MEMORY
COURT.

A memory is an answer constructed now about something that happened then.

You can remember your way home, recognise a friend's voice, and recall the smell of a kitchen from childhood. These achievements are real. The trouble begins when the subjective ease of remembering is mistaken for a guarantee that each recovered detail has been preserved unchanged. Human memory is usually useful, but it is not a camera file with a checksum. Attention was selective during the event; a later question can suggest a detail; the story may be retold; and the present situation changes what is easiest to retrieve.

Calling memory reconstructive is not the same as calling it false. Reconstruction is part of ordinary remembering: fragments of perception, knowledge, expectation and context combine to make an account. Sometimes that account is impressively accurate. Sometimes it borrows a detail from another source or confidently fills a gap. Researchers study the conditions under which these changes occur, rather than announcing that all recollections are equally unreliable. For the bigger picture, the National Research Council's 2014 assessment brings memory science into conversation with eyewitness practice.

Five distinctions that keep us from judging too quickly.

Recollection

A person's report of a past experience. It can contain genuine detail, inference, omission, and error in different proportions.

Recognition

A judgment that something seen now was encountered before. In identification tasks it differs from freely describing an event.

Source monitoring

Judging whether a detail came from perception, another person's account, imagination, or another source. Misattribution can make familiar information feel witnessed. Johnson and colleagues (1993) review this framework.

Misinformation effect

When misleading information after an event affects a later report. Effects vary by procedure, delay, questioning and other circumstances; they are not inevitable. Loftus (2005) reviews decades of findings.

Confidence

Someone's expressed certainty about an answer. Its value depends on when and how it was collected, which question was asked, and possible contamination.

One further distinction is vital: independent evidence is not simply another confident voice. A time-stamped record, physical trace, contemporaneous note or independent observation may constrain what could have happened, but its reliability and chain of custody must also be assessed. An absence in a partial recording is not the same as a contradiction proved by continuous coverage. The Court's model simplifies these evidentiary questions openly.

How a detail can travel from question to memory.

Encoding: during an event, attention is limited. A witness may notice a voice but not a face, a colour but not its exact shade. Stress, distance, lighting, divided attention and exposure duration interact; none gives a universal formula for correctness. Retention: what was encoded can be forgotten, consolidated or altered as time passes and information arrives. Retrieval: answering a question is an action: cues, expectations and wording shape what becomes available to report. Repeated retrieval can also create opportunities for change.

In the classic Loftus and Palmer (1974) car-crash experiments, wording altered participants' speed estimates. In a later question, participants who had heard a more forceful verb were more likely to report broken glass that had not appeared in the film. The result supports caution about leading questions, not a claim that every suggestive word always inserts a detailed false memory.

Gist and association: people often retain meaning rather than the exact list of elements. Roediger and McDermott (1995) demonstrated that participants could recall a strongly associated word absent from studied lists. This is a carefully controlled laboratory demonstration, not a licence to infer what happened in a disputed personal history. Source confusion: when you remember a detail after hearing it from a friend, you may later forget that it arrived through conversation rather than direct sight. Separating the detail from its source is often harder than noticing the detail itself.

Three archives; three different kinds of caution.

“Never trust confidence” is also too simple.

Eyewitness psychology has a necessary tension. One concern is that feedback, suggestive procedures, repeated exposure and memory contamination can make later confidence difficult to interpret. A different emphasis asks whether confidence recorded immediately after an initial identification, under fair, uncontaminated, properly administered lineup conditions, can carry meaningful information about accuracy. These are not logically incompatible claims: they address different testing conditions.

Wixted and Wells (2017) argue that initial high confidence can be strongly informative under appropriately controlled conditions, while warning against transferring that conclusion to compromised procedures. The National Research Council (2014) calls for better practices and evidence collection. Guidance from the US National Institute of Justice (1999) focuses on procedural safeguards. Consequently the useful question is not simply “How confident was the witness?” but “When was confidence elicited, after what exposure, through which instructions, and with what documentation?”

One mistake is to throw out all eyewitness evidence because a famous identification failed. Another is to treat a witness's emotional force as a substitute for preserved procedure. A scientifically responsible interpretation specifies the test conditions, distinguishes initial from later statements, and asks whether external traces support the specific disputed details.

What evidence can establish—and what it cannot.

An independently authenticated record can resolve a narrow question: whether a vehicle entered a particular camera frame, whether a transaction occurred within a certain logged period, whether a DNA sample excludes a person as its source. The same record may leave the wider story uncertain: why someone acted, which conversation occurred outside the frame, or how a witness experienced a frightening event. All evidence has provenance, scope and possible failure modes. Digital timestamps may be wrong; footage may be edited; a witness may accurately report an event not captured by any device.

The Court therefore never labels people truthful or deceptive. Its invented records are deliberately cleaner than real evidence; its tallies count matches under simple explicit rules, not actual likelihoods or confidence–accuracy calibration. There is no empirical validation of the interface. It cannot evaluate a legal case, detect misinformation in a real person, assess competence, diagnose trauma, or establish guilt. Scores across fictional cases would be meaningless, which is why the tool presents small transparent counts rather than rankings.

Ethically, memory science should reduce unjustified certainty without creating cynical blanket distrust. Interviewers can avoid leading questions and record first accounts promptly. Readers can ask where a detail came from while still treating a speaker with care. People who report traumatic events deserve respectful procedures and due process; psychological findings do not justify reflexively doubting, pressuring or pathologising them. The distinction between questioning a claim and attacking a person is the foundational discipline of this court.

Build your own three-column evidence ledger.

Choose a harmless, recent shared experience—not a sensitive dispute. Write a description of four observable moments from memory without consulting messages or photographs. Next to each detail, note whether it was directly perceived, heard from another person, or inferred later. In a second column, record your confidence in ordinary words rather than performing precision with a percentage. In a third, note what independent record might check that detail. Only then look at a calendar, receipt or photograph. Preserve the original notes instead of silently rewriting them.

Ask yourself what the record resolved, what it could not address, and whether one confirmed detail made the rest of your recollection feel more secure than warranted. There is no need to turn a family memory into litigation. The exercise teaches a quieter habit: separating the experience of knowing from the evidence for a particular detail. For classroom use, the printable Field Guide includes three fictional dossiers, worksheets, a bibliography and a fully offline alternative to the interactive lab.

Frequently asked questions