5.2 Memory-Safe Interviewing & Assessing Reliability

Memory-Safe Interviewing & Assessing Reliability

An investigator’s job is to obtain the witness’s memory, not manufacture a polished story. Human memory is reconstructed from fragments and can change through suggestion, repetition, feedback, discussion, and later information.

Contamination risks

Common contamination sources include:

• telling the witness the investigator’s theory;
• showing another witness’s account before free recall;
• asking “the red truck” when color has not been established;
• repeating a question until the answer changes;
• praising one answer and challenging another;
• inviting guesses;
• combining several events in one question; and
• letting witnesses compare accounts before separate interviews.

Confidence can increase after confirming feedback even when accuracy does not. Record the witness’s confidence and language at the time, before giving corroborating information.

Memory-safe sequence

1. Prepare

Define the topic, review known records, and write a short set of neutral objectives. Separate interview witnesses. Decide whether audio recording is lawful and appropriate. Document who is present and any interpreter or support person.

2. Build rapport and ground rules

Explain that accuracy is more important than speed or completeness. Tell the witness to say “I don’t know,” “I don’t remember,” or “I don’t understand.” Ask the witness to correct the interviewer and distinguish personally perceived information from what another person later said.

3. Obtain free recall

Use a broad prompt: “Please describe everything you remember from the beginning, in your own words.” Do not interrupt to fill gaps. Note topics for later follow-up.

4. Use open follow-up prompts

Prompts such as “Tell me more about the vehicle” or “What happened next?” keep ownership of the account with the witness. Narrow questions can clarify time, distance, sequence, lighting, and source of knowledge after free recall. Leading yes/no questions should be reserved for necessary confirmation and labeled in the notes.

5. Context reinstatement

A careful cognitive-interview technique invites the witness to think about the physical setting, sounds, weather, position, and emotional or sensory context as it was experienced. Do not tell the witness what should have been present. Make clear that the witness must not imagine or guess.

“Report everything” can recover details the witness initially considered trivial. It does not mean report speculation as fact.

Reverse-order or changed-retrieval prompts may sometimes cue different details, but they also increase cognitive demand and can confuse a tired, traumatized, young, or cognitively impaired witness. Use them cautiously, record the prompt, and do not treat difficulty as evidence of lying.

The former draft’s “change perspective” instruction is risky if it asks a witness to imagine another person’s thoughts or viewpoint. A witness may describe physical sightlines from remembered positions; the witness should not invent what another person saw.

Reliability is broader than demeanor

Evaluate:

• opportunity to perceive;
• distance, duration, lighting, noise, obstruction, and attention;
• impairment, stress, injury, or fatigue;
• delay between event and interview;
• exposure to other accounts or media;
• consistency on central versus peripheral details;
• spontaneous correction and acknowledged uncertainty;
• motive or bias; and
• independent corroboration.

An inconsistency can result from deception, but also from memory decay, misunderstanding, language, trauma, different question framing, or an honest correction.

No behavioral lie detector

Research does not support a universal cue that proves lying. Avoid categorical claims about gaze, crossed arms, fidgeting, self-touch, facial expressions, voice pitch, pronouns, passive voice, tense, or phrases such as “to be honest.”

A “baseline” comparison does not solve the problem. The stakes and topic change between neutral and critical questions, so behavior can change for innocent reasons. Cultural norms, autism, anxiety, medication, pain, and prior trauma can alter presentation.

Microexpressions do not reveal a specific cause or establish deception. A fleeting expression may show emotion, if it is interpreted correctly, but emotion is not guilt and guilt is not the same as falsity.

The original draft’s fixed SCAN narrative percentages—such as truthful accounts containing a particular share of incident detail—are not a defensible truth test. Narrative length depends on memory, writing ability, culture, trauma, time, instructions, and event complexity.

Evidence-based testing

Instead of declaring “the witness lied,” test the account:

1. obtain a complete account before revealing evidence;
2. pin down sources of knowledge and uncertain estimates;
3. ask for checkable details without suggesting them;
4. compare versions using exact quotations and question wording;
5. seek records, video, timestamps, receipts, location data lawfully obtained, or independent witnesses;
6. present a contradiction neutrally and invite an explanation; and
7. preserve both corroborating and disconfirming information.

A strategic decision to withhold a fact until after free recall can help test independent knowledge. It must not become a trick that misstates evidence or coerces an admission.

Reporting language

Prefer:

• “The witness’s account changed from 8:00 to approximately 9:30 after reviewing the receipt.”
• “The witness looked away and paused; no conclusion was drawn from demeanor.”
• “The account was corroborated by the timestamped entry log.”
• “The discrepancy remains unresolved.”

Avoid:

• “The witness displayed deception.”
• “The microexpression proved contempt.”
• “The use of passive voice established guilt.”
• “The statement was 72% prologue and therefore false.”

Exam method

Select the option that protects free recall, documents prompts, allows uncertainty, and tests the account against evidence. Reject any choice that equates one behavior, phrase, emotion, or narrative ratio with deception.