4 Psychology and legal proceedings

Core idea

Courts treat human memory and human judgement as if they were instruments — eyewitnesses recall events, suspects confess to crimes, jurors weigh evidence. Decades of empirical psychology have shown that each of these instruments is systematically unreliable in ways the law has been slow to absorb. Memory is not a recording but a reconstruction; confessions can be produced by procedure rather than guilt; jury verdicts track narrative coherence more closely than they track probability. Forensic psychology's contribution to legal proceedings is to document those failure modes precisely enough that the system can be redesigned around them — better lineup procedures, audited interrogation methods, and clearer instructions to juries.

Author's framing: The legal process assumes the witness, the suspect, and the juror are passive observers of fact. Psychology shows all three are active constructors of meaning, and the procedures of the court determine what they construct.

Why it matters

When a witness identifies a stranger, when a suspect admits guilt in an interview room, when a jury delivers a verdict, the legal system treats those outputs as evidence. If the process that produced them was flawed — a suggestive lineup, a coercive interrogation, an emotional closing argument — the output is contaminated regardless of how confidently it is delivered. Most wrongful convictions later overturned by DNA evidence trace back to one of three sources: mistaken eyewitness identification, false confession, or jury misinterpretation of forensic evidence. Each is a problem psychology can describe and procedure can mitigate. Ignoring those findings keeps convicting the wrong people.

Key takeaways

Mental model

Copy sourceZoom

How memory actually works in court

Eyewitness memory is reconstructive

The intuitive model of memory treats it like a video file: the event is recorded once, sits in storage, and is played back on demand. Elizabeth Loftus's research starting in the 1970s dismantled this picture. In her classic studies, participants watched a film of a traffic accident and were then asked questions about it. A single word change — "About how fast were the cars going when they smashed into each other?" versus "hit each other?" — produced significantly higher speed estimates and made participants more likely to later report having seen broken glass that was never in the film. The leading word did not just bias the answer; it was incorporated into the memory itself.

This is the misinformation effect: post-event information becomes indistinguishable from the original experience because memory is rebuilt at each retrieval rather than replayed. The witness who tells the court, on oath, that they saw broken glass is not lying — they are remembering. They are simply remembering something that was assembled, in part, from a leading question asked weeks earlier.

Confidence does not equal accuracy

The court relies on demeanour cues to weigh testimony: hesitation reads as doubt, fluency reads as truth. Psychology has shown this calibration is broken. A witness who has rehearsed their account, received positive feedback after a lineup identification ("Good — you picked the suspect"), or simply lived with their memory long enough for it to feel settled, will testify with high confidence regardless of whether the memory is accurate. Worse, confidence is malleable: post-identification feedback inflates it, raising the risk that an originally tentative pick is presented to the jury as certainty.

The lineup is itself an experiment

A police lineup is not a neutral test — it is a procedure that can be designed well or badly, and the design changes the result. Simultaneous lineups (showing all candidates at once) encourage relative judgement: the witness picks whoever looks most like the perpetrator, even if the actual perpetrator is absent. Sequential lineups (one face at a time, with an absolute yes-or-no decision on each) reduce that pressure. Double-blind administration (the officer running the lineup does not know which person is the suspect) prevents unconscious cueing — a glance, a pause, a tone of voice that signals approval.

Practical application

Interrogation: PEACE versus Reid

The two dominant interrogation models reflect opposite assumptions about the suspect.

The Reid technique, developed in the United States in the 1950s and still widely taught, treats the interview as a tool for extracting confessions from suspects assumed to be guilty. It involves a long monologue presenting the case as overwhelming, alternating between sympathetic and confrontational framing ("maximisation and minimisation"), and offering face-saving rationalisations ("anyone in your position would have done the same"). It produces confessions efficiently. It also produces false confessions — especially from young, sleep-deprived, or cognitively vulnerable suspects who eventually accept the rationalisation as a way to end the interview.

The PEACE model (Preparation and Planning, Engage and explain, Account, Closure, Evaluate), developed in the UK in the 1990s, treats the interview as information-gathering. The interviewer asks the suspect to give a free narrative, then probes inconsistencies neutrally rather than confronting them. PEACE produces fewer outright confessions but more reliable testimony, fewer retractions, and substantially fewer wrongful convictions traceable to the interview room.

Copy sourceZoom

Juries decide by story, not by score

Pennington and Hastie's "story model" of juror decision-making found that jurors do not maintain a running probability tally as evidence arrives. They build a narrative — a causally connected account of what happened — and choose the verdict that fits the most coherent story. Whichever side (prosecution or defence) tells the cleaner story tends to win, even when the underlying evidence is ambiguous. This is why the order of evidence matters, why a single vivid detail can dominate a mountain of statistics, and why expert testimony about base rates often fails to land: it asks the jury to do statistics when their cognitive machinery is wired for plot.

Special caution with children and the suggestible

Example

Consider a robbery at a corner shop. A man in a dark jacket pushes past the cashier, grabs cigarettes from behind the counter, and runs. The cashier — call her Aisha — is the only witness, and the whole episode lasts eleven seconds.

A first officer arrives within minutes and asks: "Was he white or Asian?" Aisha hesitates, then says "white". She had not consciously registered the offender's face — she had focused on his hands at the till. But the question forced a binary choice and her brain delivered an answer, which she now remembers giving.

Two days later, a detective shows Aisha six photographs simultaneously. He says, in a friendly tone, "Take your time, see if you recognise the man." A suspect — a regular customer with a prior — is among the six. Aisha picks him. The detective says, "Good, that's helpful." Her confidence rises overnight from a three to an eight out of ten.

At trial four months later, Aisha testifies that the defendant is "definitely" the man she saw. The jury believes her — the demeanour is composed, the story is clear, and the defence cannot point to anyone with a stronger alibi. He is convicted.

Each step of this sequence is psychologically predictable and procedurally preventable. The leading first-officer question planted a racial frame. The simultaneous lineup invited relative judgement. The detective's "good" confirmed and inflated the choice. The four-month gap let confidence harden. The trial saw none of this — only the polished testimony of a sincere witness. Substitute a body-cam recording of the first questioning, a double-blind sequential lineup with cautious instructions, and a jury instruction on confidence-accuracy calibration, and the same case looks very different — without anyone lying, suppressing evidence, or doing anything other than following the version of "good practice" that empirical psychology has shown is anything but.

Related lessons