By Jason Hreha

Glossary

Recall Bias

Published 2 min read

Two customers cancel the same service. One has spent an hour documenting every frustrating experience; the other barely remembers signing up. If you ask why they left, you are collecting their accounts of the past—not replaying the past itself.

What is recall bias?

Recall bias is systematic error in reports of past events caused by how those events are remembered or reconstructed. Researchers are especially concerned when the process differs between groups being compared. A group that searches its memory more thoroughly can appear to have had more of an experience even when the original experience was equally common.

In his review of retrospective epidemiological research, Steven Coughlin distinguishes biased recall from simply inaccurate recall. Forgetting an appointment is an error. A pattern of remembering appointments differently depending on a later outcome can distort an estimate of the relationship between attendance and that outcome.

How a comparison can go wrong

Imagine a hypothetical study asking 100 satisfied customers and 100 dissatisfied customers whether a delivery arrived late. Suppose records show that 20 people in each group experienced a delay. All 20 dissatisfied customers remember it, but only 10 satisfied customers do. The survey produces rates of 20% and 10%, although the recorded rates were both 20%.

That ten-percentage-point gap came from reporting, not from different delivery histories. It does not show that dissatisfaction always improves memory, or that a real association between delays and dissatisfaction cannot exist. The example isolates one way a retrospective comparison can mislead.

Equal forgetting is a different situation, but it is not automatically harmless. The consequences depend on what is misremembered and how the measure enters the analysis. Avoid the shortcut of assuming that every memory error cancels out—or that every error exaggerates the result.

Recall bias, availability, and social desirability

When people report their own past behavior, recall bias is one possible source of self-report bias.

The availability heuristic concerns judging frequency or likelihood from how easily examples come to mind. Recall bias concerns the accuracy of a report about the past. Ease of retrieval can matter to both, but the questions differ.

A person might also remember an event accurately and choose not to disclose it. That points toward a reporting problem such as social desirability bias, rather than a failure to remember. An inaccurate answer alone does not identify which process produced it.

Design the record before you need the memory

When feasible, record the relevant behavior as it happens. For the delivery example, dated tracking records answer whether a parcel was late; an interview answers what the customer experienced and considered important. Those sources serve different purposes.

For retrospective questions, specify the event and time period, pilot the wording, and check a suitable subsample against another source. Althubaiti’s review of information bias discusses recall periods, diaries, memory aids, and validation. It also notes that comparison measurements can contain errors. A diary started after the event cannot recover a missing contemporaneous record.

Ask what evidence would reveal the reporting error before choosing a correction. The research-methods guide helps separate questions about recorded behavior from questions about people’s interpretation of it.

References