Scandinavian Journal of Psychology, 2012, 53, 239–246
DOI: 10.1111/j.1467-9450.2012.00939.x
Personality and Social Psychology Recognition bias and the physical attractiveness stereotype JEAN-CHRISTOPHE ROHNER and ANDERS RASMUSSEN Department of Psychology, Lund University, Sweden
Rohner, J.-C. & Rasmussen, A. (2012). Recognition bias and the physical attractiveness stereotype. Scandinavian Journal of Psychology 53, 239–246. Previous studies have found a recognition bias for information consistent with the physical attractiveness stereotype (PAS), in which participants believe that they remember that attractive individuals have positive qualities and that unattractive individuals have negative qualities, regardless of what information actually occurred. The purpose of this research was to examine whether recognition bias for PAS congruent information is replicable and invariant across a variety of conditions (i.e. generalizable). The effects of nine different moderator variables were examined in two experiments. With a few exceptions, the effect of PAS congruence on recognition bias was independent of the moderator variables. The results suggest that the tendency to believe that one remembers information consistent with the physical attractiveness stereotype is a robust phenomenon. Key words: Attitude, stereotype, memory, recognition-bias. Jean-Christophe Rohner, Department of Psychology, Lund University, Sweden. E-mail:
[email protected]
INTRODUCTION Stereotypes influence information processing in a number of specific ways (e.g. Macrae and Bodenhausen, 2000). In this paper, we examine recognition bias for stereotype-congruent and stereotype-incongruent information about physically attractive and physically unattractive individuals. Previous studies have shown that humans believe that they remember stereotype-congruent rather than stereotype-incongruent information, independently of whether they have encountered the information or not (e.g. Lenton, Blair & Hastie, 2001; MacRae, Schloerscheidt, Bodenhausen & Milne, 2002; Payne, Jacoby & Lambert, 2004; Sherman & Bessenoff, 1999; Stangor & McMillan, 1992). Our main aim was to study whether the effect of stereotype-congruence on recognition bias is robust across a variety of conditions; that is, does recognition bias for congruent information prevail despite individual differences and variable situational factors during encoding and retrieval? Understanding how stereotypes influence memory could help explain how stereotypes influence social behavior, since both spontaneous and planned behaviors often depend on memory (Bargh, 2005; Dawes, 1998; Dijksterhuis & Bargh, 2001; Prinz, 1997). For instance, attractive children and adults are treated more favourably than unattractive children and adults (e.g. Langlois, Kalakanis, Rubenstein, Larson, Hallam and Smoot, 2000). Selective memory may be an important mediator of the effect of stereotypes on behavior. Falsely believing that an unattractive person had negative but not positive traits could lead to hostile behavior towards the person. Stereotype-congruent (compared to incongruent) information could influence memory in two main ways: (1) by producing higher or lower recognition sensitivity, and/or (2) by producing more or less recognition bias. Measures of recognition sensitivity in tests of explicit memory indicate the ability to distinguish between events that occurred and those that did not, by asking participants to report their memories; recognition sensitivity is computed by subtracting false alarms (the proportion of new items that participants incorrectly identify as old) from hits (the proportion of old items that participants correctly identify as old).
Recognition bias in tests of explicit memory is the tendency to categorize both old and new items as old, independently of variation in recognition sensitivity. The present research focuses on recognition bias. Recognition bias for stereotype-congruent combinations can result for different reasons. First, participants may use a conscious guessing strategy, responding old more often on congruent items even if they are aware that they are uncertain of the correct answer (Payne et al., 2004). Secondly, a response bias may result because of misattributions of fluency (Jacoby, Woloshyn & Kelley, 1989; Kelley & Rhodes, 2002; Payne et al., 2004; Roediger & McDermott, 2000). According to this explanation, ease of processing is sometimes interpreted as a sign of conscious memory. For instance, participants may believe that a pair of semantically related test items (that are easy to process) were presented at study even if they were not (e.g. Rajaram & Geraci, 2000; Whittlesea, 1993). Stereotype-congruent information should be processed more fluently than incongruent information; when deciding whether congruent information was encountered previously or not, people may identify both old and new congruent information as old, interpreting fluency as a sign of conscious memory (Payne et al., 2004). In the present research we focus on the physical attractiveness stereotype (PAS), the tendency to attribute social skills, potency, adjustment and intellectual ability more easily to attractive than to unattractive individuals (Adams, 1982; Dion, Berscheid & Walster, 1972; Eagly, Ashmore, Makhijani & Longo, 1991; Feingold, 1992; Langlois et al., 2000). Rohner and Rasmussen (2011) examined recognition bias for PAS-congruent information (attractive + positive and unattractive + negative) and PAS-incongruent information (attractive + negative and unattractive + positive). Participants viewed several combinations in a study phase, each consisting of a face and a word. Faces varied in physical attractiveness and words varied in valence. In a subsequent test phase participants viewed studied combinations in addition to new combinations (created by re-pairing study combinations). The task was to respond which combinations had been shown previously and which combinations were new. Results indicated that
2012 The Authors. Scandinavian Journal of Psychology 2012 The Scandinavian Psychological Associations. Published by Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA. ISSN 0036-5564.
240 J.-C. Rohner and A. Rasmussen recognition bias was higher for PAS-congruent than for PASincongruent stimulus items. In other words, participants believed that they remembered PAS-congruent items to a greater extent than PAS-incongruent items, independently of whether they had experienced the information or not. This finding is consistent with the results of studies that have examined memory for other stereotypes (e.g. Lenton et al., 2001; MacRae et al., 2002; Payne et al., 2004; Sherman & Bessenoff, 1999; Stangor & McMillan, 1992). Stereotype-congruent information is connected with more recognition bias than incongruent information. Stangor and McMillan (1992) conducted a meta-analysis on the influence of social expectations on explicit memory for expectancy-congruent and expectancy-incongruent information. A total of 54 studies were included that measured memory towards concept-attribute combinations involving individuals, groups, evaluative and descriptive traits. Overall, measures of recognition-bias revealed an advantage for expectancy-congruent information over expectancy-incongruent information. The associated effect size amounted to d = 0.63 for response-bias measures. To date, however, the results of Rohner and Rasmussen (2011) constitute the only evidence for recognition-bias in connection with the physical attractiveness stereotype. It thus seems important to (1) examine if the effect of PAS congruence on recognition bias is replicable and (2) if the effect of PAS congruence on recognition bias prevails across a variety of conditions, that is, is generalizable. Specifically, we examined the influence of the nine moderator variables in two experiments: response type, attention, subjective compatibility, difference in encoding time, total encoding-time, impression formation, motivation to control prejudice, entity theory of personality and sex. We predicted that the effect is invariant across the levels of each moderator.
EXPERIMENT 1 Experiment 1 examined the moderating effects of response type (know, remember) and attention (to sex, to attractiveness).
Response type The subjective feeling of recalling something about another person can vary greatly in phenomenal quality. Remembering denotes a subjective state of awareness in which details of past events can be brought to mind, producing an experience of vivid recollection and of reliving past events. Knowing instead denotes a subjective state of awareness in which past events feel familiar, without the experience of recollection (Gardiner & RichardsonKlavehn, 2000; Tulving, 1985). Remembering is thus connected with more confidence in memory than knowing (e.g. Dunn, 2004). In Experiments 1 and 2 we examine recognition bias for know responses and for remember responses separately.1
Attention Social perceivers attend to different physical features of other individuals, such as sex, age and physical attractiveness, in addition to the psychological traits that they possess. Recognition bias may vary as a function of whether individuals attend to if the
Scand J Psychol 53 (2012)
information is congruent/incongruent with a stereotype, or if they attend to other aspects of a stimulus. In Experiment 1 we manipulated the direction of attention, either towards PAS congruence/ incongruence (getting participants to focus on physical attractiveness), or towards sex (getting participants to focus on whether faces were female or male).
Method Participants. Participants were 40 university students; 70.00% female and 30.00% male. Mean age was 23.78 with a standard deviation of 5.43. Materials. Stimuli consisted of pairs containing a face and a word. Forty faces that received high and low ratings of physical attractiveness by a group of six judges were used as attractive faces (20) and unattractive faces (20), respectively, see Rohner and Rasmussen (2011) for validity data. Attractive and unattractive faces were matched on emotional expression (17 neutral and 3 happy in each category) and sex (10 female and 10 male in each category). Forty words that a group of judges rated as negative and positive were used as negative words (20) and positive words (20), respectively (see Rohner & Rasmussen, 2011); see Table 1. Positive and negative words were matched on frequency of occurrence in Swedish newspapers and on word length. Combinations of faces and words were constructed for each participant as follows. Twenty old PAS-congruent and 20 old PAS-incongruent pairs were made by: (1) Randomly splitting each of attractive faces, unattractive faces, negative words, and positive words into two equally large sets; (2) combining attractive faces from set 1 with positive words from set 1 and combining unattractive faces from set 1 with negative words from set 1; (3) combining attractive faces from set 2 with negative words from set 2 and combining unattractive faces from set 2 with positive words from set 2. Twenty new PAS-congruent and 20 new PAS-incongruent combinations were made by randomly recombining words and faces within old attractive-positive, old unattractive-negative, old attractive-negative, and old unattractive-positive categories. New combinations thus had the same congruency (or incongruency) as old combinations. Note that this procedure renders PAS-congruent and PAS-incongruent conditions equivalent regarding the words and faces that they contain.
Table 1. English translations of the positive and negative words used in Experiment 1 Negative Anxious Apprehensive Depressed Disagreeable Dejected Edgy Gloomy Incompetent Ineffective Listless
2012 The Authors. Scandinavian Journal of Psychology 2012 The Scandinavian Psychological Associations.
Positive Negligent Passive Slow Uncommunicative Unhappy Unimaginative Unintelligent Unproductive Unreliable Untalented
Active Acute Ambitious Amiable Assiduous Competent Dynamic Effective Energetic Enterprising
Extraverted Gifted Nice Determined Productive Reliable Resolute Sharp Trustworthy Vigorous
Scand J Psychol 53 (2012)
Recognition bias and the physical attractiveness stereotype
241
Procedure. The experiment consisted of a study phase, a distraction phase and a test phase, each being preceded by a short written instruction presented on a computer monitor. Participants were randomly allocated to receive one of two study instructions (50.00% in each group) designed to manipulate the direction of attention. In the attend to sex condition, they were told that men and women have different traits and that the task was to find out if a given trait is characteristic of men or characteristic of women by deciding if a trait word was compatible with a face or not. In the attend to attractiveness condition they were told that attractive and unattractive individuals have different traits and that the task was to find out if a given trait is characteristic of attractive individuals or characteristic of unattractive individuals by deciding if a trait word was compatible with a face or not. At study, participants then saw 20 old PAS-congruent and 20 old PAS-incongruent combinations, one at a time, in random order. Face-word combinations were shown centrally on the computer monitor (randomly face to the left and word to the right or face to the right and word to the left). Response alternatives marked Yes and No appeared below face-word combinations and participants responded by clicking with the computer mouse at the locations of response buttons. Faces were about 7 · 8 cm large and words about 0.5 · 2–6 cm. The distraction phase comprised a 14-item multiple choice vocabulary test. Here participants were required to decide the meaning of 14 words by choosing one of 5 alternative answers. Words and response alternatives were presented on the computer monitor (to the left and right, respectively) and participants made their choice by clicking on an answer with the computer mouse. The distraction phase usually took between 4 and 12 minutes. The test phase consisted of a remember/know task (Gardiner & Richardson-Klavehn, 2000). Instructions informed participants that they should respond new if they did not recognize a faceword combination, to respond remember if they recognized a combination and could remember what they thought when they saw that combination at study, and to respond know if they recognized a combination but could not remember what they thought when they saw that combination at study (Gardiner & Richardson-Klavehn, 2000). Instructions underscored the importance of categorizing combinations and not faces or words as old or new. Participants then saw 20 old PAS-congruent combinations, 20 old PAS-incongruent combinations, 20 new PAS-congruent combinations, 20 new PAS-incongruent combinations and 40 filler trials, one at time, in random order. Face-word combinations appeared centrally (again randomly left-right or right-left) and the response alternatives new, remember and know were shown below faceword combinations. Participants responded with the computer mouse, there being no time limit. Pilot testing revealed rather high accuracy levels. To reduce the risk for ceiling effects, the test phase contained 40 filler trials in addition to old and new faceword combinations. In these trials a given face appeared together with a word that had the opposite valence of the word that had appeared together with that face at study. That is, if a given face was shown with a positive word at study, on filler trials, that face was shown with a negative word. A Macintosh computer with a 15¢¢ monitor running custom software presented stimuli and collected responses.
Data analysis. Measures of recognition-bias were computed for know responses and for remember responses using the algorithm described in Snodgrass and Corwin (1988), as follows: Br = F / (1 – H + F), where H is the frequency of hits (responding know or remember to an item from the study phase), and where F is the frequency of false alarms (responding know or remember to an item that was not shown in the study phase). Br increases when participants are biased towards responding know or remember, independently of variation in recognition sensitivity. All variables were then inspected and treated for outliers as follows: Scores with a two-tailed probability lower than p = 0.01 in a normal distribution were replaced with the raw scores that correspond to p = 0.01/2 and p = 1–0.01/2 (see Tabachnick and Fidell, 2001). The maximum number of outliers in a variable was 2.50%.
Results and discussion Measures of recognition bias were analyzed in repeated measures ANOVAs that had PAS congruence (2: congruent, incongruent) as a within-subjects factor. A separate ANOVA was computed for each moderator variable; that is, response type (2: know, remember) and attention (2: sex, attractiveness). The results are shown in Table 2 and Fig. 1. PAS congruence showed a marginally significant effect (p = 0.06) on recognition bias.2 Evidently, though, the effect of congruence was qualified by response type. Participants had more recognition bias for congruent combinations when they made know responses but not when they made remember responses. The interaction between congruence and attention was not significant. Participants thus showed a tendency to make more know responses towards PAS-congruent than PAS-incongruent combinations. Direction of attention at study, towards physical attractiveness or sex, did not have a moderating effect on recognition bias. Experiment 1 only studied the influence of two moderator variables. To further examine if PAS-related recognition bias is a robust phenomenon, we conducted an additional experiment. The aim was to (1) demonstrate that PAS-congruent information produces more recognition bias than PAS-incongruent information, and to (2) study the influence of additional potential moderators.
EXPERIMENT 2 Experiment 2 examined the effect of eight moderator variables: response type (introduced in Experiment 1), subjective compatibility, difference in encoding time, total encoding time, impres-
Table 2. Analysis of variance with recognition bias (Br) as the dependent variable Effect PAS-congruence PAS-congruence PAS-congruence PAS-congruence PAS-congruence
· Response-Type | Know | Remember · Attention
F
p
g2
3.68 5.56 4.39 2.40 1.23
0.06
