Self-Consistency Bias: Why Past Opinions Look Like Present Ones

A past attitude is not stored like a photograph and read back; it is reconstructed when it is asked for, and the reconstruction starts from the present. Michael Ross argued in 1989 that people rebuild their personal histories using implicit theories about how much they change. The default theory for attitudes, traits and feelings is stability: I think this now, I am the kind of person who thinks this, so I probably thought it then. Where the present view is clear and the past one is faint, the present fills the gap, and the result is a past that agrees with the present more than the record does.

The same mechanism runs the other way when the implicit theory predicts change. Conway and Ross found in 1984 that students who had completed a study-skills program recalled their skills before the program as worse than they had rated them at the time, in effect manufacturing an improvement the course had not produced. A theory of change makes the past look different from the present; a theory of stability makes it look the same. Consistency bias is the stability case, and it is the common one for opinions, because most people believe their opinions are their own and reasonably fixed.

Panel studies show the pattern outside the laboratory. Gregory Markus reported in 1986 on respondents in a national survey who had given their views on political issues in 1973 and were asked in 1982 both what they thought now and what they had thought nine years earlier. The recalled 1973 attitudes lined up more closely with the respondents' 1982 attitudes than with the answers they had actually given in 1973. People who had changed their minds mostly did not remember doing so.

Daniel Schacter lists this under bias in his 1999 taxonomy of memory's seven sins, where consistency bias sits beside hindsight bias and egocentric bias as ways current knowledge rewrites the past. It is not a sign of dishonesty. The students in Goethals and Reckman's study were not covering their tracks; they believed their recalled attitudes, and the reconstruction felt like remembering. One visitor's two sliders cannot separate a genuinely stable opinion from a reconstructed one, which is why the result above reports the gap between your answers and nothing more.

George Goethals and Richard Reckman published "The perception of consistency in attitudes" in the Journal of Experimental Social Psychology in 1973. The participants were high school students who had filled in an attitude questionnaire that included items on the busing of schoolchildren to achieve racial integration, a live political issue in the United States at the time. From their answers the researchers selected students who were clearly for busing and students who were clearly against it.

Each group was then brought back for a small discussion of the issue. One member of each discussion group was a confederate, a high school student working with the experimenters, who had been armed with persuasive arguments against the position the group held. The anti-busing groups heard the case for busing; the pro-busing groups heard the case against. After the discussion, participants completed the attitude questionnaire again, and their positions had shifted toward the confederate's arguments.

The critical measure came next. Participants were asked to recall, as accurately as they could, the answers they had given on the original questionnaire. Those whose attitudes had changed recalled their original positions as having been close to the position they now held. The recorded first answers told a different story: the students had moved, and their memory of where they started had moved with them. A control group whose attitudes were not challenged recalled their original answers accurately.

Goethals and Reckman's finding has been the standard demonstration of consistency bias since, and Ross's 1989 review built the implicit-theories account around it. The effect has also been found with memories of past feelings and behavior, and in survey panels where recalled and recorded attitudes can be compared across years; the studies agree on the direction while the size varies with how clear the original attitude was and how long ago it was held. This page does not report figures from the 1973 paper beyond its design, because the specific percentages were not available to check.

The rows below are illustrations of where a past view gets rebuilt from a present one, not findings from the studies cited on this page.

The bias lives in reconstruction, so the counters replace reconstruction with records or make the reconstruction visible.

• Write it down at the time. A dated note of what you think before a decision, a debate or a purchase is the only version of your past opinion that cannot drift. Diaries, old emails and public posts serve the same purpose.
• When you compare then and now, name the theory you are using. If you are assuming you have not changed, say so, and ask what would have changed your mind in the meantime.
• Look for the moment of change. People who recall changing their minds usually recall the occasion; if you cannot find one, it may be that there was none, or that the change was gradual and the memory of the old view has been overwritten.
• Treat agreement between your recalled past and your present as weak evidence. It is what stability would produce, and it is also what the bias would produce.
• In research and in reviews, ask at both times. Retrospective questions about attitudes measure present attitudes with a time stamp; a panel design measures the change.
• Be generous with others. Someone who insists they always held their current view is probably remembering sincerely, and a record settles it better than an argument.

Several biases rewrite the past from the present, and they are often confused.

• It is not hindsight bias, which rewrites past predictions once the outcome is known: I knew it would happen. Consistency bias rewrites past attitudes and feelings to match present ones, with no outcome involved.
• It is not rosy retrospection, which makes the past look better than it was. Consistency bias makes it look like the present, better or worse.
• It is not choice-supportive bias, where the options in a past decision are remembered in a way that favors the one chosen. That one is about the features of options; this one is about your own earlier state of mind.
• It is not cognitive dissonance reduction, the motivated process of bringing beliefs in line with actions. Consistency bias needs no motive; Ross's account treats it as an ordinary inference from a theory of stability, and it appears in people with nothing at stake.