Depressive symptoms are associated with unrealistic negative predictions of future life events

Abstract

This study examined the relationship between depressive symptoms and bias in the prediction of future life events. Responding to internet announcements, 153 participants varying widely in self-reported depression symptom severity estimated the probability of 40 events occurring over the succeeding 30 days. After the 30-day period, participants reported which events occurred. Optimistic/pessimistic biases were related to level of depressive symptoms. A non-significant optimistic bias characterized participants with low depressive symptoms whereas a significant pessimistic bias characterized participants with high depressive symptoms. Those reporting mild symptoms did not exhibit a systematic pessimistic or optimistic bias. General imprecision in predictions for undesirable events was associated with depressive symptoms. These findings suggest that depression is associated with pessimistic bias rather than accuracy in judgment.

Introduction

Beck and his associates propose that depressed people have inaccurate, negative views. Beck asserts that depressed people have cognitive distortions and negative views of the self, world, and future (1967, 1976). Depressed people do report more negative automatic thoughts, more dysfunctional attitudes, more hopelessness, and a more pessimistic explanatory style than people who are not depressed (Beck, Riskind, Brown, & Steer, 1988; Hill, Oei, & Hill, 1989; Hollon, Kendall, & Lumry, 1986; Peterson & Seligman, 1984). They have also been shown to process negative information more thoroughly and efficiently compared to people who are not depressed (Dozois & Dobson, 2001; Gotlib, Krasnoperova, Yue, & Joormann, 2004). However, Beck's original proposal is that depressed people are not merely more negative than people who are not depressed, but they are inaccurate in their negative views. Testing this claim has proven difficult. A definitive body of empirical work has yet to emerge in this area, despite substantial efforts by several research groups.

Alloy and Abramson (1979) introduced the depressive realism hypothesis with their report of a series of experiments involving participants with varying levels of depressive symptoms. Their basic experimental procedure involved having college students participate in a series of 40 trials, on each of which participants chose either to press or not press a button, and a green light either did or did not light. The experimenters varied the probabilities of the appearance of the green light after the button was pushed, and not pushed. Alloy and Abramson defined the true degree of control in this situation as the probability of the green light being lit when the participant did press the button minus the probability of the green light being lit when the participant did not press the button. Following these trials, participants were asked to estimate the degree of control their button pressing had over the lighting of the green light. Alloy and Abramson sought to characterize accuracy in judging degree of control over the green light as a function of the level of participants’ depressive symptoms.

Although Alloy and Abramson (1979) referred to their dysphoric group as depressed, we use the term dysphoric because participants were drawn from a presumably largely non-depressed sample. While depressed college students would not have been excluded in their studies, the rate of depression in this population along with the relatively low cutoff score they used for the more depressed group suggests that only a very small percentage of their participants were likely to have had moderate to severe depressive symptoms. Alloy and Abramson identified three experimental situations in which a non-dysphoric group exhibited what Alloy and Abramson interpreted as an optimistic bias in their contingency estimates, a bias not evident in their dysphoric group. For example, in one of their experiments, participants had some control (50%) over either winning ten cents per trial or losing 30 cents per trial. The non-dysphoric group underestimated the extent to which they were responsible for their losses, and no bias was evident in the dysphoric group. In all of Alloy and Abramson's experiments, whenever any bias was evident, the non-dysphoric participants’ judgments reflected an optimistic tendency that was not present among the dysphoric participants.

These findings have led to a great deal of further consideration and empirical work. And, two versions of the depressive realism hypothesis have now been identified (Ackermann & DeRubeis, 1991). Both versions allow for optimism among people with the lowest levels of depressive symptoms. However, the hypotheses differ in their predictions for people with high levels of depressive symptoms. The first hypothesis, which we shall refer to as the depressive bias hypothesis, states that as depressive symptoms increase in severity judgments become more negatively biased. Under this hypothesis, people with the highest levels of depressive symptoms are expected to exhibit substantial bias and unrealistic pessimism (Beck (1967), Beck (1976)). The second hypothesis, which we shall refer to as the depressive accuracy hypothesis, states that as depressive symptoms increase judgments become more and more accurate, such that the judgments of people with the highest levels of depressive symptoms would be the most accurate. These two hypotheses are depicted in Fig. 1. While many psychologists have interpreted existing research to support the depressive accuracy hypothesis, there is reason to doubt the validity of this conclusion.

Because of a number of common limitations, many of the investigations of depressive realism have been inadequate to address whether and to what extent cognition among depressed people is negatively biased, and how any such bias compares to that exhibited by dysphoric or non-depressed people. Several reviewers have highlighted three common limitations of previous research (Ackermann & DeRubeis, 1991; Dobson & Franche, 1989; Haaga & Beck, 1995).

As Ackermann and DeRubeis (1991) have reviewed in detail, one limitation in previous depressive realism work is that many studies have failed to use objective standards in assessing potential bias in judgments. For example, Alloy and Abramson (1988) have suggested that in studies of attributional style dysphoric and depressed “scores for positive and negative events for each dimension did not differ significantly from a score of 0, indicating perfect even-handedness” (p. 234). However, there is no evidence that would allow one to relate the Attributional Style Questionnaire scores to objective reality (Ackermann & DeRubeis).

Other studies have used measures that were likely to have been influenced by systematic bias. For example, Lewinsohn, Mischel, Chaplin, and Barton (1980) studied judgments of people diagnosed with depression, people with other psychiatric conditions, and people not diagnosed with a psychiatric condition following a small group interaction. They examined group differences in the discrepancies between self and observer ratings of 17 desirable characteristics. While participants were simply asked to rate themselves on the 17 characteristics based on their interactions in the group, the standard used to assess their accuracy was observer ratings. Ackermann and DeRubeis (1991) noted this discrepancy, that observers’ ratings have been shown to be critical, and that “it may be the case that the judges’ harsh ratings simply happened to coincide with the depressed subjects’ low self-ratings” (p. 572). Thus, a number of studies have assessed depressive realism without an objective standard, or with a standard that is likely to have been contaminated by some systematic bias. In the studies in which this limitation was not present, other methodological problems have often clouded the interpretation of results.

A second limitation is that many researchers have assumed that findings based on dysphoric participants will generalize to clinically depressed patients. Researchers have argued that utilizing dysphoric participants to study depression may be inappropriate for a number of reasons (Coyne, 1994; Coyne & Gotlib, 1983). Reviewers of the depressive realism literature have repeatedly lamented the heavy reliance on dysphoric college student samples (Ackermann & DeRubeis, 1991; Dobson & Franche, 1989; Haaga & Beck, 1995). Nevertheless, the judgments of dysphoric people may be of interest in their own right. Researchers might include dysphoric people specifically to investigate the degree of bias among people with mild depressive symptoms rather than as a stand-in for a clinically depressed sample.

Only a few studies have employed samples that include people with a wide range of depressive symptoms. For example, Dobson and Pusch (1995) included depressed people in a non-contingent computerized version of Alloy and Abramson's (1979) procedures. Surprisingly, depressed and non-depressed groups did not differ. Using Alloy and Abramson's paradigm, Carson (2001) found that while depressed participants were consistently more negative via their lower contingency estimates, they were not consistently more accurate. Whereas the non-depressed participants were more accurate in the contingent situation, depressed participants were more accurate in the non-contingent situation. Thus, while the evidence is somewhat mixed, the evidence for the depressive accuracy hypothesis among people who are clinically depressed appears substantially weaker than the evidence generated from studies of dysphoric college students.

A third common limitation of studies of cognition among depressed and non-depressed groups is the use of artificial laboratory tasks. Reviewers have noted a tendency for artificial laboratory tasks to support depressive realism while more naturalistic, ecologically valid studies have not tended to do so (Ackermann & DeRubeis, 1991; Dobson & Franche, 1989). Beck et al.'s (1979) descriptions of depressed people's judgments have emphasized emotionally salient judgments related to personal worth. Less ecologically valid studies may not emotionally involve participants sufficiently to test Beck's ideas. Consistent with this idea, Pacini, Muir, and Epstein (1998) have provided some evidence that consequential information is processed differently among dysphoric and non-dysphoric people.

The common methodological limitations of previous work make conclusions about the validity of the depressive accuracy hypothesis premature. The best tests of the cognitive model of depression (and the depressive realism hypothesis) should investigate judgments most relevant to the model. Specifically, these tests should use procedures that examine consequential, emotionally involving judgments relevant to participants’ self-worth. Moreover, studies must be conducted in which participants having a wide range of depressive symptom severity and make judgments which are compared to an objective standard to ascertain participants’ accuracy.

A study of life event prediction meets these criteria well. In a longitudinal design, participants can report whether events occurred, thereby providing a reasonably unbiased standard by which the accuracy of predictions can be judged. Moreover, the prediction of life events is an everyday, naturally occurring phenomenon. According to the cognitive model of depression, predicting future life events is a theoretically important task. Negative expectations for future events are thought to be a hallmark of depression (Beck (1967), Beck (1976)). Patients’ negative predictions for future events are also an important target of cognitive therapy (Beck, Rush, Shaw, & Emery, 1979; Hollon & Garber, 1980). Thus, a study of life event predictions among non-depressed, dysphoric, and depressed participants would provide a test of optimistic or pessimistic bias that utilizes an objective standard, is minimally artificial, and provides a clear test of the cognitive model of depression.

Dunning and Story (1991) examined potential differences in predictions of future life events among groups with different levels of depressive symptoms. In their study, dysphoric participants’ probability estimates for life events were found to be less accurate than those of non-depressed participants. Surprisingly, this difference in accuracy was attributable to dysphoric participants being more optimistic than non-depressed participants. This appeared to be due to dysphoric participants failing to anticipate that they would experience fewer desirable events and more undesirable events than non-depressed participants. Although the results reported by Dunning and Story are surprising, the basic methodology when applied to a sample that includes people with more severe depressive symptoms appears appropriate. To our knowledge, no study has addressed the key claims of the cognitive model of depression by investigating bias in life event predictions among people with more severe depressive symptoms.

The purpose of this study is to investigate the relationship of depressive symptoms and pessimistic or optimistic bias while minimizing problems identified with previous research. We investigated this relationship in the context of people estimating the probability of future events in their lives. We endeavored to address four specific issues. First, we hypothesized that participants’ tendency to predict positive outcomes (i.e., desirable events occurring and undesirable events not occurring) would be inversely related to the severity of their depressive symptoms. Second, we predicted that the degree to which participants experienced positive outcomes would be inversely related to depressive symptoms. Third, our primary research hypothesis involved the relationship between pessimistic/optimistic bias and depressive symptoms. This pessimistic/optimistic bias is the bias relative to events that occurred for each participant. We expected that pessimistic bias would characterize the depressed participants, optimistic bias would characterize the non-depressed participants, and a little or no bias would characterize the dysphoric participants. We also planned to explore whether this relationship differed between men and women. Finally, we explored whether depressive symptoms were associated with participants’ ability to use their predictions to discriminate between more and less likely events regardless of any optimistic or pessimistic bias they exhibited. That is, we planned to examine the association of depressive symptoms and imprecision in participants’ predictions (regardless of their optimistic or pessimistic tendencies).