Biases of Retrospection?¹

Robyn M. Dawes^*

Memory belongs to the imagination.  Human memory is not like a computer which records things; it is part of the imaginative process, on the same terms as invention (Alain Robbe-Grillet, 1986).
  

While memory from our experience is introspectively a process of "dredging up" what actually happened, it is to a large extent determined by our current beliefs and feelings.  This principle has been well established both in the psychological laboratory and in surveys.  What we have at the time of recall is, after all, only our current state, which includes fragments ("memory traces") of our past experience; these fragments are biased by what we now believe (or feel) to be true to an extent much greater than we know consciously.  Moreover, the organization of these fragments of past experience into meaningful patterns is even more influenced by our current beliefs and mood — especially if we are particularly depressed or elated.

Memory is basically a "reconstructive" process.  Thus, our experience is often recalled inaccurately, even that selectively biased and possibly irrelevant experience discussed in the previous sections.  The problem is particularly acute because our recall is often organized in ways that "make sense" of the present — thus reinforcing our belief in the conclusions we have reached about how the past has determined the present.  We quite literally "make up stories" about our lives, the world, and reality in general.  The fit between our memories and the stories enhances our belief in them.  Often, however, it is the story that creates the memory, rather than vice versa.

For example, Greg Markus (1986) studied stability and change in political attitudes between 1973 and 1982.  Specifically, a national sample of 1,669 high school seniors in the graduating class of 1965, along with at least one parent in nearly every case, was surveyed in 1965, 1973, and 1982.  Fifty-seven percent of the parents (64% of those still living) and 68% of the students (70% of alive) were personally interviewed all three times.  All subjects were asked to indicate on a seven-point scale (with verbal anchors at the end) their attitudes towards five issues: guaranteed jobs, rights of accused people, aid to minorities, legalization of marijuana, and equality for women.  In addition, they were asked to characterize their political views as generally liberal or generally conservative.  Most important for analysis of the retrospective bias, Markus asked the respondents in 1982 to indicate how they had responded to each scale in 1973.

The results were quite striking.  With the exception of the ratings on the overall liberal-conservative scale, the subjects' recall of their 1973 attitudes in 1982 was more closely related to their rated attitudes in 1982 than to the attitudes they had actually expressed in 1973.  Retrospecting, they believed that their attitudes nine years previous were very close to their current one, much closer than they in fact were.  This bias was so strong that an equation set up to predict subjects' recall of their 1973 attitudes gives almost all weight to their 1982 attitudes, and virtually none at all to the attitudes they actually expressed in 1973 (with the important exception of the students' overall liberal versus conservative ratings).

In addition, what discrepancy there was between 1982 attitudes and recall of 1973 attitudes could primarily be explained in terms of stereotypic beliefs about how general attitudes in the culture had changed; the subjects believed that they had become more conservative in general, but that (again in general) they had favored equality for women all along.  Subjects whose attitude had changed in the direction counter to the general cultural change tended to be unaware of such change.  Finally, the parent group attributed much more stability to their attitudes than did the student group, which is compatible with the belief that the attitudes of older people change less.  In fact, however, the attitudes of the parent group were less stable.

Attitudes are, of course, somewhat amorphous and difficult to determine.  Linda Collins and her colleagues found quite similar results for actual behaviors when they surveyed high school students about their use of tobacco, alcohol, and illegal "recreational" drugs (Collins, Graham, Hansen, & Johnson, 1985).  They repeated the survey after one year and again after two and one-half years.  At each repetition, the students (many of them then in college) were asked how much usage they had reported on the original questionnaire.  (Collins and her colleagues had established strong rapport with this group and had reason to believe that their guarantees of confidentiality, which they honored, were in fact believed.)  Again, the subjects' belief in lack of change introduced severe retrospective bias.  For example, the recall of alcohol use for those subjects whose drinking habits had changed over the two-and-one-half-year period was more highly related to their reported use at the time of recall than to the reports they had made two and one-half years earlier.

Thus, change can make liars of us, liars to ourselves.  That generalization is not limited to change in an undesirable direction.  As George Vaillant (1977), who has studied the same individuals for many years throughout their adult lives, writes: "It is all too common for caterpillars to become butterflies and then to maintain that in their youth they had been little butterflies.  Maturation makes liars of us all."

But not always.  Sometimes, when our belief is in change, we recall change even when it has not occurred.  In order to make our view compatible with this belief, we resort (again not consciously) to changing our recall of the earlier state.  We can, for example, reinforce our belief in a nonexistent change for the better by simply exaggerating how bad things were before the change.  Certainly there have been times before a religious or psychiatric conversion, for example, when the individual was badly off (we all are at times), and memories of those times persist; recall can be organized around the traces of these memories.  A dieter who has not succeeded in losing a single pound can certainly recall periods of time prior to embarking on a diet when he or she was heavier than when he or she completes the ineffective diet; by carefully not recording his or her weight before starting the diet, those times can be recalled as an evidence for its success.

Experimental evidence supports the contention that when we believe a change has occurred we are apt to distort the past in the direction compatible with the change.  For example, in two separate but similar experiments, Conway and Ross (1984) randomly selected participants for a university program designed to improve study skills and a control group of students who had indicated a desire to be in the program and were on the waiting list for it.  Participants and controls were questioned before the study skills program began and at its conclusion.  At both times they were asked to assess their study skills (e.g., how much of their study time was well spent, how satisfactory their note-taking skills were, etc.) and the amount of time they studied.  At the second interview they were also asked to recall what they reported during the first session concerning their skills and study time.

At the initial interview, participants and controls did not differ significantly on any measure of skill, study time, or other variables.  Both groups performed equally well and — most important to the study — the program itself was not found to improve study skills.  Nor did it improve grades.  When asked to recall their situations before the program started (or before they were put on the waiting list), however, the subjects did differ.  There was no difference between the two groups in their memory of the amount of time they spent studying, but their recall of their skills was markedly different.  Program participants recalled their study skills as being significantly worse than they had initially reported, while on the average, waiting-list subjects recalled their skills as being approximately the same as they had reported initially.  Thus, program participants exaggerated their improvement in a direction consistent with their beliefs of what ought to be (improved skills due to taking the course), not by exaggerating their current skills, but rather by reconstructing their memory of the past to fit with the belief that they should have improved.  In short, they recalled themselves as having been worse off before they entered the program than they had in fact been.  There was no such distortion on the part of the subjects who had been put on the waiting list.

Mood also affects recall.  It has, for example, been strongly established experimentally by Gordon Bower (1981) and others that recall of material learned in a particular mood is facilitated by recreation of that mood.  Does the same principle apply to our recall of our own lives?  Is our recall of events that occurred when we were in a bad mood — which are usually negative events — facilitated by a current bad mood, and vice versa for good moods?  The answer is yes.

Lewinsohn and Rosenbaum (1987) studied the recall of parental behavior by acute depressives, remitted depressives (that is, people who had once been depressed and were no longer depressed), nondepressives (people who had never been depressed), and "predepressives" (people who were to become depressed) in a group of 2000 people over a three-year period.  One focus of this research was on the relationship between current mood states and memory; one possibility is that recollections of one's parents are influenced by a current state of depression or nondepression; another is that people who are prone to depression recall their parents differently from those who are not (the nondepressives).  Theories that depression follows from childhood problems would predict that the childhood of those of us who are depression-prone is different from that of those who are not and hence would be recalled differently, while theories about the effect of current mood on past recall predict that the primary difference in recall should be between people who are currently depressed and those who are not.

The results were consistent with the hypothesis that recollection of one's parents as rejecting and unloving is strongly influenced by current moods; it is not a stable characteristic of depression-prone people.  "Whereas the currently depressed subjects recalled their parents as having been more rejecting and as having used more negative controls than the normal controls, the remitted depressives did not differ from the never depressed controls in their recall of parental behavior.  Similarly, the subjects who were about to become depressed shortly after the initial testing did not differ from the controls in their recollections of the degree to which their parents used negative control methods (Lewinsohn & Rosenbaum, 1987).  One particularly important aspect of this study was that the subjects were drawn from the general population; they were not sampled on the basis of having any particular psychiatric problems.

(Note: There is no way of determining the accuracy of this recall.  It is possible, for example, that when the depression-prone people become depressed, their memory of their parents' behavior loses the "rosy glow" that appears to be a concomitant of "good mental health," and that the reports are in fact more accurate than when the people are not depressed.  Some evidence supports that depressed people are more accurate in their perceptions — of themselves anyway — than are people judged to be in good psychological health; these latter people tend to become "Pollyannish" as they get over their depression.  See Lewinsohn, Michel, Chaplin, and Barton, 1980.)

This study of depression is particularly important in that it casts doubt on the degree to which adult problems are related to childhood ones.  Given a biasing effect of mood on memory, people who are distressed as adults tend to remember distressing incidents in their childhood.  One result is the view that the sources of the problems encountered lie in early life is reinforced.  To the degree to which the people accept this view, it may serve as an organizing principle for even greater distortion of recall, which in a circular way reinforces the "child is father to the man" view of life.  (Freud himself emphasized that he knew people who had childhood problems similar to those of his patients but who never became distressed.  One view of Freudian psychology is that when adults become distressed, the form of the distress will mirror childhood problems.  That is not the same, however, as saying that these childhood problems cause adult distress.  These problems may even be necessary for adult neurosis and psychosis, but again that does not make them sufficient.  The idea that childhood problems necessarily lead to adult ones [i.e., are sufficient to cause adult problems] is due more to the neo-analytic followers of Sigmund Freud than to Freud himself — particularly those who have popularized their view of neo-Freudian psychology; see, for example, Harry Overstreet's book The Mature Mind.)

Finally, our retrospective bias that we (usually) haven't changed can lead us to expect that we will not change with changing circumstances; in particular, that our intentions and motives — which serve as background in our judgment of our experience — will not change.  For example, consider the statement, "You have nothing to fear from me now that I have this power over you; I have always been benign."  (Reagan administration officials assert that the Soviets have nothing to fear if "we" develop an invulnerable "star wars" defense, because "our" intentions have always been peaceful, mainly defensive.  First, we must remember that we are not always the best judges of our own intentions, particularly not of what they have been in the past.  Second, intentions can change as capabilities do, and the person who changes is often "the last to know."  Third, and most importantly, the Soviets have no way of knowing who "we" in charge of U.S. policy will be when and if such a defense system is ever perfected; nor do we.)
  

References and Notes

Bower, G. (1981). Mood and memory. American Psychologist, 36, 129-148.

Collins, L. N., Graham, J. W., Hansen, W. B., & Johnson, C. A. (1985). Agreement between retrospective accounts and substance use and earlier reported substance use. Applied Psychological Measurement, 9, 301-309.

Conway, M., & Ross, M. (1984). Getting what you want by revising what you had. Journal of Personality and Social Psychology, 47, 738-748.

Lewinsohn, P. M., Michel, W., Chaplin, W., & Barton, R. (1980). Social competence and depression: The role of illusory self-perceptions? Journal of Abnormal Psychology, 89, 203-212.

Lewinsohn, P. M., & Rosenbaum, M. (1987). Recall of parental behavior by acute depressives, remitted depressives, and nondepressives. Journal of Personality and Social Psychology, 52, 611-620.

Marcus, G. B. (1986). Stability and change in political attitudes: Observe, recall, and "explain." Political Behavior, 8, 21-44.

Overstreet, H. (1949). The Mature Mind ()()(). New York: W. W. Norton. See particularly Chapter 10, "The home as a place for growth." Earlier, Overstreet alleges: "The human individual is a fairly tight-knit pattern of consistency" (p.73).

Robbe-Grillet, A. (1986). The art of fiction XCI. The Paris Review, Spring, 46.

Vaillant, G. E. (1977). Adaptation to Life (). Boston: Little, Brown and Company.