Ronald Fisher and Edward Geiselman started from a question police interviewers rarely ask: not how do we tell if this witness is lying, but how do we get more out of a witness who is trying to help.
The answer they built is called the cognitive interview, and unlike almost everything else in Part 4 it is derived directly from laboratory memory research rather than from practitioner intuition.
Four instructions form the core.
Mentally reinstate the context. Before recalling anything, the witness is asked to return to the scene in their mind — the weather, the sounds, the light, what they were feeling, what they had been doing beforehand. This comes from Endel Tulving's encoding specificity principle: memory retrieval depends on the overlap between the cues present at retrieval and those present at encoding, and reinstating the context supplies cues that are otherwise unavailable. It is the same reason walking back into a room can recover what you came for.
Report everything. The witness is asked to say every detail regardless of whether it seems important or complete. Witnesses filter aggressively, suppressing anything they judge trivial or uncertain, and that metacognitive filter discards material the investigator may need. It also matters that partial recollections can cue fuller ones.
Recall in varied order. Narrating backwards or from the middle disrupts the schema-driven reconstruction that fills gaps with expectation rather than memory.
Change perspective. Describing what another person present would have seen can surface details from a different retrieval route, though this instruction is the most contested and carries some risk of confabulation.
The enhanced version adds a substantial social component: rapport-building, transfer of control to the witness so that they are doing the talking, open questions, pauses, and no interruptions.
Meta-analyses find the cognitive interview produces a large increase in correct details — commonly around a third to a half more than a standard interview — with a much smaller increase in incorrect details, so the ratio of accurate to inaccurate information is maintained or improved.
Two limits matter. The interview takes considerably longer than a standard one, which is the main obstacle to its use. And the gains evaporate entirely if leading questions are introduced, because the contamination effect of chapter 65 overwhelms everything else.
The technique has one further quality that makes it worth learning outside police work. It is a good-faith instrument: it makes a cooperative witness more useful and does nothing to a hostile one. Anyone conducting a serious conversation about a past event — an incident review, a medical history, a due diligence interview, a difficult conversation with someone about something that happened — will get materially more by using context reinstatement, an uninterrupted free narrative, and open questions than by working from a list.
Cross-ref: Loftus leading questions, Part 3