Processing of Ambiguous and Unambiguous ...

Journal of Personality and Social Psychology 1989, Vol. 56, No. 3,431-445

Copyright 1989 by the American Psychological Associalion, Inc. 0022-3514/89/$00.75

Processing of Ambiguous and Unambiguous Feedback by Depressed and Nondepressed College Studentsi Schematic Biases and Their Implications for Depressive Realism Benjamin M. Dykman and Lyn Y. Abramson University of Wisconsin--Madison

Lauren B. Alloy and Shirley Hartlage Northwestern University Explored schematic processing as a mechanism for predicting (a) when depressed Ss would be negative relative to nondepressed Ss and (b) when depressed and nondepressed Ss would show biased or unbiased (i.e., "realistic") processing. Depressed and nondepressed Ss performed multiple trials of a task under conditions in which the two groups held either equivalent or different schemas regarding this task. Ss received either an unambiguous or objectively normed ambiguous feedback cue on each trial. In full support of schematic processing, depressed Ss showed negative encoding relative to nondepressed Ss only when their schemas were more negative, and both depressed and nondepressed Ss showed positively biased, negatively biased, and unbiased encoding depending on the relative feedback cue-to-schema match. Depressed and nondepressed Ss' response latencies to unambiguous feedback also supported the occurrence of schematic processing. We discuss the methodological, treatment, and "realism" implications of these findings and suggest a more precise formulation of Beck's schema theory of depression.

unrealistic or biased relative to nondepressed individuals. (By the term characteristically we mean always or across all situations, content domains, etc.). It is noteworthy that "characteristic processing" interpretations of depressive cognition have not been limited to Beck's schema theory but extend even to findings that run counter to his view. Specifically, a number of studies conducted on the judgment of control (e.g., Alloy & Abramson, 1979, 1982; Golin, Terrell, & Johnson, 1977; Golin, Terrell, Weitz, & Drost, 1979), recall of feedback (e.g., Nelson & Craighead, 1977, punishment condition; Wenzlaff & Berman, 1985), and perception of social competence (e.g., Lewinsohn, Mischel, Chaplin, & Barton, 1980) have also been interpreted by many as suggesting that depressed and nondepressed individuals process information in characteristically different ways (i.e., that depressed people are characteristically accurate, whereas nondepressed people are characteristically positively biased). Hence, perhaps owing to a "fundamental attribution error" (Ross, 1977), we hold that there has been a general tendency to interpret both clinical and empirical evidence as indicating that depressed and nondepressed individuals process information in traiflike, characteristic ways. In this research, we examined depressed and nondepressed individuals' encoding and recall of ambiguous and unambiguous feedback to determine (a) when depressed individuals will and will not make negative inferences relative to nondepressed individuals, and (b) when depressed and nondepressed individuals will be biased or unbiased relative to a normative baseline. In addressing these issues, we discuss and contrast the predictions one would make on the basis of existing characteristic processing accounts of depressive and nondepressive cognition with those one would make on the basis of contemporary research and theorizing on the schema.

A core postulate of Beck's (1967, 1976; Beck, Rush, Shaw, & Emery, 1979) cognitive theory of depression is that depressed individuals possess stable and enduring negative schemas that serve as conceptual filters for the coding, screening, and general evaluation of impinging stimuli. According to Beck, these negative schemas and the "faulty" processing strategies they engender (e.g., arbitrary inference, dichotomous thinking) lead to pervasive negative themes in the depressed individual's cognitions about the self, world, and future. Beck hypothesized that these cognitions are not only consistently negative relative to those ofnondepressed individuals, but are consistently exaggerated and hence unrealistically negative as well. Given Beck's (1967, 1976, Beck et al., 1979) portrayal of depressive cognition, a frequent interpretation of his schema theory that has guided much research is that depressed individuals are both "characteristically" negative and "characteristically"

This research is based on a doctoral dissertation submitted by Benjamin M. Dykman to the University of Wisconsin--Madison. Support for this research came from a Romnes Fellowship awarded to Lyn Y. Abramson and by a grant from the John D. and Catherine T. MacArthur Foundation awarded to both Lyn Y. Abramson and Lauren B. Alloy. Portions of this research were presented at the 21st Annual Convention of the Association for the Advancement of Behavior Therapy, Boston, November 1987. We would like to thank Jim Embke, Carrie Fuller, Angelique Hemery, Nancy Lightfoot, Marion Meyer, and Joy WaUin for their assistance with data collection and Judy Markgraf for her assistance in the preparation of this article. Correspondence concerning this article should be addressed to Benjamin M. Dykm.an, Department of Psychology,University of Wisconsin, 1202 W. Johnson Street, Madison, Wisconsin 53706. 431

432

DYKMAN, ABRAMSON, ALLOY, AND HARTLAGE

Characteristic Negativity Versus Schematic Negativity: Predicting Depressed-Nondepressed Differences in Encoding Although many studies have demonstrated that depressed individuals process information more negatively than nondepressed individuals, studies comparing these groups in their encoding of ambiguous feedback have not followed this trend. Specifically, two such investigations (Craighead, Hickey, & DeMonbreun, 1979; DeMonbreun & Craighead, 1977) obtained no support for the prediction that depressed individuals negatively bias their encoding of ambiguous feedback relative to nondepressed individuals, and a third study (Dykman & Volpicelli, 1983) obtained only partial support for this prediction. Although these ambiguous feedback studies apparently contradict Beck's (1967, 1976) schema theory of depression, this may be the case only if one views Beck's theory as a characteristic processing theory rather than as a schema theory. That is, as explicated by current research and theorizing on the schema (cf. Alba & Hasher, 1983; Markus & Zajonc, 1985; Taylor & Crocker, 1981), depressed and nondepressed individuals should interpret the same information differently only when each group assimilates that information (via schematic processing) to different schemas. Quite possibly, however, depressed and nondepressed subjects in prior ambiguous feedback research may not have held different schemas with regard to such novel tasks as dot counting (Dykman & Volpicelli, 1983) or syllable matching (Craighead et al., 1979; DeMonbreun & Craighead, 1977), a consideration supported by work indicating that depressed and nondepressed individuals hold equivalent performance expectancies across a wide range of tasks (e.g., Abramson, Garber, Edwards, & Seligman, 1978; Smolen, 1978). One objective of the current research was to determine whether depressed individuals are characteristically negative in their encoding of ambiguous feedback relative to nondepressed individuals, or whether such negativity depends on depressed people having more negative schemas than nondepressed people with regard to the task at hand. To achieve this objective, we first determined various content areas that discriminated depressed and nondepressed individuals with regard to their self-schemas as opposed to content areas that did not (i.e., Study 1). We then incorporated this knowledge in designing the two experimental conditions of our second and main study of interest. In the schema-discriminating condition, depressed and nondepressed subjects performed a task ostensibly requiring abilities that, according to Study 1, discriminated these two groups vis-fi-vistheir self-schemas. In the schema-nondiscriminating condition, a second group of depressed and nondepressed subjects performed an identical task that, by contrast, ostensibly required abilities that did not discriminate the selfschemas of these two groups. In both conditions, subjects received multiple trials of ambiguous and unambiguous feedback, and following each trial subjects were required either to identify or to infer the identity of each feedback cue administered. We reasoned that if depressed individuals are characteristically negative in their encoding of ambiguous feedback relative to nondepressed individuals, then depressed-nondepressed encoding differences should obtain in both experimental conditions. On the other hand, if schematic differences between depressed and nondepressed individuals are a determinant of de-

pressed-nondepressed encoding differences, then depressed individuals should negatively bias their encoding of ambiguous feedback relative to nondepressed individuals in the schemadiscriminating but not the schema-nondiscriminating condition. ~ Last, insofar as both Beck (1967, 1976) and contemporary social-cognitive psychologists (Markus & Zajonc, 1985) have pointed to the schema's heightened influence under conditions of ambiguity, we also explored whether depressed-nondepressed encoding differences would be greater for ambiguous compared with unambiguous feedback trials. Characteristically Biased Processing Versus Schematically Biased Processing: Predicting Nonnormative Bias

Biases in the Encoding of Ambiguous Feedback As described in the preceding section, two major viewpoints exist that address the normativeness of depressive and nondepressive cognition, with each viewpoint suggesting the possibility that depressed and nondepressed individuals process information in characteristically biased or unbiased ways. One view relates to Beck's assertion (1967, 1976; Beck et al., 1979) that, relative to a normative baseline, depressed individuals make negatively biased or distorted interpretations of reality, whereas nondepressed individuals make unbiased or realistic interpretations (which we term depressivepessimism and nondepressive realism). The second view derives from work (see Alloy & Abramson, 1988, and Alloy, Albright, Abramson, & Dykman, in press, for reviews) that suggests the converse effect: that is, that relative to a normative baseline, depressive inference is unbiased or realistic, whereas nondepressive inference leads to positively biased or distorted interpretations of reality (i.e., depressive realism and nondepressive optimism). A second objective of this research was to determine whether depressed and nondepressed individuals do in fact have characteristic biasing tendencies that fit either of the viewpoints described previously. To assess this possibility, we presented subjects with ambiguous feedback in the form of different feedback "pairs" and required subjects to choose which of the two competing feedback cues in each pair was their "true" feedback. Following Kahneman and Tversky (1973)and Langer (1975), we adopted a probability-based standard as our normative baseline and accordingly defined an unbiased outcome for any given feedback pair as a pattern of judgments in which both members of the feedback pair are selected with equal probability. On the other hand, we defined a biased outcome for any given feedback pair as a pattern of judgments in which one member of the feedback pair is selected with a higher probability than the other member. On this basis, we predicted that if depressive pessimism and nondepressive realism reflect the use of characteristic processing strategies, then in both experimental conditions de-

On the basis of the general failure of previous encodingof ambiguous feedback studies to obtain depressed-nondepressed processing differences, we acknowledge that a third possible prediction is that group differenceswould not obtain in either the schema-discriminating or schema-nondiscriminatingcondition. Such a findingwould support the view that depressed and nondepressed individuals simply do not differ in their encodingof ambiguousinformation.

SCHEMATIC BIASES AND DEPRESSION pressed subjects should choose the more negative member of each given feedback pair with a higher probability than the more positive member, whereas nondepressed subjects should choose both members with equal probability. On the other hand, if depressive realism and nondepressive optimism reflect the use of characteristic processing strategies, we predicted the appropriate counterpart effect. The preceding predictions rest on the premise that depressed and nondepressed individuals are either characteristically unbiased or biased information processors. Suggesting an alternative view, however, are the mixed findings that have been obtained with regard to displays of bias among these groups. That is, whereas some studies have found depressed people to be the more "realistic" group (e.g., Alloy & Abramson, 1979; Nelson & Craighead, 1977, punishment condition; Lewinsohn et al., 1980), other studies have found nondepressed people to be the more "realistic" group (e.g., DeMonbreun & Craighead, 1977; Gotlib, 1983; Nelson & Craighead, 1977, reinforcement condition). As these mixed findings suggest that biased and unbiased outcomes are not specific to either group (Kuiper & Higgins, 1985), an intriguing additional hypothesis is that both depressed and nondepressed individuals use the same processing strategy, one that leads to both realistic (i.e., unbiased) and unrealistic (i.e., biased) outcomes for both depressed and nondepressed individuals alike, depending on conditions. One processing strategy that current research and theorizing suggests might meet these requirements is schematic processing (Neisser, 1976), whereby information is interpreted so as to be consistent with the content of a schema. Thus, we also explored whether both depressed and nondepressed people engage in schematic processing and thus bias their encoding of ambiguous feedback in the direction of their respective schemas. If this were the case, we predicted that both biased and unbiased outcomes would occur for each group, with both types of outcomes being a function of the match between the valence of each group's schema and the particular feedback pair being processed. Thus, for depressed and nondepressed subjects in both experimental conditions, the encoding of ambiguous feedback pairs negatively positioned relative to the schema should be positively biased, the encoding of ambiguous feedback pairs positively positioned relative to the schema should be negatively biased, and the encoding of ambiguous feedback pairs that match the schema should be unbiased. Last, if we combine across ambiguous feedback pairs, then schematic processing allows either depressed or nondepressed individuals to be the less biased group overall, depending on which group's schema more closely matches the average value of this combined feedback group.

Biases in Response Latencies to Unambiguous Feedback In addition to examining biases in the encoding of ambiguous feedback, we also examined possible biases in depressed and nondepressed individuals' identification of unambiguous feedback. To gauge this bias, we unobtrusively recorded subjects' response latencies to the differently valenced unambiguous feedback cues presented. The distinction between characteristic and schematic processing again served as a framework for prediction. We reasoned that having a particular characteristic processing bias

433

(i.e., depressive pessimism or nondepressive optimism) would result in a linearly increasing delayed response (or "resistance") to feedback whose valence was increasingly inconsistent with that bias. Alternatively, characteristically unbiased processing (i.e., depressive realism or nondepressive realism) would result in some constant reaction time regardless of feedback valence. A different prediction, however, would follow if both depressed and nondepressed individuals engaged in schematically biased processing. On the basis of a number of findings indicating faster processing times for schema-congruent as opposed to schema-incongruent material (e.g., Lingle & Ostrom, 1979; Markus, 1977; Payne, Connor, & Colletti, 1987), we would ex: pect the response latencies of both depressed and nondepressed subjects to vary as a function of the degree to which a particular feedback cue and a particular group's schema are congruent. Thus, for depressed and nondepressed subjects in both experimental conditions, response times should be fastest to the feedback cue lying closest to each group's schema and progressively slower to feedback cues lying progressively farther from each group's schema (i.e., a curvilinear latency pattern if a group's schema lies between the extremes of the feedback valences administered).

Biases in Summary Recall of Unambiguous Feedback We obtained a final measure of processing biases in this study by using a recall task in which subjects were asked to recall how many trials of each different unambiguous feedback type they had received. We predicted that if depressive pessimism and nondepressive realism are characteristic tendencies, then depressed subjects should negatively bias their recall of feedback, whereas nondepressed subjects should be unbiased (i.e., across both experimental conditions). On the other hand, if depressive realism and nondepressive optimism are characteristic tendencies, we predicted the appropriate counterpart effect. However, if both depressed and nondepressed subjects engaged in schematically biased processing, then biased or unbiased recall could occur among either group depending on the relative match between a group's schema and the average value of the various unambiguous feedback cues administered (with the direction of bias again going in the direction of each group's schema). Study 1 As suggested earlier, a number of investigators have presumed that depressed and nondepressed subjects approach various laboratory tasks with different schemas without first subjeering this presumption to test. The goal of Study 1 was to compare the self-schemas of depressed and nondepressed subjects on each of 50 trait dimensions and thereby determine the content domains in which depressed subjects' self-schemas are more negative than those of nondepressed subjects. We then applied the knowledge gained from these comparisons to Study 2 to assess the conditions under which depressed subjects would negatively bias their encoding of ambiguous and unambiguous feedback relative to nondepressed subjects.

Method Subjects, materials, and procedure. Introductory psychology students servedas subjectsand receivedextra-credit points for their partici-

434


pation in the study. In a mass screening session at the start of the semester, male and female subjects were given the Beck Depression Inventory (BDI; Beck, 1967) followed by a 50-item self-schema inventory entitled the "Self-Description Questionnaire." The BDI served as a screening measure; 37 subjects with BDI scores of 0 or l were classified as nondepressed (M = .27) and 21 subjects with scores of 13 or above (M = 19.10) were classified as depressed. We used the self-schema inventory to assess and compare the selfschemas of subjects classified as depressed and nondepressed on each of 50 randomly ordered trait dimensions. For each trait dimension, subjects rated how they generally viewed themselves on that trait dimension and then made a rating of how important that trait dimension was to their self-view. All ratings were made on a 9-point Likert scale (e.g., 1 = Very unmotivated, 9 = Very motivated; or 1 = Not at all important, 9 = Very important). For a group to qualify as having a self-schema on a particular trait dimension, we required that depressed and nondepressed subjects' mean self-ratings be accompanied by a mean importance rating of at least 5.0, signifying that a trait dimension was of at least moderate importance and thus could be considered a well-articulated structure in memory (cf. Markus, 1977; Markus, Smith, & Moreland, 1985).

Results and Discussion For each trait dimension, we tested for depressed-nondepressed differences in self-ratings and importance ratings using t tests. Table 1 presents these results, with self-ratings rank ordered by ascending t values. Examination o f Table 1 indicates that the self-ratings o f depressed and nondepressed subjects differed significantly (in the expected direction) on 41 out of the 50 trait dimensions presented, but that the remaining 9 trait dimensions did not significantly distinguish the two groups. 2'3 As noted earlier, for a group to qualify as having a schema along a given trait dimension, we required that the trait receive a mean importance rating by that group of at least 5.0. As Table 1 indicates, for both depressed and nondepressed subjects, all 50 trait dimensions met this criterion (this unanimity perhaps reflecting our initial intuitively guided attempt to select important trait dimensions). Taken as a whole, these results suggest that although depressed individuals' self-schemas are generally negative, they are not characteristically (i.e., universally) negative relative to those of nondepressed individuals. Instead, depressed and nondepressed subjects appear to hold statistically equivalent selfschemas on a select n u m b e r o f trait dimensions.

Study 2 Given Study l's results that depressed and nondepressed subjects' self-schemas differed in some content areas but not others, one goal o f Study 2 was to determine whether these groups would show counterpart differences in the encoding of ambiguous and unambiguous feedback. Thus, we incorporated into Study 2's design a n u m b e r of content areas from Study 1 that did and did not discriminate the self-schemas o f depressed and nondepressed subjects. In Study 2 we also compared depressed and nondepressed individuals' encoding, response latencies, and recall of ambiguous and unambiguous feedback to determ i n e whether either group is "characteristically" biased or unbiased, or whether biased and unbiased outcomes for both groups are a function o f the feedback-to-schema match.

Method Subjects. As in Study 1, subjects were drawn from a pool of introductory psychology students, but in the subsequent semester. In another mass screening session, subjects completed the BDI, and individuals with BDI scores of 13 and above (depressed) or 0 or 1 (nondepressed) were later phoned to participate in the experiment. Just prior to each subject's particiPation, subjects again completed the BDI and, in order to match the subject selection criteria of Study 1, only subjects who scored 13 or above and 0 or 1 at retest were included in the final data analysis (the average intertest interval being approximately 3 weeks). The final subject sample consisted of 84 subjects, 30 men and 54 women. Experimental design andfeedback conditions. We randomly assigned subjects to one of four experimental conditions in a 2 × 2 × 2 factorial design with between-subjects variables of group (depressed vs. nondepressed) and condition (schema-discriminating vs. schema-nondiseriminating) and a within-subjects variable of feedback clarity (ambiguous vs. unambiguous). In both the schema-discriminating and schema-nondiscriminating conditions, subjects performed an identical task, that is, 45 trials of a computer-administered dot-counting task. In the schema-discriminating condition, performance on the dot-counting task was described as a measure of one's tendency to be successful, productive, and motivated. In Study 1, these were trait dimensions on which a similar sample of depressed and nondepressed subjects held different schemas. In the schema-nondiscriminating condition, by contrast, subjects were told that performance on the dot-counting task was a measure of one's tendency to be polite, considerate, and courteous. In Study I, these were trait dimensions on which a similar sample of depressed and nondepressed subjects held equivalent schemas. Although theoretically other schema-discriminating and sehema-nondiscriminating trait dimensions could also have served as task descriptors, our final choices reflected our preference for traits that lay at or near the extremes of the t value continuum (to enhance the robustness of our effects) and that were conducive to the construction of a compelling cover story. 4 Manip-

2 We also assessed whether depressed-nondepressed differences in self-ratings covaried as a function of subjects' importance ratings. Comparing depressed and nondepressed subjects' self-ratings using respective importance ratings as covariates yielded adjusted t values whose rank ordering paralleled closely the rank ordering presented in Table 1 (Spearman's p = .98). Similarly, subjecting the self-ratings presented in the first 2 rows of Table 3 to a two-way analysis of covariance (with importance ratings as the covariate) yielded an interaction F value of 16.53, which is essentially equivalent to the nonadjusted two-way analysis of variance F value of 16.80. 3 It is important to point out that although we attempted to sample broadly in constructing the self-schema inventory, our sampling approach was largely exploratory and intuitively based. Until a more exhaustive and bias-protected sampling is undertaken, we caution that the proportions of schema-discriminating and nondiscriminating traits we obtained (i.e., 41:50 and 9:50, respectively) may not be representative. 4 We also assessed whether the two sets of traits used as task descriptors were factorially independent of each other. This was done by subjecting subjects' ratings on the six trait dimensions to a principal-factor analysis with orthogonal rotation. Two factors having eigenvalues greater than 1.0 emerged, accounting for 80% of the total variance in the matrix. Invoking a .50 inclusion cutoff, Factor ! emerged as an Achievement factor with loadings of .96, .88, and .79, respectively, for the traits of unproductive-productive, unmotivated-motivated, and unsuccessful-successful. Factor 2 emerged as a Sensitive-to-Others factor, with loadings of.88, .69, and .68, respectively, for the traits of rudepolite, discourteous-courteous, and inconsiderate-considerate. All remaining loadings were .29 or lower. This analysis thus corroborates the

SCHEMATIC BIASES AND DEPRESSION ulation checks appraising the plausibility o f the cover story are described later. Following each trial o f the dot-counting task, subjects received computer-generated but bogus feedback evaluating the accuracy of their dot estimate. In both experimental conditions, feedback was conveyed by way of a vertically arrayed 5-point scale. Each point on the scale was anchored with an evaluative label, but in each experimental condition these labels differed in order to correspond to the schema of interest. Thus, in the schema-discriminating condition, the 5-point scale was anchored in the following way: 1 You are a person who is VERY SUCCESSFUL (very motivated, productive) 2 You are a person who is SOMEWHAT SUCCESSFUL (somewhat motivated, productive) 3 You arc a person who is EQUALLY SUCCESSFUL AND UNSUCCESSFUL 4 You are a person who is SOMEWHAT UNSUCCESSFUL (somewhat unmotivated, unproductive) 5 You are a person who is VERY UNSUCCESSFUL (very unmotivated, unproductive) This contrasts with the 5-point scale in the schema-nondiseriminating condition, which was anchored as follows: l You are a person who is VERY POLITE (very considerate, courteous) 2 You are a person who is SOMEWHAT POLITE (somewhat consideratc, courteous) 3 You are a person who is EQUALLY POLITE AND RUDE 4 You are a person who is SOMEWHAT RUDE (somewhat inconsiderate, discourteous) 5 You are a person who is VERY RUDE (very inconsiderate, discourteous) in personalizing the feedback in both experimental conditions (i.e., "You are a person who i s . . . . " ) , we hoped to increase the likelihood that subjects' responses to the feedback would be mediated by the particular schema of interest. In both experimental conditions, the full 5-point feedback display appeared each time a feedback trial was administered. For unambiguous feedback trials, subjects were told that the number corresponding to their feedback would start flashing. Hence, unambiguous feedback consisted o f a single flashing number. For the ambiguous feedback trials, however, subjects were administered two simultaneously flashing numbers instead of one, in which case subjects were required to choose which of the two competing feedback candidates was their "true" feedback. For the ambiguous feedback trials, repetitions of four different feedback pairs were presented, that is, the 1-2, 2-3, 3-4, and 4-5 pairs. Procedure. Subjects were run individually by an experimenter who was blind to the subjects' depression status. On arriving for the experiment, subjects were requested to read and sign a consent form and then complete the BDI. Subjects were then seated in front of a Commodore 64 computer keyboard and an Apple lie green screen monitor. The experimenter, seated approximately 3 ft away, then started a taped set of instructions. At the start of the taped instructions, all subjects were told that during the experiment 45 dot displays containing between l0 and 500 dots

internal consistency and factorial independence o f our task descriptor sets. More generally, it is interesting to note that traits in Table l having small t values tended to be Sensitive-to-Others-related, whereas traits having large t values tended to be Achievement-related. This suggests that there may exist independent but broad domains along which the schemas of depressed and nondepressed individuals differ more (i.e., the achievement domain) or less (i.e., the sensitive-to-others domain).

435

would appear briefly on the computer screen in front oftbem, and that their task was to estimate the number o f dots that appeared in each display. Following this general introduction, the taped instructions differed for subjects in the two experimental conditions. Subjects in the schema-nondiseriminating condition received instructions that portrayed the dot-counting task as a measure o f one's tendency to be polite, considerate, and courteous (i.e., schema-nondiscriminating trait dimensions from Study 1). This task description read as follows: The reason we know that doing well on this task is related to one's ability to evaluate ambiguous situations quickly and accurately is through studies where we have looked at the way people handle real-life situations. What we have found is that people who do well on the dot counting task also tend to do well in certain real life situations that are ambiguous and that require quick and accurate judgments. Perhaps the best example of such a real-life situation is one in which we must quickly decide what to do in order to behave politely towards another person. That is, oftentimes we find ourselves in situations with other people where there are no clear guidelines as to what we should say or how we should act. In such situations, we must size up the situation quickly and find out what the other person's needs are and how we should respond to those needs. Thus, in many social situations that require politeness, there's a lot that goes on in a short amount of time just as a lot goes on in a short amount o f time in the dot counting task that you are about to perform. Evidence that supports a close relationship between one's dot counting ability and one's tendency to behave rudely or politely comes from a study we conducted on over 300 U.W. undergrads. The data from this study indicate that those same students who performed well on the dot counting task also tended to respond with politeness, courteousness, and considerateness when placed in ambiguous social situations that required quick and accurate judgments. On the other hand, students who did poorly on the dot counting task were people who tended to come across as rude, discourteous, and inconsiderate when placed in such ambiguous social situations. Thus, how well you perform on the dot counting task is highly predictive o f how polite or rude you will come across in ambiguous social situations that require a quick and accurate sizing up of the situation. For subjects in the schema-discriminating condition, on the other hand, performance on the dot-counting task was portrayed as a measure o f one's tendency to be successful, motivated, and productive (i.e., schema-discriminating trait dimensions from Study I). This task description read: As you can well guess, the ability to evaluate ambiguous situations quickly and accurately is an ability that is needed for doing well in a large number o f real-life situations. This is because many situations in life, especially high pressure achievement situations, require us to make quick and accurate judgments even when the right answer is not readily apparent. This often occurs in many high pressure achievement situations where there are no clear guidelines as to what we should say, how we should act, or what the right answer is. Thus in many high pressure achievement situations there's a lot that goes on in a short amount o f time and our job is to come up with a correct judgment. As you will soon see, there's a lot that goes on in a short amount o f time in the dot counting task but you must nevertheless try to come up with a correct judgment. Evidence that supports a close relationship between one's dot counting ability and one's ability to perform well in high pressure achievement situations comes from a study we conducted on over 300 U.W. undergrads. This study indicates that students who performed well on the dot counting task also tended to be much more successful in achievement situations than students who performed poorly on the dot counting task. These students not only tended to be more successful, but they also tended to be more motivated and more productive compared to students who performed poorly on the dot counting task. Thus, because the dot counting task is representative o f high pressure, ambiguous achievement situations that

436


Table 1

Mean Self-Ratings and Importance Ratings by Depressed and Nondepressed Subjects on 50 Trait Dimensions: Study I Mean self-ratings

Mean importance ratings

Trait dimension

Depressed

Nondepressed

t

Depressed

Nondepressed

Unsympathetic-sympathetic Rude-polite Boastful-humble Discourteous-courteous Inconsiderate-considerate Untrustworthy-trustworthy Unreliable-reliable Dishonest-honest Unfriendly-friendly Stingy-generous Undependable-dependable Unkind-kind Disagreeable-agreeable Immature-mature Bad-good sense of humor Disliked-liked by others Disrespectful-respectful Illogical-logical Unpredictable-predictable Introverted-extroverted De fensive-n ondefensive Unassertive-assertive Shy-outgoing Sluggish-energetic Timid-brave Unenthusiastic-enthusiastic Helpless-masterful Undecisive-decisive Socially unskilled-skilled Impulsive-nonimpulsive Deceiving-genuine Unambitious-ambitious Ineffective-effective Unintelligent-intelligent Dull-interesting Fearful-fearless Unmotivated-motivated Unlovable-lovable Incapable-capable Shameful-praiseworthy. Incompetent-competent Unproductive-productive Unconfident-confident Unsuccessful-successful Pessimistic-optimistic Purposeless-purposeful Finds life meaninglessmeaningful A loser-a winner Unhappy-happy Worthless-worthy

7.57 7.57 6.43 7.38 7.24 7.95 7.81 7.05 6.95 6.48 7.86 7.10 6.10 6.57 6.62 6.33 7.10 6.38 4.33 5.29 3.67 5.10 4.86 6.19 5.00 5.81 5.43 4.67 5.10 3.71 6.05 6.14 5.55 5.86 5.43 4.33 5.19 5.71 5.95 5.43 5.71 5.33 4.33 5.43 4.43 5.14

7.62 7.64 6.51 7.78 7.62 8.24 8.19 7.78 7.70 7.30 8.32 7.81 7.16 7.70 7.70 7.43 7.94 7.54 5.75 6.69 5.19 7.08 6.81 7.68 6.69 7.35 7.05 6.84 7.00 5.32 7.75 7.81 7.70 7.76 7.31 6.62 7.72 7.73 8.14 7.46 8.24 7.70 7.32 7.59 7.19 7.81

0.14 0.21 0.27 1.16 1.25 1.29 1.73 1.90 1.91 2.03* 2.12" 2.32 * 2.38* 2.44* 2.45* 2.50* 2.61 * 2.65* 2.83** 2.88** 3.09** 3.31 ** 3.39** 3.44** 3.45** 3.45** 3.51 ** 3.51 ** 3.57** 3.60*** 3.68** 3.79*** 4.00*** 4.22*** 4.31 *** 4.33*** 4.38*** 4.53*** 4.73*** 4.94*** 4.97*** 5.14"** 5.45*** 5.5 l*** 5.71 *** 5.79"**

7.43 7.14 6.48 7.43 7.67 8.19 8.14 7.43 8.10 6.67 8.33 8.05 6.62 7.62 7.67 8.14 7.24 7.43 5.71 6.67 6.57 7.00 7.43 7.14 6.14 7.14 7.43 6.81 7.62 5.86 7.19 7.57 7.33 8. l0 8.00 5.67 8.00 8.14 8.00 7.00 8.24 7.86 7.95 8.00 7.43 7.52

7.24 7.69 6.14 7.4 l 7.89 8.41 8.49 7.9 l 8.08 7.05 8.38 8.22 6.59 7.41 7.27 7.68 7.81 6.76 5.06 6.83 5.43 7.08 7.32 6.81 6.08 6.86 7.22 7.27 7.16 5.38 7.58 7.59 7.35 7.97 7.22 5.65 7.6 l 8.03 8.22 7.14 8.27 7.70 7.78 7.86 7.00 7.46

5.05 4.95 4.43 4.81

8.08 7.65 7.84 7.95

6.13*** 6.29*** 7.25*** 7.62***

7.40 7.67 7.62 7.33

7.62 7.65 8.16 7.46

Note. Except for the trait dimensions of dull-interesting and defensive-nondefensive,the importance ratings of depressed and nondepressed subjects did not differ significantly. * p < .05, two-tailed. ** p < .0 l, two-tailed. *** p < .00 l, two-tailed.

require quick and accurate judgments, the dot counting task is also predictive of one's general level of success as well as one's general level of motivation and productivity. Again, students who perform well on the dot counting task are high in general success, motivation, and productivity, whereas students who perform poorly on the dot counting task are low on these characteristics. Following these contrasting cover stories, all subjects were told that after each dot display they were to make their best estimate of how many

dots appeared and then enter that estimate into the computer. Subjects were also told that following their estimate, they would receive feedback from the computer evaluating the accuracy of their judgment. Subjects were further informed that the feedback they received would tell them where they stood on the personality dimension of interest. Thus, subjects were instructed "the more accurate you are on the dot counting task, the more Polite (or Successful) your feedback will indicate you are." At this point, subjects were shown a sample 5-point feedback display

SCHEMATIC BIASES AND DEPRESSION and were told that following each dot estimate, a number on the display would start flashing to indicate what their feedback was. (As discussed earlier, a single flashing number constituted an unambiguous feedback cue.) Subjects were instructed that after receiving a feedback cue in the form of a flashing number, they were to indicate which of the five feedback cues they received "by simply pressing the number on the number keyboard which exactly matches the number of the evaluation you were administered." Subjects were then informed that for some of the dot estimates they made, their feedback would take the form of two flashing numbers instead of one (i.e., an ambiguous feedback cue). Subjects were told that when this occurred, their task was to "determine which of the two flashing numbers is really your feedback . . . After you decide which of the two flashing numbers is actually your feedback, we would like you to press the number on the number keyboard which matches the number you think is your true feedback." Following a brief procedural summary, subjects completed a brief pretask questionnaire (to be described) and then began the dot-counting task. Dot displays and feedback administrations were equated in the two experimental conditions. In actuality, dot displays contained between 66 and 255 dots, and because of the extremely brief presentation time (. 12 s), subjects were not overly suspicious (as revealed at debriefing) of the completely randomized feedback administered by the computer. In both experimental conditions, five successive groupings of randomly ordered feedback were generated by the computer such that one of each of the five unambiguous feedback cues and one of each of the four ambiguous feedback pairs appeared within each grouping. Thus, across the 45 feedback trials, subjects received five presentations of each of the five unambiguous feedback types and five presentations of each of the four ambiguous feedback types. In both experimental conditions, however, randomization was restricted by positioning the Number 3 feedback cue as the first feedback presentation and by not permitting two identical feedback cues to occur in succession. With the aid of procedural prompts from the computer, subjects selfadministered all 45 experimental trials at a self-determined pace. To assess whether subjects' response times differed across the five gradations of unambiguous feedback, elapsed time between feedback onset and feedback identification was unobtrusively recorded by the computer. Self-schema indexes and manipulation checks. Following the taped instructions, subjects were asked to complete a four-item pretask questionnaire. We constructed counterpart questionnaires for each experimental condition such that the two questionnaires differed only with regard to the specific trait dimensions addressed (e.g., very successful vs. very polite). Two items on each questionnaire jointly served as indexes of the schema. The first asked subjects to "Rate how you generally view yourself on the following trait dimension." Subjects indicated their self-rating on a 5-point scale that was anchored at each point in a manner identical to the 5-point feedback scale described earlier (with the deletion of the prefacing phrase, "You are a person who is"). The identical labeling of points on the self-rating scale and the feedback scale was purposeful in that this allowed us to superimpose the two scales to assess whether a relation existed between the valence of subjects' schemas and the direction of bias in subjects' encoding of each of the four ambiguous feedback types. A second item pertinent to our definition of the schema asked subjects to rate how important the condition-relevant trait dimension was to their view of themselves. Subjects responded on a 5-point scale (1 =

This dimension is very important to my view of myself. 5 = This dimension is not at all important to my view of myselJ). Similar to Study 1, we considered depressed and nondepressed groups in each experimental condition as holding a schema only if each condition-relevant trait dimension received a mean importance rating of 3.0 or lower, signifying that subjects regarded the trait dimension to be of at least moderate importance. In addition, to help rule out plausible alternative interpretations of our results, we deemed it necessary that subjects not differ

437

Table 2

Beck Depression Inventory Characteristics for Each Subject Group: Study 2 Group

Nper cell

M

SD

Depressed-Schema-discriminating Depressed-Schema-nondiscriminating Nondepressed-Schema-discriminating Nondepressed-Schema-nondiscriminating

27 17 21 19

18.89 17.41 0.43 0.37

5.42 5.69 0.5 l 0.50

across the two conditions in the importance ascribed to each trait dimension. The two remaining items on each pretask questionnaire served as indirect manipulation checks (phrased indirectly to offset heightening subjects' suspicions). The first of these asked subjects to rate how well they expected to perform on the dot-counting task that would follow. Subjects responded on a 5-point scale ( 1 = Very well, 5 = Very poorly). We predicted that if subjects in each condition actually regarded the experimental task to be a measure of their tendency to be successful or polite (as described in the cover story), then subject's performance expectancies should correlate with subject's self-ratings on these abilities. As a second manipulation check, we asked subjects to rate how important it was for them to perform well on the dot-counting task that would follow. Subjects again responded on a 5-point scale (1 = Very important, 5 = Not at all important). As an indication that our cover stories were equally plausible in the two experimental conditions, we reasoned that subjects should not differ across the two conditions in how important it was to perform well. Summary recall measure. On completing the dot-counting task, subjects in each experimental condition filled out a bogus "Social Perception Inventory" that had been constructed to serve as a filler task for the recall measure that immediately followed. Subjects, on the average, completed the seven-item filler task in approximately 4 rain. As our interest in subjects' summary recall was limited to unambiguous feedback trials, subjects were instructed to limit their responses on the recall measure to only those feedback trials that had consisted of one flashing number. For each experimental condition, the five relevant unambiguous feedback types were listed, and subjects placed their estimate of how many of each unambiguous feedback type they had received in the blanks provided. In addition, subjects were informed that a total of 25 unambiguous feedback stimuli had been administered, and thus their recall total should similarly sum to 25. On completing the recall measure, subjects were thoroughly debriefed.

Results Subject characteristics. G r o u p m e a n s and standard deviations for scores on the BDI, as well as cell sizes, are p r e s e n t e d in Table 2. A G r o u p (depressed vs. nondepressed) X Condition (schema-discriminating vs. s c h e m a - n o n d i s c r i m i n a t i n g ) twoway analysis o f variance (ANOVA)p e r f o r m e d on subjects' BDI scores revealed only the e x p e c t e d group m a i n effect, F ( l , 80) -398.50, p < .001. Subsequent t tests c o n f i r m e d the presence o f significantly higher BDI scores for depressed subjects c o m p a r e d with nondepressed subjects in b o t h the s c h e m a - d i s c r i m i n a t i n g condition, t(80) = 12.72, p < .00 l, and the s c h e m a - n o n d i s c r i m inating condition, t(80) = 15.64, p < .001. As initial analyses o f our m a i n d e p e n d e n t measures indicated n o sex differences, the results r e p o r t e d in the following sections are collapsed across the sex variable.

438


Table 3

Mean Ratings of Self-Schema Indexes and Manipulation Checks as a Function of Depression Status and Experimental Condition: Study 2 Experimental condition Group

Schemadiscriminating

Schemanondiscriminating

2.89 1.71

1.88 1.84

2.44 2.52

2.29 2.26

3.22 2.14

2.77 2.53

2.82 2.86

2.82 2.68

Self-rating a Depressed Nondepressed Importance of trait dimension to self-view Depressed Nondepressed How well subjects expected to perform" Depressed Nondepressed Importance of performing well Depressed Nondepressed

Note. Lower numeric scores indicate higher or more positive ratings. Statistically significant interaction.

Self-schema indexes and manipulation checks. Consistent with Study l's results, we predicted that depressed and nondepressed subjects would differ in their self-ratings in the schemadiscriminating condition but not in the schema-nondiscriminating condition. A two-way ANOVA performed on subjects' self-ratings revealed the predicted Group × Condition interaction, F(1, 80) = 16.80, p < .001. Mean self-ratings for each group are presented in the first 2 rows of Table 3. Pairwise comparisons revealed that nondepressed subjects rated themselves significantly higher than did depressed subjects on the composite trait dimension (i.e., successful, motivated, productive) presented in the schema-discriminating condition, t(80) = 6.46, p < .00 I, but that group differences in self-ratings were nonsignificant on the composite trait dimension (i.e., polite, considerate, courteous) presented in the schema-nondiscriminating condition, t(80) = . 19. For a trait dimension to qualify as a schema, we required that it be at least moderately important to a group's self-view, as reflected by a mean importance rating of 3.0 or lower on the 5point scale used. As indicated in the second 2 rows of Table 3, this criterion was met by each participating subject group. In addition, to rule out interpretations of our results as being due to the differential importance of the trait dimensions used, we conducted a Group × Condition ANOVA on subjects' importance ratings. No significant interactions or main effects emerged from this analysis (all Fs < l), and all pairwise comparisons both across and within conditions were similarly nonsignificant (all Fs < l). As part of a manipulation check, we reasoned that if subjects actually regarded the experimental task to be a measure of their politeness or successfulness, as described in our cover story, then how well subjects expected to perform on the dot-counting task should correlate with their self-ratings on these respective dimensions. This correlation was evident in two ways. First, across all experimental conditions, subjects' performance ex-

pectancies were highly correlated with their self-ratings, r(82) = .61, p < .001. Second, a two-way Group × Condition ANOVA conducted on subjects' performance expectancies yielded a significant interaction, F(I, 80) = 6.70, p < .02, that paralleled the significant two-way interaction obtained for subjects' self-ratings (the parallel form of these interactions being confirmed by a nonsignificant Group × Condition × Rating triple interaction, F < l). The mean performance expectancies of each group are presented in the third 2 rows of Table 3. As an indication that our cover story was equally plausible in the two experimental conditions, we reasoned that subjects in the two conditions should not differ with regard to how important it was for them to perform well. A two-way Group × Condition ANOVAperformed on subjects' ratings of how important it was for them to perform well revealed the preferred absence of a significant interaction ( F < l) as well as the preferred absence of significant main and simple effects (Fs < l), both across and within conditions. Group means can be found in the last 2 rows of Table 3. Encoding of ambiguous feedback. Prior to these analyses, each subject's feedback identifications were assigned values of 1 through 5 for identifications ranging from very successful (or very polite) through very unsuccessful (or very rude), respectively. Then, for ambiguous and unambiguous trials taken separately, these individual values were summed and averaged. The resulting means or "identification scores" served as composite measures of a given subject's identification of ambiguous and unambiguous feedback. For the analyses discussed in the following paragraphs, it should be noted that the unbiased expected value averaging across all ambiguous feedback trials is 3.0. To test our hypothesis that group differences in the processing of feedback would be more pronounced under ambiguous conditions, we performed a Group × Condition × Feedback Clarity (ambiguous vs. unambiguous) repeated measures ANOVA on subjects' mean identification scores, with feedback clarity serving as the repeated measures variable. The three-way interaction proved significant, F( l, 80) = 5.82, p < .02. To analyze the three-way interaction, we conducted separate two-way Group × Condition ANOVASon each class of feedback. These analyses revealed a nonsignificant interaction, F(I, 80) = 2.32, as well as nonsignificant main effects (Fs < 1) for unambiguous feedback, in contrast to a significant interaction for ambiguous feedback, F(1, 80) = 5.07, p < .03. The nonsignificant effects obtained for unambiguous feedback support the view that feedback of this type is nonconducive to differential processing by depressed and nondepressed subjects. The significant Group × Condition interaction for ambiguous feedback, on the other hand, suggests that feedback of this type is amenable to differential processing, but only under certain conditions. That is, pairwise comparisons revealed that depressed subjects encoded ambiguous feedback more negatively than did nondepressed subjects in the schema-discriminating condition, t(80) = 2.88, p < .01, whereas depressed and nondepressed groups did not reliably differ in the schema-nondiscriminating condition, t(80) = .48. The patterning of this interaction (displayed in Table 4) coupled with the absence of a significant depressed-nondepressed main effect across experimental conditions, F(1, 80) = 2.35, suggest that depressive negativity in encoding is not characteristic, hut rather de-

SCHEMATIC BIASES AND DEPRESSION Table 4 Mean Identification Scores for Ambiguous Feedback as a Function of Depression Status and Experimental Condition: Study 2 Condition Group Depressed M SD Nondepressed M SD

Schemadiscriminating

Schemanondiscriminating

2.89 0.25

2.78 0.18

2.72 0.16

2.82 0.21

Note. The unbiased expected value is 3.0, the most positive attainable mean value is 2.5, and the most negative attainable mean value is 3.5.

pends on depressed individuals having a more negative schema than nondepressed individuals with regard to the task at hand. The data in Table 4 are also relevant to the normativeness of depressed and nondepressed subjects' encoding of ambiguous feedback. Initial evidence against the characteristic processing patterns proposed earlier is the fact that no subject group in this study was realistic (i.e., unbiased) in their encoding of ambiguous feedback. That is, one-sample t tests revealed that all groups showed significant positive biases relative to the unbiased expected value of 3.0 (all ps < .05), with depressed subjects in the schema-discriminating condition being the least positively biased group, nondepressed subjects in the same condition being the most positively biased group, and depressed and nondepressed subjects in the schema-nondiscriminating condition both falling in between. In direct support of schematic processing, on the other hand, is the fact that the rank ordering of means in Table 4 corresponds closely to the rank ordering of means for subjects' selfratings (i.e., self-schemas) as displayed in the first 2 rows of Table 3. This comparable rank ordering suggests that for both depressed and nondepressed subjects, the more positive the selfschema, the more positive the degree of bias. The correspondence between subjects' self-ratings and degree of bias was also supported at a more idiographic level by a significant positive correlation between individual subjects' self-ratings and their mean identification scores for ambiguous feedback, r(82)-.46, p < .001. Hence, on both the group and idiographic level, degree of bias in the encoding of ambiguous feedback appears directly related to the relative positioning of the self-schema. In addition to positing a direct relation between degree of bias and the relative positioning of the schema, schematic processing would also predict that both depressed and nondepressed subjects alike would show evidence of positively biased, negatively biased, and unbiased processing, with the direction of bias for each group converging toward that group's schema. An internal analysis of the encoding patterns of depressed and nondepressed subjects in each experimental condition allowed us to assess whether this prediction obtained. For each of the four ambiguous feedback pairs administered in this study, we computed the average amount by which each subject group deviated from an unbiased response (i.e., the average amount by

439

which responses to the 1-2, 2-3, 3--4, and 4-5 feedback pairs deviated from the unbiased values of 1.5, 2.5, 3.5, and 4.5, respectively). For each ambiguous feedback pair, mean deviations from an unbiased response and the direction of those deviations are listed in Figure 1. Also mapped onto Figure 1 is the mean self-rating (i.e., self-schema) of each subject group. By inspection, the data displayed in Figure 1 show a pattern in which both depressed and nondepressed subjects in each experimental condition both positively and negatively biased their encoding of ambiguous feedback, with this biasing occurring in the direction of each group's respective schema. Interestingly, further inspection of Figure 1 indicates that for each group, the greater the difference between the schema and the feedback pair being processed, the greater the magnitude (hence, the probability) of bias. This tendency for all groups to probabilistically bias their encoding of ambiguous feedback in the direction of their respective schemas was supported by a highly significant linear trend in the deviation means, F(l, 80) = 313.99, t7 < .00 l, and by the absence of a significant Group × Linear Trend interaction (F < 1). Moreover, as the lettered values in Figure 1 indicate, a number of the individual deviation means of both depressed and nondepressed subjects alike reached significance (as revealed by one-sample t tests), indicating that both subject groups showed significant positive biases and significant negative biases in their encoding of various individual ambiguous feedback pairs.5 Unbiased encoding also typified each subject group and, in further support of schematic processing, unbiased encoding for three of the four subject groups occurred at the point where a particular ambiguous feedback pair and a group's schema were most nearly matched in valence. Response latencies to unambiguous feedback. To reduce random variance in response latencies attributable to the initial novelty of the task, we did not include subjects' response latchties to the first 5 unambiguous feedback trials in the final data analysis. Mean response latencies to each unambiguous feedback type as well as each group's mean self-ratings (i.e., self-schemas) are depicted in Figure 2. To analyze the patterning of means in Figure 2, we conducted a 4 × 5 (Group × Feedback Type) repeated measures ANOVAon subjects' response latencies, with feedback type serving as the repeated measures variable. In contrast to a nonsignificant group effect (F < 1), the quadratic trend across the repeated measures variable was highly significant, F(1, 80) = 33.92, p < .001. To determine whether this quadratic trend reflected the occurrence of schematic processing, we inspected the patterning of means for each subject group. In strong support of schematic processing, this inspection revealed that for depressed and nondepressed subjects in the schemanondiscriminating condition, and for nondepressed subjects in 5 Note that "significant" negative biases were not displayed by depressed subjects in the schema-nondiscriminating condition or by nondepressed subjects in the schema-discriminating condition. We point out that such attenuation in negative biases is entirely consistent with the schematic processing view as the schemas of these groups in these conditions most closelymatched the most positive feedback pair administered in this study. On the basis of the pattern of data, we conjecture that administering even more positive ambiguous feedback (e.g., super successful-very successful) would have resulted in significant negative biases in these conditions, as well.

440


A.

B. Feedback Pair Nondepressed

Depressed

1

1

Nondepressives' ~

s~-~]~ 2 ~ ~ (2.89.)

LJepresslves'

Feedback Pair Nondepressed

"z.--J

Se~schen~ 4 ~

a

Nondepressives' ~

~ :7-062 ...............7~130....... Self-sch¢~ (1.84) ................................... (1.8~) 2~.. Depressxves' Self-schema / . ~ +.320c -.004 3-.... +'424c +'270c /

h

-- 174"

Depressed

- 124

...................... "........................................................

+. 122

c

" ..............................

+. 194 b c

+.364

+.394

+.426

+.406

4

. ~ 5/

+'452c

+.292c 5

"

Figure 1. Mean deviations from an unbiased response for each ambiguous feedback pair. (A = schemadiscriminatingcondition; B = schema-nondiscriminating condition. Negative (-) and positive (+) deviation scores represent negative and positive biases, respectively. Superscript a signifies p < .05, one-tailed; superscript b, p < .05, two-tailed;, and superscript c, p < .001, two-tailed.)

the schema-discriminating condition, response times were, in fact, fastest to the feedback stimulus lying closest to each group's schema and progressively slower to feedback stimuli lying increasingly farther from each group's schema. The group that did not follow a schematic processing pattern, in the strict sense, was depressed subjects in the schema-discriminating condition. We would expect this group to respond fastest to the Number 3 feedback stimulus as the Number 3 stimulus is the closest match to their schema. Instead, this group responded fastest to the Number 2 stimulus, as did the other groups ( F < l, for the Group X Quadratic Trend interaction). It is nevertheless noteworthy that except for the reverse ordering of these two points, depressed subjects in the schema-discriminating condition did conform to the general quadratic trend indicative of a schematic processing effect. Across the four subject groups, the marginal means for reaction times to the Numbers l, 2, 3, 4, and 5 unambiguous feedback types were 2.39, 2.16, 2.26, 2.37, and 2.58 s, respectively. The tendency for all groups to conform to this quadratic trend is strong evidence that neither depressed nor nondepressed subjects were characteristically biased or unbiased in their response latencies to unambiguous feedback. Summary recall. For each subject, we calculated a summary recall score by multiplying the number of each unambiguous feedback type recalled, that is, very successful (and very polite) through very unsuccessful (and very rude), by values of 1-5, respectively. These products were then summed across the five unambiguous feedback types, with the resulting total serving as a summary recall score. As subjects were instructed that their

total recall of unambiguous feedback stimuli should sum to 25, a summary recall total of 75 represented accurate or unbiased recall. A two-way Group × Condition ANOVA performed on subjects' summary recall scores revealed that neither of the main effects, nor their interaction, proved significant (all ps > .20). In addition, one-sample t tests revealed that only depressed subjects in the schema-discriminating condition deviated significantly, and in a negatively biased direction, from the unbiased value of 75, t(26) = 2.12,p < .05. The absence of positive recall biases for any subject group (as one would predict on the basis of the positively valenced schemas held by all subject groups) and the absence of a depressed-nondepressed main effect for recall across experimental conditions ( F < 1) are evidence, respectively, that neither schematic nor characteristic processing biases occurred. Task performance. As pointed out earlier, dot displays were equated across conditions. To assess whether obtained encoding effects were possibly related to the differential dot-counting accuracy of the different subject groups, we summed subjects' dot estimates across trials and subjected the resulting totals to a two-way Group x Condition ANOVA. No significant effects emerged from this analysis.

General Discussion In two important respects, our findings strongly challenge the view that depressed and nondepressed individuals process information in characteristically different ways. First, depressed

SCHEMATIC BIASES AND DEPRESSION A

441

2.7

d 111 (n

2.6'

(n iii mm

2.5

0 Z iii

!