Forgetting by remembering: Stereotype inhibition through rehearsal of alternative aspects of identity

Abstract

We applied previous research on retrieval-induced forgetting to the issue of stereotype inhibition. All participants learned about a target person who belonged to a stereotyped group, and then practiced retrieving a subset of the target’s characteristics. When participants practiced individuating information about the target, they showed inhibited memory for the target’s stereotypic traits. When participants practiced stereotypic information about the target, they showed inhibited memory for: (a) traits associated with another stereotyped aspect of the target’s identity; (b) individuating traits of the target; and (c) other, unpracticed traits of the target associated with the same stereotype. Stereotype belief moderated these inhibition effects; the more strongly participants believed in the stereotype, the less inhibition of stereotype-relevant traits they showed.

Introduction

Imagine a White female college student interacting with her Black male professor. Although the student may sincerely wish to avoid viewing the professor through the lens of racial stereotypes, her conscious attempts to avoid stereotypical thinking may backfire, particularly if her mental resources are partially occupied by other tasks (e.g., Wyer, Sherman, & Stroessner, 2000). Is the student doomed, then, to the inescapable influence of racial stereotypes in perceiving and interacting with the professor? Perhaps not. After all, the professor is not only an African-American, but also a professor—not to mention, a complex person with a range of individuating personality traits (e.g., an exceptional sense of humor). By focusing on the teacher’s professorial or individuating characteristics, then, the student may be able to successfully inhibit stereotypes about African-Americans.

A handful of recent studies provide initial evidence that such inhibitory processes may play an important role in stereotyping and person perception (Dijksterhuis and van Knippenberg, 1995, Dijksterhuis and van Knippenberg, 1996, Dijksterhuis and van Knippenberg, 1998; Hugenberg & Bodenhausen, 2001; Macrae, Bodenhausen, & Milne, 1995; Rudman & Borgida, 1995; Stroessner, 1996). For example, Macrae et al. (1995) showed participants a video of a Chinese woman either putting on makeup (priming the female stereotype) or eating noodles with chopsticks (priming the Chinese stereotype). In the lexical decision task that followed, participants who had seen the makeup version were faster to identify female-relevant words (e.g., emotional), but slower to identify Chinese-relevant words (e.g., calm), compared to unprimed control participants; the reverse pattern emerged for participants who had seen the chopsticks version. Macrae et al. assert that “both applicable categories (i.e., Chinese and woman) are activated in parallel and that a competition for mental dominance ensues,” leading to the excitation of the winner and the inhibition of the loser (p. 398).²

This research suggests that inhibition may be involved in the category-selection process that occurs when we initially encounter multiply-categorizable targets. However, what about the later stages of person perception when we have already learned about a person and need to remember information about the person’s characteristics—might similar inhibitory processes come into play then as well? If focusing on one stereotyped aspect of a person inhibits another stereotyped aspect, might focusing on one stereotyped aspect inhibit memory for individuating information about the person? More optimistically, might focusing on individuating information inhibit the stereotype? And what role would belief in such stereotypes play?

To examine these questions we borrowed a technique from the cognitive literature called the retrieval-induced forgetting paradigm (see Levy & Anderson, 2002, for a review). This technique provides a method for creating and examining inhibition in long-term memory. First, we describe the retrieval-induced forgetting paradigm and how it has been applied to person perception. Next, we describe important factors to consider in applying this paradigm to the stereotyping domain, including the potential role of stereotype belief. Then, we discuss mechanisms through which belief in the stereotype might moderate inhibition. Finally, we present three experiments that address the questions posed above.

The phenomenon of retrieval-induced forgetting (RIF) was first demonstrated by Anderson, Bjork, and Bjork (1994) and has been replicated many times using various kinds of stimuli (see Levy & Anderson, 2002, for a list; see also Bäuml, 2002; Bäuml & Hartinger, 2002; Macrae & Rosevare, 2002). The standard version of the retrieval-induced forgetting paradigm has four phases: a learning phase, a retrieval-practice phase, a distracter phase, and a test phase. For example, in Anderson, Bjork, and Bjork (1994), participants studied category–exemplar pairs from eight target categories, each of which had six exemplars (e.g., Fruit–orange; Hobbies–coins; Fruit–banana). Participants were told to study each pair by relating the exemplar to the category. Next, participants practiced retrieving half of the items from half of the target categories; they were shown a category name and the first two letters of an exemplar (e.g., Fruit–or__) and told to write down an item that they had studied earlier that would fit the word stem. After a distracter task, participants were shown the name of each previously studied category in turn and asked to write down all of the words they had studied from that category.

These manipulations create three kinds of items: practiced items from practiced categories (e.g., orange); unpracticed items from practiced categories (e.g., banana); and baseline items—that is, items from categories in which no items were practiced (e.g., coins). Not surprisingly, Anderson et al. (1994) found that practiced items were recalled better than baseline items. The interesting question, however, is what would happen to the unpracticed items like banana? Would these words be remembered better than the baseline items because of their association with the strengthened items from their category? Or would memory for them be impaired? In fact, memory for the unpracticed items was impaired relative to the baseline items. This phenomenon was called “retrieval-induced forgetting” because it results from the retrieval of competing items from memory.³

What is the mechanism behind this forgetting? Anderson and Spellman (1995) pointed out that although the previous results might be due to inhibition of the concept banana, the evidence is ambiguous. Because the impaired items (e.g., banana) were both studied and tested under the same cue (e.g., Fruit) as the strengthened items (e.g., orange), the impairment could be due to other, non-inhibitory, mechanisms such as blocking. According to the blocking explanation, the strengthened item orange repeatedly comes to mind when the test probe Fruit is presented. The representation of banana itself might not be inhibited; rather, access to it through the cue Fruit might be blocked by orange. (See Anderson & Neely, 1996, for a review of the possible non-inhibitory mechanisms.) Using a more complicated design, Anderson and Spellman demonstrated that the representations of the impaired items were, in fact, inhibited. (See also Anderson & Bell, 2001.)

According to Anderson and his colleagues, this form of inhibition occurs because when given the cue Fruit-or__ both orange and banana (plus all the other fruits studied) are likely to be initially activated. In order to retrieve the target orange from out of the activated noise, the other items that were activated need to be suppressed. As a result of this suppression, non-practiced Fruit words become difficult to remember when given the original Fruit cue or an alternative cue.

Social stimuli in retrieval-induced forgetting. Recent work by Macrae and MacLeod (1999) and MacLeod and Macrae (2001) illustrates that the phenomenon of retrieval-induced forgetting extends to some social stimuli. Participants were asked to form impressions of two men by studying a series of index cards, each of which listed one man’s name along with a single personality trait. Afterward, participants rehearsed half of the traits associated with one of the men. Following a distracter period, participants were asked to write down all of the traits that had originally been associated with each of the men. Consistent with previous research on RIF, participants: (a) exhibited enhanced memory for the traits they had practiced while (b) displaying inhibited memory for the unpracticed traits associated with the man whose other traits had been rehearsed.

In these experiments, each man was described by 10 traits that were unrelated to each other. For example, one man was described by the traits “sensible,” “romantic,” and “vigorous,” a combination that almost precludes forming a clear impression of him. In utilizing traits that do not hang together, the experimental design blunts the perceiver’s usual role in organizing the information based on pre-existing knowledge about the social world. While a perfectly appropriate strategy for a first investigation of RIF in the social domain, this work does not attempt to address how perceivers’ social knowledge interacts with the general phenomenon of retrieval-induced forgetting. For example, if one learns that a man is sensible and vigorous and then rehearses “sensible,” one may easily forget that an unrelated trait (i.e., “vigorous”) also describes the man. But would it be so easy to forget a trait such as “rational,” given that sensible people also tend to be rational? We would argue that such pre-existing knowledge about the social world may influence the extent to which people exhibit RIF in forming impressions of others.

Indeed, rather than defining others by a string of unconnected traits, people are likely to organize information about other individuals around important aspects of their identity. Because group membership often represents a central aspect of people’s identity (e.g., Tajfel & Turner, 1986), information about others may frequently coalesce around their membership in various social groups, information that, for better or worse, often takes the form of stereotypes. Thus, in applying RIF to person perception, it is useful to consider the special case of stereotypes. In the present series of studies, we examine the role of retrieval-induced forgetting in memory for stereotypical traits associated with a social group or group member, compelling a careful consideration of the special nature of social stereotypes.

Structural differences in stimuli. Importing the retrieval practice paradigm into the stereotype arena is not as straightforward as it may initially seem. Typical RIF experiments use category–exemplar pairs (e.g., Fruit–orange); the kind of pairs we are interested in using are category–feature pairs (e.g., Artist–creative). These pairings are not analogous: oranges are a type of fruit; creativity is a common trait (i.e., feature) among artists or belonging to particular artists. We know from previous research that category–exemplar pairs can demonstrate inhibition (e.g., Anderson et al., 1994); we also know that using social stimuli, individual–feature pairs can demonstrate inhibition (e.g., Macrae & MacLeod, 1999). We do not know, whether for social stimuli, category–feature pairs can demonstrate inhibition.

In applying previous RIF research to the issue of stereotype inhibition, it is crucial to consider how social knowledge interacts with the cognitive process of retrieval-induced forgetting. There are two findings in the cognitive RIF literature that suggest ways in which stereotype belief might moderate inhibition. These findings have to do with: (a) the strength of the links between the category and the exemplars and (b) the coherence or “integration” of a category’s exemplars. Extending the analogy from those two findings to research on stereotype inhibition leads to competing predictions about how belief will affect inhibition.

The strength of the links between the category and the exemplars. Clever cognitive experiments not withstanding, for most standard categories, most of the time, most people agree that membership is clear: oranges and bananas are fruits. But for stereotypes it is not so clear that most of the time most people agree that the stereotypic traits actually apply, for example, that Asian-Americans are intelligent and reserved. Although most people may be aware of common social stereotypes, belief in such stereotypes varies a great deal from individual to individual (Devine, 1989). The belief that a banana is a fruit is likely to vary little across individuals, but some fruits are fruitier than others; when asked to generate exemplars of the category fruits, participants are more likely to come up with apple, orange, and pear, than pomegranate, blueberry, and mango (exemplars 1, 2, and 3, and 21, 22, 23 from Battig & Montague, 1969). To the extent that people who strongly believe in a given stereotype tend to think of stereotype-relevant traits when faced with a member of that category, stereotype belief could be conceptualized as a set of strong links between the category and its stereotypic traits—just as the category Fruit is strongly linked to apples and other exemplars that come to mind when presented with this category.

The cognitive literature shows that exemplars that are strongly associated with their category suffer more RIF—both absolute and proportional—than weakly associated items; exemplars that are strongly associated with a category are likely to come to mind upon seeing the category and therefore must be strongly inhibited in order to promote successful recall of other items from the same category (Anderson et al., 1994; Bäuml, 1998). Thus, if belief in a stereotype strengthens the association between the relevant social category and its stereotypic traits, then higher levels of stereotype belief should be associated with greater retrieval-induced forgetting of these traits.

Coherence or “integration” of exemplars. Apart from strengthening the links between a category and its associated traits, stereotype belief may strengthen the links between the traits themselves via cross-item integration. Cross-item “integration” has to do with the extent to which each item is associated with and evokes thoughts of the others. For example, oranges and bananas have associations to each other beyond the fact that they are similar because they are both fruits: they are both breakfast foods, they might both be in your smoothie, and they both make good afternoon snacks. But traits like “intelligent” and “reserved” would not have those sorts of independent associations—until the label is provided. Indeed, stereotypic traits often seem to cohere only through their common association with the stereotype label.

Integrating items within a category has been shown to reduce RIF; if participants think back to other items associated with the same category whenever they see a new item, the unpracticed items from the category will show resistance to inhibition (Anderson & McCulloch, 1999). The reduction in RIF is shown both if the participants spontaneously choose to integrate items while studying or if the experimenter instructs participants to do so. Why does it happen? One possible explanation is that of “mediated retrieval.” When thinking back to previous items, connections are being created between the to-be-practiced and to-be-non-practiced items. Later, during the final recall test, such cross-item integration provides additional retrieval routes from the strengthened practiced items (which are likely to be recalled) to the unpracticed items from the same category, benefiting the latter relative to baseline items. If one embraces a stereotype, it may provide a powerful integrative framework. For example, if one really believes that Asian-Americans are studious, intelligent, reserved, diligent, educated, and precise, then remembering any of these words may bring the others back to mind because the traits are closely linked to each other by way of their common association with the stereotype. If stereotype belief strengthens the links between traits related to the stereotype in this manner, then people who believe a given stereotype should show less retrieval-induced forgetting of stereotypic traits.

In the studies that follow, we examined how RIF applies to social stereotypes. All of the present experiments use a similar four-phase procedure based on the methods of Anderson and colleagues described above. In the initial study phase, participants learn to associate trait words with several different people or groups. In the retrieval-practice phase, participants’ attention is focused on a subset of the information they have just learned. They are shown the name or description of one of the people or groups (“target”) and are asked to retrieve half of the associated traits from memory. After a distracter period, participants are asked to recall all of the trait words they saw during the initial study phase. Finally, participants fill out a questionnaire regarding their belief in the stereotypes addressed in the experiment.

In Experiments 1 and 2, we predicted that studying a compound category (e.g., Asian-American artist) and rehearsing stereotypic traits associated with one of the sub-categories (e.g., artist) would inhibit memory for traits associated with the other sub-category (e.g., Asian-American). Further, in Experiment 3, we predicted that practicing individuating traits would inhibit stereotypic traits, and vice versa. In this experiment, we also investigated whether practicing a sub-set of the characteristic traits of a stereotype would lead to inhibited memory for other unpracticed traits related to the same stereotype.

Throughout these studies, we examined the moderating role of stereotype belief in retrieval-induced forgetting of stereotype-relevant traits. If believing in a stereotype serves merely to strengthen the links between a group label and relevant traits, then strongly held stereotypes would be expected to suffer greater inhibition than weakly held stereotypes. If, however, believing in a stereotype serves to integrate stereotype-relevant traits, then strongly held stereotypes should prove relatively resistant to inhibition.