Choice-Supportive Bias: Why Your Choice Improves in Memory

Choice-supportive bias is a fault in source memory, not a change of heart. After a decision, people are usually able to recall the features that were on offer; what they lose is the record of which option each feature belonged to, and the errors they make favor the option they took. That is the angle this page covers. The reasons a person assembles to defend a purchase after the fact are a separate pattern, covered on the page for post-purchase rationalization.

Marcia Johnson's source-monitoring framework, set out with Shahin Hashtroudi and Stephen Lindsay in 1993, explains why the record is fragile. A memory does not arrive labeled with where it came from. Its source is worked out at the moment of recall, from the detail, the context and the plausibility of the candidates, and that reconstruction can be nudged by whatever makes one candidate fit better. Once an option has been chosen it becomes the better fit for a good feature, so a feature that was never attached to it can be attached now.

The same mechanism runs in the other direction for the option that was rejected. Its good points are harder to place, its bad points easier, and the gap between the two memories grows while the actual lists stay where they were written down. The distortion matters because it protects the decision from review: a person comparing the options again is comparing a remembered chosen option that has improved with a remembered rejected one that has got worse.

How far the effect reaches is still disputed. Lind, Visentini, Mantyla and Del Missier's 2017 review sorts the evidence into four kinds of misremembering: attributing a feature to the wrong option, remembering a fact in distorted form, remembering a feature that was never there, and forgetting selectively. They conclude that the support is solid for the first, when memory is tested by recognition, and weaker for the other three and for other kinds of memory test. The direction of the effect is well reported; the claim that memory alone produces it is the part to hold loosely.

Mara Mather, Eldar Shafir and Marcia Johnson published "Misremembrance of options past: Source monitoring and choice" in Psychological Science in 2000, and the name choice-supportive bias comes from that work. Participants chose between two options, each presented as a list mixing positive and negative features, and were later asked which option each feature had belonged to. The errors were not random. Positive features were attributed to the chosen option more often than the lists warranted, and negative features to the rejected one.

Participants were put in one of three scenarios: interviewing people for a job, choosing a blind date, or selecting a roommate, which is where the roommate framing of the demo above comes from. The design matters more than its size. Because each feature really did belong to one of the two lists, a misattribution is an error with a direction, and the direction is what the paper measures: the errors did not cancel out, they leaned toward the option the participant had taken.

The finding sits inside a larger program on source monitoring. Johnson, Hashtroudi and Lindsay's 1993 review had already shown that people confuse what they saw with what they imagined, what they read with what they were told, and one speaker with another, and that these confusions follow the plausibility of the candidate sources rather than chance. Choice-supportive memory is that machinery running with a decision as the thing that makes one source more plausible. Mitchell and Johnson reviewed fifteen further years of the work in 2009.

Mather and Johnson reported in 2000 that older adults showed larger choice-supportive distortions than younger adults, a result they connected to the broader finding that older adults remember emotional material more positively, which Mather and Carstensen set out in 2005. The 2017 review by Lind and colleagues is the strongest caution in the literature: it sorts choice-supportive misremembering into four kinds, and concludes that the support is solid only for the first of them, attributing a feature to the wrong option, and only when memory is tested by recognition.

The pattern shows up wherever a decision is reviewed from memory rather than from the record. The rows below are illustrations, not findings from the studies cited on this page.

Memory for which option had what is the thing at risk, so the counters are about keeping a record that memory cannot edit.

• Write the comparison down before you decide, option by option and feature by feature, and keep the sheet. A later review reads the sheet, not the recollection.
• Record the reasons for the choice on the day, including the good points of the option you turned down. Those are the ones that migrate first.
• When you revisit a decision, get the original materials back in front of you: the listing, the offer, the specification, the quote.
• Ask someone who was there and did not make the choice. A person without the decision has no option to protect, though their memory has its own gaps.
• Separate the question "was it a good decision" from the question "what were the options". The first invites a defense; the second has an answer on paper.

Several effects make a past decision look better than it was. They differ in what is being changed.

• It is not post-purchase rationalization, which is the reasons and ratings a person produces after committing. That one changes the evaluation of the option; this one changes the memory of what the options contained.
• It is not source confusion in general, which is any failure to recall where a piece of information came from. Choice-supportive bias is source confusion with a direction: the errors favor the option that was chosen.
• It is not rosy retrospection, the tendency to recall an experience as better than it was rated at the time. That is about the tone of a remembered episode rather than about which of two options held a feature.