Judgements of Learning

The gap between confidence and recall

A judgement of learning (JOL) is exactly what it sounds like: an estimate, made during or just after study, of how well you will remember something later. The problem is that people are systematically overconfident. The material feels familiar, and familiarity gets mistaken for durable memory. This is the fluency illusion in its most consequential form — ease of processing registers as knowing, not merely as having recently seen.

Research going back decades, and consolidated by work at Washington University in St. Louis by Henry Roediger and colleagues, shows that the conditions which produce high JOLs are often the conditions that produce poor long-term retention.

Re-reading is the clearest example: exposure to the same text a second time raises confidence measurably while adding comparatively little to what you will actually recall a week later.

The subjective sense of mastery is doing real work — it shapes how long people continue studying, when they move on, and whether they return to material at all. When that sense is inflated, people stop too soon.

Timing matters, too. Give yourself a minute, or better a day, and your prediction improves because the initial fluency signal has faded and only genuine retained knowledge is left to introspect on.

The finding is robust enough to have survived scrutiny across different materials, populations, and labs, though the size of the miscalibration varies with how the question is asked and what is being remembered. It is not the same as claiming people are hopeless at metacognition — they are not — but it does mean that feelings of readiness measured immediately after study are a weak guide to actual readiness.

The practical implication is analytical, not prescriptive: confidence at the end of a session is a lagging indicator, not a leading one.