, , , , , , , ,

It’s not unusual for people to think that an outcome was inevitable after the fact, a phenomenon known as “hindsight bias”. This remains the case even when test subjects are instructed to ignore their knowledge of the outcome and rate the likeliness of several possible outcomes. Once an outcome is known, people tend to see causal links in antecedents (the events preceding the outcome); this builds into a causal chain that makes the outcome seem an inevitable consequence of the antecedents. Studies have shown that, in addition to constructing such links, people also tend to exercise selective memory, preferentially remembering details that are consistent with the seemingly inevitable outcome.  Linda Carli (of Wellesley College) conducted a study extending these findings even further. She found that people will “misremember” additional antecedents that are stereotypically associated with a given outcome; these novel memories reinforce the hindsight bias – their perception that the outcome is inevitable – which affects their judgement of the behaviour and character of the individuals involved. These findings shed light on how & why we judge that people “get what they deserve”, often over-simplistically attributing credit for success and unfairly blaming victims.

Carli conducted two studies to test how knowledge of an outcome affects our memory, judgement and perception of a series of events; the first study examined the effect of a negative outcome, while the second included both a positive and negative outcome. The participants in the study were told they were taking part in a study of how people form impressions; they were each given an invented story which they thought was an interview with a woman and told not to discuss it with the others. In each study, the participants were divided into two groups who received different endings to the story, which was otherwise identical. Some time after reading the story, they were given a memory test (without fore-warning); the memory test included both actual events from the story and additional events which hadn’t been in the story but were stereotypical for a given ending. These “stereotypical antecedents” were generated by a group of four students and then evaluated as stereotypes by a larger survey. In addition, a questionnaire was used to determine how inevitable they judged the outcome to be (hindsight bias) and to assess their opinion of the behaviour and character of the individuals in the story.

In the first study, the “case study” was an interview with Pam, who described her experience with Peter. Pam ran into Peter, an acquaintance, while out at a club with friends. They talked and danced and she gave him her number. He called her later and they went out on a date; Pam wore a new outfit that she thought he would like. After the date, Peter said he wanted to stop at his apartment before taking her home and she agreed. For half of the participants (the “no-ending group”), the story ended here; for the other half (the “rape-ending group”), there was an additional sentence stating that Peter raped Pam while they were at his apartment.

After one week, the participants were given the questionnaire and memory test. Both groups showed hindsight bias, rating the ending they had been exposed to as more likely on the questionnaire. Although both groups remembered events from the story equally well, the rape-ending group also remembered more rape-stereotypical antecedents that weren’t in the story, such as Pam having a voluptuous figure, dressing suggestively or having only a high school education or Peter having a blue collar job or not having had sex in a while. A statistical analysis confirmed that the novel memories were responsible for generating the hindsight bias, rather than being a product of it. In other words, the novel, reconstructed memories don’t result from the inevitability of the outcome; rather, by reinforcing the apparent antecedent causal chain, they reinforce the impression that the outcome is inevitable.

The rape-ending group gave a more negative rating of Pam’s character and behaviour than the no-ending participants. There was also a negative correlation between the assessment of her character and the likelihood of rape – that is, the more likely (or inevitable) a participant considered the rape outcome, the worse they judged Pam’s behaviour (and character). Perhaps unexpectedly, none of these results was affected by the gender of the participants – men and women similarly blamed Pam for behaving in a manner that would “inevitably” lead to a rape scenario when, in fact, the seeming inevitability is only an artifact of their reconstructed memories of the events in the story.

The second study was similar, but included both a positive and negative ending to the story. In this case, the “case study” was an interview with Barbara about her relationship with Jack. The two met in a graduate school business class and were assigned to work on a project together. After doing their work, they would go out for coffee and hang out. Over the course of the semester, they became close. The story included two accounts of Jack losing his temper, once with a professor and once when he yelled at Barbara. At the end of the semester, they went out for drinks to celebrate. They stayed out all night and Jack invited Barbara to his parents’ ski lodge. They kissed during their first night at the lodge; the following evening, Jack took Barbara out to a special restaurant, for which she wore a new outfit. On the way back to the lodge, they held hands and said sweet things to each other. Depending on the version of the story, Jack then either raped Barbara or proposed marriage. In order to make sure that both endings were plausible, a pre-test was conducted with a different group of participants, who each read a version of the story and rated how likely the ending was; both were judged to be equally plausible.

Participants again misremembered antecedents in accordance with the ending with which they had been presented – the “rape-ending group” misremembered more antecedents stereotypically associated with rape while the “proposal-ending group” misremembered more antecedents stereotypically associated with a marriage proposal. Hindsight bias was also confirmed again. In this case, however, participants were asked to rate four possible endings: Jack rapes Barbara; they have a one-night stand; they begin dating; Jack proposes to Barbara. In addition to considering the rape outcome more likely, the rape-ending group also rated a one-night stand more likely than did the proposal-ending group, who rated the dating ending as more likely. It seems clear that hindsight bias works just as well for positive as for negative outcomes, but these results also nicely show that this bias isn’t limited to a specific outcome. Rather, participants exposed to the rape ending built an antecedent causal chain that cast Jack & Barbara’s relationship in a less favourable light, making a one-night stand seem a more likely outcome than dating.

As in the first study, the ending had an effect on participants’ perception of behaviour; the rape-ending group had a more negative view of both Jack and Barbara’s behaviour than did the proposal-ending group. However, the same effect wasn’t seen in terms character evaluation. While the rape-ending group rated Jack’s character more negatively, this wasn’t the case for Barbara’s character. Carli suggests that the difference in effect between behaviour and character might be because the behaviours were explicitly described in the story (and so available for judgement) while Jack and Barbara’s character traits had to be inferred from their actions. It may be that these sorts of inferences are more easily drawn for perpetrators than for victims, explaining why Jack’s character was judged more negatively and not Barbara’s. It’s important to realize that these results don’t just hold for the negative outcome – participants who had received the positive outcome had a more favourable view of Jack and Barbara’s behaviour. In both cases, knowledge of the outcome causes observers to misremember antecedents that are stereotypical for that outcome and to blame or credit the characters for behaviour that seems to inevitably lead to the outcome, although the behaviours being judged were identical, regardless of the outcome.

We tend to ground ourselves quite strongly in our perceptions and memories of events. Studies such as this one highlight how shaky that foundation actually is. We judge behaviours as foolish, risky, manipulative, or clever in retrospect through an unconscious dependence on foreknowledge of the outcome. We also remember untrue behaviours and events that seem to lead inevitably to a given outcome but are, in fact, simply cognitive reconstructions, fabrications of our mind which reinforce the seeming inevitability. This can have very real implications, particularly given the common requirement to take “reasonable precautions” to prevent damage and the weight assigned to witness testimony in court proceedings; it may also lead us to unfairly blame victims and to credit others for a success actually based on luck. Together with my earlier piece about cognitive dissonance, this portrays the human mind as a dynamic process of awareness which readily sacrifices accuracy for internal consistency – we don’t really see (or remember) the world around us as much as we reconstruct it and our mind will go to great lengths to maintain the consistency of that construct.

Carli, L. (1999). Cognitive Reconstruction, Hindsight, and Reactions to Victims and Perpetrators Personality and Social Psychology Bulletin, 25 (8), 966-979 DOI: 10.1177/01461672992511005