is guided by a motivation to maintain existing stereotypes, then changing that motivation should result in a change to information-seeking strategies. The tendency to engage in stereotypic thinking and stereotype maintenance may be persuasive but social perceivers are capable of avoiding the influence of stereotypes on judgments in situations where there is a motivational impetus to do so (Bodenhausen, Kramer & Siisser, 1994; Fiske & Taylor, 1991; Forgas, 1989). Importantly, a number of processing goals (e.g. outcome dependency, anticipated future interaction) which encourage perceivers to judge target individuals in individual-based rather than group-based terms have been identified (Brewer, 1988; Fiske & Neuberg, 1990). Under such motivational conditions, stereotype inconsistent information is

804 L. Johnston

integrated into evaluations of the targets and stereotype-based judgments correspondingly moderated.

A number of studies have shown motivational factors to also influence information- seeking. Unfortunately, these information-seeking studies did not assess the impact of such strategies on judgments of the target. Outcome dependence increased the number of expectancy-relevant questions perceivers asked about a target (Darley et al., 1988) but questions were only categorized as expectancy relevant or irrelevant and so it is unclear whether the perceived dependence resulted in an increase in the number of expectancy-mismatching questions asked. Anticipated future interaction resulted in balanced information-seeking (Johnston et al., 1994). When no interaction was anticipated subjects chose to ask targets more questions referring to stereotypic than to counter-stereotypic characteristics but when interaction was anticipated equal numbers of questions referring to stereotypic and counter-stereotypic characteristics were chosen. Accuracy instructions alone were not, however, sufficient to overcome stereotype-preservation information-seeking strategies; perceivers still asked more questions about stereotype confirming than disconfirming characteristics (Snyder & Swann, 1978). Information about how such question asking could be used as an impression management strategy (as either open-minded or narrow-minded) resulted in more balanced information-seeking, subjects asking equal numbers of questions about each type of characteristic. These studies suggest that certain processing goals can lead to more balanced information-seeking strategies. Whether this leads, in turn, to moderation of stereotype-based perceptions of targets remains to be investigated. In Experiment 3 salient processing goals were varied. Specifically, processing goals which conflict with stereotype maintenance were employed (accuracy, justification). Under such conditions it was predicted that subjects would not show a stereotype- preservation bias but rather would show a balanced information search, selecting information about both stereotypic and counter-stereotypic traits'. As a consequence, subjects in these conditions would be expected to show a moderation in their stereotype-based judgements of the group.

The present research

In summary, the present research seeks to identify any systematic information- seeking strategies used by social perceivers in gathering information about a stereotyped group, using a new bulletin board technique. The impact of information- seeking on stereotype-based judgments of the group is considered, especially perceivers' response to stereotype-inconsistent information. Factors which may influence information-seeking are also considered; Experiment 1 includes a time pressure manipulation and Experiment 3 manipulates the salient processing goals under which perceivers complete the information-gathering task.

'The emphasis on selecting counter-stereotypic topics to avoid stereotyping may be an artefact of the experimental situation, normally avoiding using stereotypes may be reflected in avoidance of all stereotype-relevant information.

Information seeking and stereotype change 805

EXPERIMENT 1

Overview

Information about members of a target group was presented via a bulletin board, under free or forced presentation conditions. In the free presentation conditions subjects could select the amount and type (trait topic type and matching or mismatching questions) of questions they wished to ‘ask’ members of the target group. In the forced presentation condition, perceivers were given (rather than selected) information. After information gathering, subjects rated the target group on a number of stereotypic and counter stereotypic dimensions. The bulletin board allowed for biases in information-seeking and underlying search strategies to be identified and their impact on stereotype-based judgments assessed.

Method

Subjects and design

Forty-nine undergraduates participated in the experiment in return for payment. The experiment had a 3 (presentation mode: f’orced/free - no time pressure/free -high time pressure) x 3 (trait topics: stereotypic/counter-stereotypic/stereotype-neutral) x 2 (questions: stereotype-matching/stereotype-mismatching) design with the first factor a between-subjects factor and the second and third factors within-subjects factors.

Stimulus materials

A series of pilot studies were conducted to identify a suitable target group and its stereotype. In the first pilot study 10 students were asked to list any groups in society which were easy to distinguish and which they could characterize by a few key features. One frequently identified group was doctors. In the second pilot study 20 students rated doctors on 50 personality traits, including those mentioned by subjects in the first pilot study (1 ~ ‘not at all characteristic’, 5 - ‘unrelated to group’, 9 - ‘extremely characteristic’). From these ratings stereotypic (mean ratings within the top 10 ratings and no more thn two subjects rating the trait in the opposite direction from the scale midpoint to the mean), counter stereotypic (mean ratings among the lowest 10 ratings and no more than two subjects rating the trait in the opposite direction from the scale midpoint to the mean) and stereotype-neutral (mean ratings within two points of the scale midpoint and no more than two subjects rating the trait more than two points from the scale midpoint) traits were identified. Four of each type of trait were chosen for use in the experiment proper: stereotypic: compassionate, tolerant, rational, dependable; counter stereotypic: aggressive, rude, hostile, prejudiced; stereotype-neutral: stable, modest, confident, wasteful. An analysis of variance (ANOVA) on the trait ratings revealed a significant effect of trait type, F(2,119) = 44.63, p < 0.0001. All differences between the mean ratings were significant (Fisher’s PLSD; Ms = 6.58, 3.05 and 5.58 for the stereotypic, counter-stereotypic and

806 L. Johnston

neutral traits respectively). Importantly, the stereotypic and counter stereotypic traits were equally distant from the scale midpoint ( M s = 1.58 and 1.73) suggesting that their strength of association with the target group was equally strong. An additional 10 students rated the target group as a whole on the 12 selected traits to provide baseline data against which the experimental subjects could be compared for evidence of actual stereotype change (1 -‘not at all characteristic’, 7- ‘very characteristic’).

For each of these 12 traits two questions were to be used in the experiment proper, one matching (a ‘yes’ answer confirming the stereotype of the group and a ‘no’ answer disconfirming the stereotype) and one mismatching (a ‘no’ answer disconfirming the stereotype) and one mismatching (a ‘no’ answer confirming the stereotype of the group and a ‘yes’ answer disconfirming the stereotype). In order to select appropriate questions 20 subjects rated each of a pool of 48 questions, constructed by the experimenter and assistants. First, they rated how useful, or diagnostic, the answer to the question would be in terms of the given trait. That is, how useful the question and its answer would be in determining the level of the trait possessed by the person answering (1 - ‘not at all useful’, 9 -‘extremely useful’). Subjects were told to give a higher rating the more the question would distinguish between those with a high and a low level of the trait. For example, ‘Did you join in the brawl at the pub on Friday night?’ may be considered more diagnostic for the trait aggressive than ‘Did you watch the boxing match on televison on Friday night?’ Second, subjects estimated the percentage of doctors who would respond ‘yes’ to each question. For each trait the two questions chosen for use in the experiment proper were matched on their diagnosticity or usefulness ratings ( M s = 6.46,6.60 and 6.08 for the stereotypic, counter-stereotypic and neutral traits respectively) but one was a stereotype-matching (mean % ‘yes’=72 per cent) and one a stereotype- mismatching question (mean % ‘yes’=33 per cent). Sentence pairs were also matched for word length.

Listed below are examples of each type of sentence:

Stereotypic trait (compassion), matching question: Do you worry that there are millions of people at risk from starvation in many parts of the world?; mismatching question: Do you think it is necessary and acceptable for some people to sufler,for the greater good of the majority? Counter-stereotypic trait (rudeness), matching question: Do you think it is important to be polite when meeting people for the first time?; mismatching question: Do you say what you think even if it may offend people? Stereotype-neutral trait (wastefulness), matching question: Do you make an eflort to recycle as much household rubbish as possible?; mismatching question: Do you think it is better to spend money when you have it, rather than saving for the future?

Across the total question set, three-quarters of the stereotype-relevant questions (i.e. those relating to either stereotypic or counter-stereotypic but not to stereotype- neutral traits) were answered in a stereotype-consistent manner (i.e. matching questions answered ‘yes’ and mismatching questions ‘no’) and one-quarter in a stereotype-inconsistent manner (i.e. matching questions answered ‘no’ and mismatching answered ‘yes’). Across the entire question set, then, 0.167 of answers were stereotype-inconsistent. The stereotype-inconsistent answers were evenly distributed across target group members (each gave one inconsistent answer), trait

Infbrmution seeking und stereotype chunge 807

topic (there were two inconsistent answers in each of the stereotypic and counter- stereotypic trait topics) and question type (there were two inconsistent answers each to matching and mismatching questions). It was, of course, impossible to control the exact proportion of stereotype-inconsistent answers received by subjects in the free conditions. However, the distribution across targets, trait types and question types makes avoidance of all the inconsistent information unlikely. As detailed in each results section, the proportion of inconsistent answers received by subjects in the free conditions was, in fact, very close to that received by subjects in the forced condition. The time allowance for subjects in the high time pressure was set at one third of the time it took five pilot subjects to complete the task (M=7.5 minutes). In this condition a clock counting down from 2.5 minutes was visible on the screen so that subjects were aware of how much time they had left.

Procedure

All subjects were informed that they were participating in an experiment on social perception, concerned with how different groups in society (e.g. political, religious, sports groups) are seen by others. They were randomly allocated to one of the three presentation mode conditons, with an equal ratio of males and females in each condition, and tested individually. The experiment was run on an Apple Macintosh (Mac IIsi) computer. Instructions, stimulus materials and dependent measures were all presented via the computer. The first computer screen informed subjects that they would be asked for their perceptions of one particular group, doctors. Before giving their own perceptions of the group, however, they were to be given examples of the behaviour of some members of the target group, which had previously been collected in a series o f interviews of doctors conducted by the experimenter. Subjects were told that, owing to time restrictions, they would receive information about just four of the doctors interviewed. These four had been chosen as representative of doctors in general. Each of the doctors answered questions about different topic areas, each on three different topics. These topic areas were the trait topics identified in the pilot studies. Subjects were instructed to use this information to help them in cementing their perceptions of the target group as a whole. Subjects in the forced condition were instructed to work their way through all the presented information. Subjects in the free condition were instructed to only read as much of the information as they felt they needed in order to be confident in their perceptions of the group as a whole. These subjects were free to terminate the information gathering at any stage. When subjects had read the instructions they clicked on a 'run' button to proceed with the experiment. At this point the experimenter left the room until the subjects had completed the experiment.

The first experimental screen displayed the names of the four doctors. To receive information about any of the doctors subjects had to click on their name. A new screen appeared with the three topics describing that doctor. Across the four doctors these topics were close synonyms of the 12 traits identified in the pilot studies. For each target there was one stereotypic, one counter-stereotypic and one neutral trait topic. To receive information about any of these topics the subject clicked on the topic name. A screen with the two questions related to the topic, one stereotype- matching and one stereotype-mismatching, then appeared. To see the doctor's

808 L. Johnston

Screen4 Screen 1 Screen 2 Screen 3 TARGETS I R A 1 1 TOPICS QUESllONS ANSWERS

MATCHING

COUNTER- STERKOTY PIC

MlSMATCHINO

NEIJTRAL

Figure 1. A schematic representation of the bulletin board technique. This is the format of the bulletin board for Experiments 1 and 3, for Experiment 2 screens 3 and 4 are merged as there is no question choice. The position of targets, trait topics and questions on the screen is randomized across subjects

answer to one of the questions subjects clicked on a box labelled ‘answer’ next to each question. A new screen with either ‘yes’ or ‘no’ on it appeared. Subjects then clicked on an ‘OK’ button to return to the first experimental screen, containing the doctors’ names. After a question had been selected it was removed from the screen. This bulletin board procedure is illustrated in Figure 1. It should be noted that the position of the target names, the traits and the questions was randomized across subjects such that the stereotypic trait was not always at the top of the screen (as is illustrated in Figure 1).

Subjects in the forced presentation condition were required to continue their information search until all the questions and answers had been selected, and therefore disappeared from the screen. The trait rating dependent measure then automatically appeared. The trait rating task asked subjects to rate how characteristic a number of personality traits were of doctors as a whole (1 not at all characteristic’, 7 ~ ‘very characteristic’). The traits were the four stereotypic, four counter-stereotypic and four neutral traits identified in the pilot testing and rated by the baseline subjects. In the free presentation condition subjects could stop the information search at any time by clicking on a ‘finished’ button on the first experimental screen. The trait rating task appeared either when the ‘finished’ button was selected or when all the questions had been asked. If subjects in the free, high time pressure condition had not selected the ‘finished’ button by the end of their allotted time the trait task would appear along with the message, ‘Your time for reading the information is up’.

Infornzation seeking and stereotype change 809

Results and discussion

Information seeking’

A one-way ANOVA on the total number of questions asked revealed a main effect of time pressure, F(1,34) = 24.83, p < 0.0001. Subjects in the high time pressure condition asked fewer questions than those in the no time pressure condition ( M s = 9.72 and 17.18 respectively). To test for information-seeking biases the proportion of total questions asked which came from each trait topic was calculated for each subject. These proportions were subjected to a 2 (time pressure: high/ none) x 3 (trait topics: stereotypic/counter-stereotypic/stereotype-neutral) x 2 (questions: stereotype-matching/stereotype-mismatching) ANOVA. There was a main effect of trait type, F(2,68) = 5.14, p < 0.01. Post-hoc tests (Tukey, p < 0.05) showed that the proportion of questions asked which were about stereotypic traits was greater than that asked about either counter-stereotypic or neutral traits (Ms = 0.392, 0.300 and 0.312 respectively). This bias toward stereotypic traits was reflected in the first questions chosen (Skov & Sherman, 1986). In both the high time pressure and low time pressure conditions the proportion of first questions asked which came from stereotypic trait topics exceeded that expected by chance (t-tests, p < 0.05; Ms = 0.625, 0.571 and 0.333 respectively).

As predicted, perceivers showed a bias in their information-seeking. Support was specifically for a stereotype preservation bias, perceivers showed a bias toward stereotypic topics. There was no evidence of a positive test strategy, subjects asked equal proportions of stereotype-matching and mismatching questions. An explanation for the bias in terms of difficulty of processing negative information (Fazio et a/., 1982; Jenkins & Sainsbury, 1969, 1970; Newman et al., 1980) is not, then, supported. Contrary to predictions, however, the confirmatory bias was no stronger under high than low time pressure conditions. The reliance on stereotypes in information processing may, therefore, be greater in naturalistic, information gathering settings than in passive experimental settings. When perceivers are passive recipients of information, cognitive busyness is an important factor in stereotype usage. Perceivers rely more on stereotypes when cognitive resources are low rather than plentiful (e.g. Bodenhausen, 1990; Bodenhausen & Lichtenstein, 1987; Bodenhausen & Wyer, 1985; Gilbert & Hixon, 1991; Macrae et af., 1993; Pratto & Bargh, 199 1). When perceivers are active information gatherers, stereotypes are used in information-seeking, even in the absence of a high cognitive load.

Information seeking strategies

Evidence of strategic information search was assessed by considering the impact of different types of answer to previous questions on perceivers’ subsequent question choice. The probability of choosing a question from each topic type, given a stereotype-consistent, -inconsistent or -neutral answer to the previous question

’In none of the reported experiments was there an effect of target. Subjects asked an equal number of questions about each of the four target group members. This factor is not, therefore, discussed.

810 L. Johnston

Table 1. received (Experiment 1)

Conditional probabilities for question choice as a function of the previous answer

~~

Trait topic

Previous answer Stereotypic Counter-stereotypic Neutral

Stereot ype-consistent Stereotype-inconsistent Stereotype-neutral

0.347 0.325 0.330 0.453,,, 0.281, 0.186,d 0.253, 0.360 0.364,

Nofore: Items with a common subscript differ significantly from one another.

asked, was calculated for each subject3. These conditional probabilities were then subjected to a 3 (previous answer received: stereotype-consistent/stereotype- inconsistent/stereotype-neutral) x 3 (trait topics: stereotypic/counter-stereotypic/ neutral) within-subjects ANOVA. There was a significant previous answer by trait type interaction, F(4,136) = 3.55, p < 0.01, means in Table 1. Simple main effects analysis was computed. There was an effect of trait topic only after a stereotype- inconsistent answer, F(2,68) =4.61, p < 0.05. After an inconsistent answer the probability of selecting the next question about a stereotypic trait was greater than that of selecting a question about either a counter-stereotypic or neutral trait (Tukey, p < 0.05; Ms = 0.453, 0.28 1 and 0.186 respectively). There was no effect on trait topic choice of receiving either a stereotype-consistent or a neutral answer. There were also main effects of the previous answer received on the probability of selecting a question about a stereotypic, F(2,68) = 4.75, p < 0.05, or a neutral, F(2,68) = 4.65, p < 0.05, trait topic. The probability of selecting a stereotypic trait topic was higher after a stereotype-inconsistent than a stereotype-neutral answer (Tukey, p < 0.05; Ms = 0.453 and 0.253), no other differences reaching significance. The probability of choosing a neutral trait was higher after a stereotype-neutral answer than a stereotype-inconsistent answer (Tukey, p < 0.05; Ms = 0.364 and 0.186), no other differences reaching significance. There was no effect of the nature of the previous answer received on the probability of selecting questions about counter-stereotypic traits.

Perceivers in the free, information-gathering conditions were sensitive to the nature of the information they received. Perceivers' information-gathering strategies were a function of both the hypothesis under consideration and the evidence received (Evett et al., 1994). After receiving a stereotype-inconsistent answer subjects were likely to ask the next question about a stereotypic trait (Pyszczynski & Greenberg, 1981). Selection of a stereotypic trait topic was, in fact, most likely after a stereotype- inconsistent answer.

'These analyses were also conducted using only two categories of answer, stereotype-consistent and stereotype-inconsistent. In this classification any question answered in an expectancy-mismatching manner was deemed as stereotype-inconsistent, even when it referred to stereotype-neutral traits. The results of this analysis were consistent with those reported in the text. Overall, though, it was felt that thc classification reported in the text was the more appropriate, since even when neutral questions are answered in an unexpected manner they are still stereotype-neutral.

Infbrmation seeking and stereotype change 8 1 1

Trait ratings

An index of stereotypicality was created for each subject by subtracting the mean rating for the counter-stereotypic traits from that for the stereotypic traits4. The higher the index, the more stereotypical the trait ratings. A single-factor (presentation mode: forced/free ~ no time pressure/free - high time pressure/ baseline) between-subjects ANOVA revealed a marginally significant effect of presentation mode, F(3,55) = 2.37, p < 0.0X. Planned contrasts between presentation conditions revealed a significantly lower index of stereotypicality in the forced than free conditions, F(1,55) = 5.96, p < 0.01 ( M s = 1.66 and 3.14); a significantly lower index of stereotypicality in the forced than baseline condition, F( 1 3 5 ) = 4.61, p < 0.05 (Ms = 1.66 and 3.48) but no difference between the free and baseline conditions, F < 1 . Results are shown in Figure 2. Note that, although the proportion of stereotype-inconsistent answers received by subjects in the free presentation conditions varied as a function of the actual questions subjects selected, the proportion received which was stereotype-inconsistent was very similar to that in the forced condition, fixed at 0.167 (Ms = 0.160 in the no time pressure, 0.190 in the high time pressure condition).

The results of the free information presentation conditions demonstrated that simply receiving stereotype-inconsistent information is not sufficient for moderation of stereotype-based judgments of the group (Evett et al., 1994). Subjects in the free, information-gathering conditions received a similar proportion of stereotype- inconsistent information to subjects in the forced, information-given condition in which moderation of the stereotype-based judgments was seen relative to baseline conditions. The proportion of stereotype-inconsistent information received in the information-gathering conditions was also comparable to that of previous studies demonstrating stereotype change in response to disconfirming information (Hewstone et al., 1992; Johnston & Hewstone, 1992; Johnston et al., 1994; Weber & Crocker, 1983). The same proportion of stereotype-inconsistent information had less effect on stereotype-based judgments when subjects controlled the information they received, than when subjects were forced to process all the available information. Possible explanations for this effect are considered in the General Discussion section.

The results of Experiment 1 supported some of the predictions. Perceivers in the free, information-gathering condition were indeed strategic in their information search, despite controlling for diagnosticity, showing a preference for information relating to stereotypic topics (i.e. a stereotype-preservation bias). This information- seeking was reflected in the trait rating task where there was no moderation of stereotype-based judgments. However, the same pattern of information-seeking would have resulted from a positivity bias. For the doctors target group all the stereotypic traits were positive and the counter-stereotypic traits negative. The bias toward stereotypic traits may, then, have been due not to a stereotype-preservation strategy but to a positivity bias. Experiment 2 addressed this possibility using a negative target group.

4There were no effects of processing condition on the ratings of the stereotype-neutral characteristics in any of the reported experiments and so these traits are not considered further.

812 L. Johnston

x v .- 3 3 - V a x 0

.- Y

2 2 -

rz Index of Slereotypicality

--O- Haseline

1 1 I Forced Free - ntp Free - htp

Presentation Mode

Figure 2. Index of stereotypicality as a function of presentation condition (Experiment 1). Free - ntp, free no time pressure information presentation mode; Free ~ htp, free high time pressure information presentation mode

EXPERIMENT 2

Overview

Experiment 2 was conceptually similar to Experiment 1, except that the target group was one with a negative stereotype (people with schizophrenia). Subjects were again presented with information via a bulletin board under either forced or free presentation conditions. After information gathering, subjects rated the target group on a number of stereotypic and counter-stereotypic dimensions. A number of procedural changes were made from Experiment 1. There was less information available to subjects (12 items) and the question type (stereotype- matching/stereotype-mismatching) and time pressure manipulations were dropped.

Method

Subjects and design

Fifty-nine undergraduates participated in return for payment. The experiment has a 2 (presentation mode: forced/free) x 3 (trait topic: stereotypic/counter-stereotypic/ stereotype-neutral) design, the first factor being a between-subjects and the second factor a within-subjects factor.

Informution seeking and stereotype change 8 13

Stimulus materials

The target group for this experiment had to be one for which there was an easily identifiable negative stereotype and for which characteristic behaviours could be identified. People with schizophrenia was selected as a group which met these two criteria. The key features of the schizophrenic stereotype were taken from Neuberg and Fiske (1987). Ten control subjects rated the target group on 20 personality characteristics, including those identified in Neuberg and Fiske (1987), (1 -‘not at all characteristic’, 5 - ‘unrelated to group’, 9 - ‘extremely characteristic’). From these ratings stereotypic, counter-stereotypic and stereotype-neutral traits were identified, as in Experiment 1. Four of each type of trait were chosen for use in the experiment proper: stereotypic: dangerous, suspicious, nervous, unpredictable; counter-stereotypic: assertive, adaptable, determined, sociable; stereotype-neutral: honest, sporting, generous, humorous. An ANOVA on the trait ratings revealed a significant effect of trait type, F(2,18) = 9.90, p < 0.01 (Ms = 6.35, 3.96 and 5.13 for the stereotypic, counter-stereotypic and neutral traits respectively). Again the stereotypic and counter-stereotypic traits were equally deviant from the scale midpoint ( M s = 1.35 and 1.04) suggesting that their strength of association with the target group was equally strong. An additional 10 subjects then rated the target group on the 12 selected traits (1 --‘not at all characteristic’, 7-‘very characteristic’) in order to provide baseline data against which the ratings of the experimental subjects could be compared for evidence of actual stereotype change.

For each of the trait topics an example of behaviour illustrating that trait was needed for use in the experiment proper. These behaviours were said to have been performed by a member of the target group. A number of behaviours were generated by the experimenter and assistants. Ten subjects rated how good an example of the trait the behaviours were (1 - ‘very poor example’, 9 - ‘very good example’) and how high a level of that trait the behaviour demonstrated (1 -‘very low level of trait’, 9-‘very high level of trait’). Ten subjects rated how useful the behaviour would be in determining the level of the given trait possessed by the person answering. Subjects were told to give a higher rating when the question would more strongly distinguish between those with high and low levels of the trait (1 -‘not at all useful’, 9- ‘extremely useful’). Subjects were then asked for feedback about their ratings and how behaviours could be changed in order to increase the ratings. After these changes were made additional groups of eight subjects rated the behaviours as above. For each of the 12 experimental traits a behaviour which was considered both a good example of the trait (mean rating greater than 7) and was considered useful (mean rating greater than 7) was identified and used in the experiment proper. There were no differences in the ratings for sentences across trait topics. For each of the trait topics two of the traits were described by consistent behaviours, that is behaviours indicating a high level of the trait (mean ratings greater than 7) and two by inconsistent behaviours, that is behaviours indicating a low level of the trait (mean ratings less than 2).

For the stereotype-relevant traits (i.e. not including the stereotype-neutral traits) half of the information was stereotype-consistent (two stereotypic traits described by consistent behaviours and two counter-stereotypic traits described by inconsistent behaviours) and half stereotype-inconsistent (two stereotypic traits described by inconsistent behaviours and two counter-stereotypic traits described by consistent

814 L. Johnston

behaviours). Across the set of behaviours, then, one-third of the information was stereotype consistent, one-third was stereotype inconsistent and one-third was stereotype neutral. The stereotype-inconsistent answers were evenly distributed across target group members (each gave one inconsistent answer) and trait topic (two inconsistent answers for each of the sterotypic and counter-stereotypic trait topics).

Listed below are examples of each type of behaviour:

Stereotypic trait topic (threatening), consistent behaviour: Occasionally John does have to be heavily sedated and kept in his room for his own safety and that of others. Stereotypic trait topic (anxious), inconsistent behaviour: Generally Pete is quite calm, very little hassles him. He usually copes very well with new places or people. Counter-stereotypic trait topic (sociable), consistent behaviour: Mark doesn’t have very many close friends. He doesn’t often go out with others, he especially doesn’t like large groups or parties. Often he’ll go to the cinema alone.

.Counter-stereotypic trait topic (assertive), inconsistent behaviour: When he returned some ill-fitting jeans to the shop John refused to be fobbed ofJ by the sales assistant. He demanded his money back.

.Neutral trait topic (generous), consistent behaviour: Mark is happy to share his food, drinks, records etc, with others. Neutral trait topic (sporting), inconsistent behaviour: Pete doesn’t play any sport himself and will not watch it on the television.

Procedure

The procedure was similar to that followed in Experiment 1. The major difference was that rather than two questions for each trait there was only one statement. This resulted in some modifications to the bulletin board as represented in Figure 1; screens 1 and 2 and screens 3 and 4 were merged.

Results and discussion

Information seeking

To test for an information-seeking bias the proportion of total questions asked which came from each trait topic was calculated for each subject. These proportions were subjected to a single factor (trait topic: stereotypic/counter-stereotypic/neutral) within-subjects ANOVA. There was a significant effect of trait topic, F(2,56) = 5,02, p < 0.01. Post-hoc tests (Tukey, p < 0.05) showed the highest proportion of questions asked to be from stereotypic trait topics and the lowest from neutral trait topics with the counter-stereotypic topics in between, all differences reaching significance (Ms=0.381, 0.289 and 0.340 respectively). In addition 0.733 of the first questions asked were about a stereotypic trait topic. As predicted, then, perceivers did show a stereotype preservation bias in their information-seeking about a negative target group.

Injbrmation seeking and stereotype change 8 15

Information-seeking strategies

Evidence of strategic information search was again assessed by considering the impact of different types of answer to previous questions on perceivers’ subsequent question choice. The probability of choosing a question from each topic type, given a stereotype consistent, inconsistent or neutral answer to the previous question asked, was calculated for each subject. These conditional probabilities were then subjected to a 3 (previous answer received: stereotype-consistent/stereotype-inconsistent/ stereotype-neutral) x 3 (trait topics: stereotypic/counter-stereotypic/neutral) within- subjects ANOVA. There was a significant previous answer by trait topic interaction, F(4,112) = 8.1 1, p < 0.0001, means in Table 2. Simple main effects analysis was conducted. There was a significant effect of trait topics after a stereotype- inconsistent answer, F(2,56) = 3.92, p < 0.05. Follow-up tests revealed that after a stereotype-inconsistent answer the probability of choosing a counter-stereotypic item was lower than that of choosing either a stereotypic or a neutral item ( M s = 0.220, 0.360 and 0.420 respectively). There were also significant effects of the previous answer on the probability of choosing both a counter-stereotypic trait topic, F(2,56) = 5.59, p < 0.01, and a stereotypic trait topic, F(2,56) = 7.02, p < 0.01. Follow- up tests (Tukey, p < 0.05) revealed that perceivers were significantly less likely to choose a counter-stereotypic trait topic if the previous answer was stereotype- inconsistent than if it was either stereotype-consistent or stereotype-neutral ( M s = 0.220, 0.408 and 0.419 respectively). Similarly, subjects were more likely to select a stereotypic trait after receiving a stereotype-inconsistent or stereotype- neutral than a stereotype-consistent answer ( M s = 0.360, 0.428 and 0.161 respectively).

The nature of information received had an impact on subjects’ information- seeking strategies. After receiving stereotype-inconsistent information perceivers were less likely to select their next piece of information from a counter-stereotypic topic than from either a stereotypic or neutral topic. Choosing a counter- stereotypic trait topic was less likely after a stereotype-inconsistent answer than after either a consistent or neutral answer and selecting a stereotypic trait was more likely after an inconsistent than a consistent answer. This trait selection strategy is again consistent with a motivation to maintain existing stereotypic perceptions of the target group.

Table 2. received (Experiment 2)

Conditional probabilities for question choice as a function of the previous answer

Trait topic ~~ ~

Previous answer Stereotypic Counter-stereotypic Neutral ~~ _ _ ~ ~ ~~ -~

Stereotype-consistent Stereotype-inconsistent Stereot ype-neutral

0.195,f 0.408, 0 391 0.360,, o.220qhcd 0.4201, 0.428, 0.419, 0.118

Note. Items with a common subscript differ significantly from one another.

816 L. Johnston

Information seeking and stereotype change 33

+ Index o f Stereotypicality

--O- Baseline

Presentation Mode

Figure 3. Index of stereotypicality as a function of presentation condition (Experiment 2)

Trait rat ings

An index of stereotypicality was again created for each subject by subtracting the mean rating for the counter-stereotypic traits from that for the stereotypic traits. A single-factor (presentation mode: forced/free/baseline) between-subjects ANOVA again revealed a marginally significant effect of presentation mode, F(2,65) = 2.5 1, p < 0.08. Planned contrasts between presentation conditions revealed a significantly lower index of stereotypicality in the forced than free conditions, F( 1,65) = 4.06, p < 0.05 ( M s = 0.741 and 1.57); a significantly lower index of stereotypicality in the forced than baseline condition, F(1,65) = 3.89, p < 0.05 ( M s = 0.741 and 1 .88) but no difference between the free and baseline conditions, F < 1 . These results are shown in Figure 3. In the forced presentation condition the proportion of the information received which was stereotype inconsistent was fixed at 0.333; in the free presentation condition this proportion was 0.335.

As in Experiment 1 the results of the free information presentation condition demonstrated that simply receiving stereotype-inconsistent information is not sufficient for moderation of stereotype-based judgments of the group. Subjects in the information-gathering condition received a similar proportion of stereotype- inconsistent information to subjects in the forced, information-given condition in which moderation of stereotype-based judgments was seen. Possible explanations of this effect will be considered in the General Discussion section.

Information seeking and stereotype change 8 17

Together the results of Experiments 1 and 2 demonstrate a bias in information- seeking toward stereotypic topics, regardless of whether the stereotype is positive or negative. This strategic information gathering is consistent with a motivation to maintain existing stereotypic perceptions of the target group. Experiment 3 further investigates the role of motivational factors in information search by varying the salient processing goals under which perceivers select information about a target group.

EXPERIMENT 3

Overview

In Experiment 3 the impact of processing goals on perceivers’ information search and subsequent impressions of the target group were assessed. Specifically, processing goals which conflict with stereotype maintenance were employed. Under such conditions it was predicted that subjects would not show a stereotype preservation bias and would show moderation in their stereotypic beliefs. Accuracy instructions had no impact on information search with experimenter-created expectancies (Snyder & Swann, 1978; Snyder et al., 1982), but when considering preexisting stereotypic beliefs it is predicted that perceivers asked to be accurate in their impressions will not show a stereotype-preservation bias in their information search (Kruglanski, 1989; Pyszczynski & Greenberg, 1987). Anticipating having to justify one’s opinions may motivate perceivers to increase accuracy (Kruglanski 62 Freund, 1983; Tetlock, 1983; Tetlock & Kim, 1987) in response to a desire to be seen positively by the justification group. To whom perceivers expect to justify their judgements will be an important factor. When perceivers anticipate justifying their opinions to a group who share their positive stereotype of the target group (e.g. a doctor’s support group) there is no conflict between the desire to be seen positively by the justification group and the goal of stereotype maintenance. When perceivers anticipate justifying their opinions to a group who do not share their stereotype of the target group (e.g. a patient action group) there is a conflict between the desire to be seen positively by the justification group and the goal of stereotype maintenance. The accuracy and justification, negative group instructions, produce a conflict with stereotype maintenance and so no stereotype preservation bias is predicted. The justification, positive group instructions do not, however, produce any conflict and so stereotype preservation is predicted in this condition.

Method

Szlhj~cts and design

Fifty-two psychology undergraduates participated in the experiment in return for course credit. The experiment had a 3 (processing goal: accuracy/justification ~

positive group/justification -negative group) x 3 (trait topics: stereotypic/counter- stereotypic/stereotype-neutral) x 2 (questions: stereotype-matching/stereotype- mismatching) mixed model design, the first factor being a between-subjects factor and the second and third factors within-subjects factors.

818 L. Johnston

Stimulus materials

The same stimulus materials as for Experiment 1 were used. In addition to the instructions presented at the beginning of the experiment subjects also received one of three processing goal instruction sets prior to starting the computer task. In the accuracy condition, subjects were instructed to try to be as accurate in their perceptions as possible. They were told that a monetary prize would be awarded to the person whose judgments of the target group best met those of experts who, allegedly, had received the same information and made the same judgments as the subjects would. In fact the prize winner was randomly selected. In the two justification conditions subjects were told that after they had given their perceptions of the target group they would be asked to justify those perceptions. These justifications would then be used in the preparation of a report discussing current perceptions of doctors. A microphone and tape recorder to be used for this justification stage of the experiment were visible to subjects throughout the experiment. For the justification, positive group condition subjects were told that their justifications were to be used by a doctors’ support group. For the justification, negative group condition subjects were told that their justifications were to be used by a patient pressure group. Pre-testing showed students to believe that the doctors’ support group, but not the patient pressure group, would endorse the positive stereotype of doctors identified in the pilot testing described in Experiment 1.

Procedure

Subjects were randomly allocated to one of the three experimental conditions and tested individually. On entering the laboratory subjects were told that they were participating in a social perception experiment concerned with how different social groups are seen by others. They were told that the study they were taking part in was concerned with perceptions of doctors. Subjects were then given the processing goal instructions for their experimental condition. The procedure was then the same as the free presentation, no time pressure mode of Experiment 1. Subjects in the justification conditions were not actually required to justify their opinions but were debriefed with regard to the purpose of this deception.

Results and discussion

Information seeking

To test for information-seeking biases the proportion of total questions asked which came from each trait topic was calculated for each subject. These proportions were subjected to a 3 (processing goal: accuracy/justification, positive group/justification, negative group) x 3 (trait topics: stereotypic/counter-stereotypic/stereotype- neutral) x 2 (questions: stereotype-matching/stereotype-mismatching) ANOVA. There was a significant processing goal by trait topic interaction, F(4,78) = 5.68, p < O . O O l , means in Table 3. Simple main effects analyses were conducted. Most importantly, there was an effect of trait topic only in the justification, positive group

Information seeking und stereotype change 8 19

Table 3. topic and processing goals (Experiment 3)

Proportion of questions asked and index of sterotypicality as a function of trait

Trait topic Index of

Processing goal Stereotypic Counter-stereotypic Neutral stereotypicality

Accuracy 0.316, 0.320 0.364 1.53,f Justification ~~ positive 0.486,,, 0.1 so,, 0.342 3.05,,

Justification -- negative 0.280b 0.482, 0.240 1.42, group

group Baseline 3.48,,

Note. Items with a common subscript differ significantly from one another.

condition, F(2,78) = 3.03, p < 0.05. A significantly higher proportion of the questions asked in this processing condition came from stereotypic than from counter stereotypic trait topics (Tukey, p < 0.05, Ms = 0.486 versus 0.1 SO), no other differences reaching significance. Only perceivers in the justification, positive group condition showed a stereotype-preservation bias in their information- seeking, as predicted. Subjects in the other two processing conditions selected equal proportions of questions about stereotypic and counter-stereotypic traits. The confirmatory bias in the justification, positive group condition was further demonstrated by the significant effect of processing goal on the proportion of questions asked about stereotypic traits, F(2,78) = 5.28, p < 0.01. Post-hoc tests (Tukey, p < 0.05) showed that the proportion of questions asked about stereotypic traits was greater in the justification, positive group condition than in either the accuracy or justification, negative group condition (Ms = 0.486, 0.280 and 0.316 respectively). In addition only in the justification, positive group condition was there a greater than chance preference for any trait topic on the first question asked (t-test, p < 0.05; Ms = 0.583 versus 0.333).

When there was a conflict between the salient processing goal and stereotype maintenance (i.e. accuracy and justification, negative group conditions) there was no evidence of the stereotype preservation bias seen in Experiments 1 and 2, as predicted. Information-seeking strategies are indeed sensitive to processing conditions. When there was no conflict between the processing goal and stereotype maintenance (i.e. the justification, positive group condition) there was, again, evidence of a stereotype-preservation bias.

Information-seeking strategies

Since a stereotype-preservation bias was only seen in the justification, positive group condition, evidence of strategic information search was only assessed in this condition. The impact of different types of answer to previous questions on perceivers’ subsequent question choice was considered. The probability of choosing a question from each topic type, given a stereotype-consistent, -inconsistent or -neutral answer to the previous question asked, was calculated for each subject. These

820 L. Johnston

conditional probabilities were then subjected to a 3 (previous answer received: stereotype-consistent/stereotype-inconsistent/stereotype-neutral) x 3 (trait topics: stereotypic/counter-stereotypic/neutral) within-subjects ANOVA. There was only a significant main effect of trait topic, F(2,22) = 6.82, p < 0.005. Post-hoc tests (Tukey, p < 0.05) showed the probability of selecting a stereotypic trait topic to be greater than that of selecting a counter-stereotypic trait topic ( M s = 0.463 versus 0.141), as in the information-seeking analysis above. Although the previous answer by trait topic interaction was not significant, a planned contrast revealed a significant difference in the probability of selecting a question from a stereotypic and counter-stereotypic trait topic after a stereotype-inconsistent answer, F(1,44) = 6.66, p < 0.01. After an inconsistent answer perceivers were more likely to select the next question from a stereotypic than counter-stereotypic trait topic (Ms = 0.444 versus 0.028).

Perceivers in the justification, positive group condition showed a stereotype- preservation bias in their information-seeking. The conditional probability data showed little evidence of strategic information-seeking in response to the information received. After an inconsistent answer, however, perceivers were more likely to select the next question from a stereotypic than a counter-stereotypic trait topic, as seen in Experiments 1 and 2, and consistent with the motivation to maintain existing stereotypic beliefs.

Trait ratings

An index of stereotypicality was again created for each subject by subtracting the mean rating for the counter-stereotypic traits from that for the stereotypic traits. A single- factor (processing goal: accuracy/justification, positive group/justification, negative group/baseline) between-subjects ANOVA revealed a significant effect, F(3,47) = 4.67, p < 0.01. Post-hoc tests (Tukey, p < 0.05) showed the index to be significantly lower in the accuracy and justification, negative group conditions than in the baseline and justification, positive group conditions ( M s = 1.53 and 1.42 versus 3.48 and 3.05). These results are shown in Table 3. The mean proportion of answers received which were stereotype-inconsistent were 0.248,0.261 and 0.205 for the accuracy, justification, positive group and justification, negative group conditions, respectively.

As predicted, in conditions where there was no stereotype preservation bias in information-seeking (accuracy and justification to a negative group), there was moderation of stereotype-based judgements of the group. The results for the justification, positive group condition matched the free conditions of Experiments 1 and 2; the bias in information-seeking was reflected in the trait ratings which showed no change from baseline conditions.

GENERAL DISCUSSION

The results from all three reported experiments demonstrate the important role of information-seeking strategies in the stereotyping domain, even when controlling for diagnosticity. Perceivers show a preference for information relating to stereotypic topics over information relating to counter-stereotypic topics. This preference is

Injormution seeking and stereotype change 82 1

particularly pronounced after perceivers have received stereotype-inconsistent information. The reported information-seeking bias is consistent with a motivation to maintain existing stereotypic beliefs about a target group. A bias toward stereotypic information is consistent with both the motivation to maintain stereotypic beliefs and with the preservation of cognitive resources (Fiske & Taylor, 1984, 1991). When the salient processing goals in the information-gathering situation conflicted with the maintenance of existing stereotypes, however, the stereotype- preservation bias disappeared and was replaced by more balanced information search.

This information search had consequences for stereotype-based judgments of the target group. The adoption of a stereotype-preservation bias by perceivers in free, information-gathering conditions was reflected in stereotype maintenance. There was no moderation of stereotype-relevant trait ratings relative to a baseline, no information condition. Subjects in the forced and free presentation conditions received very similar proportions of stereotype-inconsistent information and yet only those in the information-given condition showed a moderation of stereotype-based judgments. Although perceivers in the forced condition did receive more items of information, including more items of stereotype-inconsistent information, than perceivers in the free conditions it is unlikely that the difference in stereotype moderation is due to the difference in the absolute number of inconsistent items received. Stereotype moderation in response to stereotype-inconsistent information in forced conditions has been demonstrated across a large range of absolute amounts of information (e.g. Weber & Crocker, 1983)5. This finding illustrates the importance of the view of hypothesis testing as a multistage process that involves not just selecting questions to ask but also interpreting information (answers to the questions asked) and drawing inferences from it (Evett et al., 1994; Slowiaczek et al., 1992).

The bias toward stereotypic topics was most pronounced after the receipt of stereotype-inconsistent information suggesting that receiving such information does not change perceivers' expectancies regarding the answers to questions they select. Rather, the inconsistent evidence must be explained away somehow in order to maintain both those expectancies and pre-existing stereotype-based beliefs. There are a number of possible explanations for the differential impact of disconfirming information across conditions. Perceivers may attach more weight and importance to hypothesis-supporting answers than to hypothesis-opposing answers (Slowiaczek et al., 1992). It is possible that the inconsistent information was processed more thoroughly in the forced than free conditions. Perceivers in the free, information- gathering conditions may be more ready to dismiss inconsistent information. Evett et al. (1994) for example, showed information seekers to make situational attributions for inconsistent but dispositional attributions for consistent information. Situational attributions show less generalization across targets and situations than do dispositional attributions and so the impact of the stereotype discrepant information is weakened as a consequence of the situational attribution. The ability of subjects in the free conditions to choose a stereotypic topic as the next item may also reduce the impact of the stereotype-inconsistent information received. The

'Yoking perceicers under forced conditions to those in free, information-gathering conditions so that both receive exactly the same information would be a more precise way to demonstrate this.

822 L. Johnston

ability to jump between targets collecting information may also have resulted in perceivers in the free condition dismissing targets who provided stereotype- discrepant information. Perceivers in the forced conditions had to work through all the available information about each target member, simply dismissing one after a piece of inconsistent information was not possible and so subjects in this condition may have tried harder to integrate all the available information into the overall perception of the target group (and hence resulting in stereotype moderation). Although not explicitly instructed to do so, subjects in the forced condition tended to work through the information about each target in turn whilst subjects in the free conditions jumped between targets. This mode of processing may simply have made the disconfirmation more salient to perceivers (Hewstone, Macrae, Griffiths, Milne & Brown, 1994; Experiment 2). These possibilities await future research.

The role of information-seeking strategies in stereotype change may help to explain why stereotypes continue to persist in everyday settings despite the success of researchers in demonstrating stereotype change in the laboratory (e.g. Gurwitz & Dodge, 1977; Hewstone et al., 1992; Johnston & Hewstone, 1992; Johnston et al., 1994; Weber & Crocker, 1983). Simply receiving disconfirming information is not, however, sufficient to cause stereotype change (Johnston & Macrae, 1994). Stereotype moderation effects in naturalistic information-gathering settings appear to be contingent upon having salient processing goals (e.g. accuracy, negative justification) which conflict with stereotype maintenance and hence require perceivers to seek information about both stereotypic and counter-stereotypic characteristics of the target group. The ease with which processing goals can overcome information-seeking biases may be a function of how strongly the target group’s stereotype is held (Trope & Bassok, 1983) and how affectively laden the target group is (Pyszczynski & Greenberg, 1987).

ACKNOWLEDGEMENTS

This research was conducted when the author was at the School of Psychology, University of Wales College of Cardiff. The financial assistance of that department in funding the research project is gratefully acknowledged. Thanks are extended to Simon Handley who developed the bulletin board methodology used in the research and collected the data for Experiments 1 and 3; to Louise Giles who collected the data for Experiment 2; to Trish Devine, Garth Fletcher, Steve Hudson, Arie Kruglanski, Neil Macrae, Yaccov Trope, anonymous reviewers and especially Vance Locke for their insightful comments on an earlier version of the manuscript.

REFERENCES

Abrams, D., & Hogg, M. A. (1985). Comments on the motivational status of self-esteem in social identity and intergroup discrimination. European Journal of Social Psychology, 18, 3 17-334.

Bassok, M., & Trope, Y. (1984). People’s strategies for testing hypotheses about another’s personality: Confirmatory or diagnostic? Social Cognition, 2, 199-216.

Beyth-Marom, R., & Fischhoff, B. (1 983). Diagnosticity and pseudodiagnosticity. Journal of Personality and Social Psychology, 45, 1185-1 195.

Infkrmntion seeking and stereotype chungr 823

Bodenhausen, G. V. (1 990). Stereotypes as judgmental heuristics: Evidence of circadian variations in discrimination. Psychological Science, 1, 3 19-322.

Hodenhausen, G. V., Kramer, G. P., & Susser, K . (1994). Happiness and stereotypic thinking in social judgments. Journal of Personality and Social Psychology, 66, 621-632.

Bodenhausen, G. V., & Lichtenstein, M. (1987). Social stereotypes and information- processing strategies: The impact of task complexity. Journal of’ Personulity and Social

Bodenhausen, G. V., & Wyer, R. S. (1985). Effects of stereotypes on decision making and information processing strategies. Journal Of‘Personulity and Social Psychology, 48,267-282.

Brewer, M. B. (1988). A dual process model of impression formation. In T. K. Srull & R. S. Wyer (Eds), Advances in social cognition (Vol. 1, pp. 1-36). Hillsdale, NJ: Erlbaum.

Darley, J. M., & Gross, P. H. (1983). A hypothesis-confirming bias in labelling effects. Journal of’ Personality and Social Psychology, 44, 20-33.

Darley, J. M., Fleming, J. H., Hilton, J. L., & Swann, W. B. (1988). Dispelling negative expectancies: The impact of interaction goals and target characteristics on the expectancy confirmation process. Journal of Experimental Social Psychology, 24, 19-36.

Devine, P. G., Hirt, E. R., & Gehrke, E. M. (1990). Diagnostic and confirmation strategies in trait hypothesis testing. Journal of Personality and Social Psychology, 58, 952-963.

Doherty, M. E., Mynatt, C. R., Tweney, R. D., & Schiavo, M . D. (1979). Pseudodiagnosticity. Acta Psvchologica, 43, 1 1 1-1 2 I .

Evett, S. R., Devine, P. G., Hirt, E. R., & Price J. (1994). The role of the hypothesis and the evidence in the trait hypothesis testing process. Journal of Experimental Social Psychology,

Fazio, R. H., Sherman, S. J., & Herr, P. M. (1982). The feature-positive effect in the self- perception process: Does not doing matter as much as doing? Journal of’ Personality and Social Psychology, 42, 404-41 I .

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. Zanna (Ed.), Advances in experimental social psychology (Vol 23, pp. 1-74). New York: Academic Press.

PsJ)chology, 52, 871--880.

30, 456-48 I .

Fiske, S. T., & Taylor, S. E. (1984). Social cognition. New York: McGraw-Hill. Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed.). New York: McGraw-Hill. Forgas, J. P. (1989). Mood effects on decision making strategies. Australian Journal of’

Gilbert, D. T., & Hixon, J. G. (1991). The trouble of thinking: Activation and application of stereotypic beliefs. Journal of Personality and Social Psychology, 60, 509-5 17.

Gurwitz, S. B., & Dodge, K. A. (1977). Effects of confirmations and disconfirmations on stereotype-based attributions. Journal of’ Personality and Social Psychology, 35, 495-500.

Hamilton, D. L., & Sherman, J. W. (1994). Stereotypes. In R. S. Wyer Jr. & T. K . Srull (Eds), Handbook of social cognition (2nd ed.). Hillsdale, NJ: Erlbaum.

Hewstone, M., Johnston, L., & Aird, P. (1992). Cognitive models of stereotype change (2): Perceptions of homogeneous and heterogeneous groups. European Journal of Social

Hewstone, M., Macrae, C. N., Griffiths, R., Milne, A. B., & Brown, R. (1994). Cognitive models of stereotype change: ( 5 ) . Measurement, development and consequences of subtyping. Journal of Experimental Social Psychology, 30, 505-526.

Higgins, E. T., & Bargh, J. A. (1987). Social cognition and perception. Annual R ~ v i e w qf’

Jenkins, H. M., & Sainsbury, R. S. (1969). The development of stimulus control through differential reinforcement. In N. J. Mackintosh & W. K. Honig (Eds), Fundamental issues in associative learning. Halifax, Nova Scotia, Canada: Dalhousie Univ. Press.

Jenkins, H. M., & Sainsbury, R . S. (1970). Discrimination learning with the distinctive feature on positive or negative trials. In D. Mostofsky (Ed.), Attention: Contemporary theory and anal.ysis. New York: Appleton-Century-Crofts.

Johnston, L., & Hewstone, M. (1992). Cognitive models of stereotype change (3): Subtyping and the perceived typicality of disconfirming group members. Journal of’ Experimental Social Psychology, 28, 360-386.

P.Yychology, 41, 197--2 14.

P~yychologj, 22, 23 5--249.

P.yychology, 38, 369--425.

824 L. Johnston

Johnston, L., & Macrae, C. N. (1994). Changing social stereotypes: The case of the information seeker. European Journal of Social Psychology, 24, 581-592

Johnston, L., Hewstone, M., Pendry, L., & Frankish, C. (1994). Cognitive models of stereotype change (4): Motivational and cognitive influences. European Journal of Social Psychology, 24, 237-267.

Kruglanski, A. W. (1989). Lay episternics and human knowledge. New York: Plenum. Kruglanski, A. W., & Freund, T. (1983). The freezing and unfreezing of lay-inferences: Effects

of impressional primacy, ethnic stereotyping and numerical anchoring. Journal of Experimental Social Psychology, 19, 448-468.

Locke, V., MacLeod, C., & Walker, I. (1994). Automatic and controlled activation of stereotypes: Individual differences associated with prejudice. British Journal of Social Psychology, 33, 2946.

Macrae, C. N., Hewstone, M., & Griffiths, R. J. (1993). Processing load and memory for stereotype-based information. European Journal of Social Psychology, 23, 76-87.

Macrae, C. N., Milne, A. B., & Bodenhausen, G. V. (1994). Stereotypes as energy-saving devices: A peek inside the cognitive toolbox. Journal of Personality and Social Psychology, 66, 3747.

Neuberg, S. L., & Fiske, S. T. (1987). Motivational influences on impression formation: Outcome dependency, accuracy-driven attention and individuating processes. Journal of Personality and Social Psychology, 53, 43 1-444.

Newman, J., Wolff, W. T., & Hearst, E. (1980). The feature-positive effect in adult human subjects. Journal of Experimental Psychology: Human Learning and Memory, 6,

Nisbett, R. E., & Ross, L. (1980). Human inference: Strategies and shortcomings of social judgment. Englewood Cliffs, NJ: Prentice Hall.

Pratto, F., & Bargh, J. A. (1991). Stereotyping based upon apparently individuating information: Trait and global components of sex stereotypes under attentional overload. Journal of Experimental Social Psychology, 27, 2647.

Pyszczynski, T. A,, & Greenberg, J. (1981). Role of disconfirmed expectancies in the instigation of attributional processing. Journal of Personality and Social Psychology, 40, 3 1- 38.

Pyszczynski, T. A,, & Greenberg, J. (1987). Toward an integration of cognitive and motivational perspectives on social influence: A biased hypothesis-testing model. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 20, pp. 297-340). San Diego, CA: Academic Press.

Ray, J . J. (1983). Reviving the problem of acquiescence response bias. Journal of Social

Skov, R. B., & Sherman, S. J. (1986). Information-gathering processes: Diagnosticity, hypothesis confirmatory strategies and perceived hypothesis confirmation. Journal of Experimental Social Psychology, 22, 93-- 121.

Slowiaczek, L. M., Klayman, J., Sherman, S. J., & Skov, R. B. (1992). Information selection and use in hypothesis testing: What is a good question and what is a good answer? Memory and Cognition, 20, 392405.

Snyder, M. (1981). Seek, and ye shall find: Testing hypotheses about other people. In E. T. Higgins, D. C. Herman, & M. P. Zanna (Eds), Social cognition: The Ontario Symposium on personality and social psychology (Vol. 1, pp. 277-303). Hillsdale, NJ: Erlbaum.

Snyder, M., & Campbell, B. H. (1980). Testing hypotheses about other people: The role of the hypothesis. Personality and Social Psychology Bulletin, 6, 421-426.

Snyder, M., Campbell, B. H., & Preston, E. (1982). Testing hypotheses about human nature: Assessing the accuracy of social stereotypes. Social Cognition, 1, 330-342.

Snyder, M., & Cantor, N. (1979). Testing hypotheses about other people: The use of historical knowledge. Journal of Experimental Social Psychology, 15, 330-342.

Snyder, M., & Swann, W. B. (1978). Hypothesis-testing processes in social interaction. Journal of Personality and Social Psychology, 36, 1202-1 212.

Tetlock, P. E. (1983). Accountability and complexity of thought. Journal of Personality and Social Psychology, 45, 74-83.

630-650.

Psychology, 121, 81-96.

Information seeking and stereotype change 825

Tetlock, P. E., & Kim, J. I. (1987). Accountability and judgment processes in a personality

Trope, Y., & Bassok, M. (1982). Confirmatory and diagnosing strategies in social information

Trope, Y., & Bassok. M. (1983). Information gathering strategies in hypothesis testing.

Trope, Y., Bassok, M., & Alon, E. (1984). The questions lay interviewers ask. Journal of

Wason, P. C. (1960). On the failure to eliminate hypotheses in a conceptual task. Quarterly

Wason, P. C. (1968). Reasoning about a rule. Quarterly Journal of Experimental Psychology,

Weber, R., & Crocker, J. (1983). Cognitive processes in the revision of stereotypic beliefs. Journal of Personality and Social Psychology, 45, 961-977.

Yzerbyt, V. Y . , & Leyens, J-P. (1991). Requesting information to form an impression: The influence of valence and confirmatory status. Journal of Experimental Social Psychology, 27,

Zuckerman, M., Knee, C. R., Hodgins, H. S., C? Miyake, K. (1995). Hypothesis confirmation: The joint effect of positive test strategy and acquiescence response set. Journal of Personality and Social Psychology, 68, 52-60.

prediction task. Journal of Personality and Social Psychology, 52, 700-709.

gathering. Journal of Personality and Social Psychology, 43, 22-34.

Journal of Experimental Social Psychology, 19, 560-576.

Personality, 52, 9&106.

Journal of Experimental Psvc'hology, 12, 129--140.

23, 273-281.

337-356.