Interviewing basics: victims, witnesses and complainants

Mira Osei is a nurse who works nights at Port Ardent General. On the night of the Lakeshore Hardware break-in she was off duty, asleep in her apartment above the shop next door, when breaking glass woke her. By the time the probationary officer knocks on her door at 2:21 a.m., she has already called her sister, posted about it on a neighborhood social media group and read two replies from people who say they saw 'a guy in a hoodie' on Pine Street. What she remembers, and what she tells the officer, will be shaped by everything that happens in the next twenty minutes, including how the officer asks.

Interviewing is the core skill of investigation. Most crimes are solved because someone tells police something useful: a victim, a witness, a neighbor, sometimes the person who committed the crime. This lesson covers interviews with victims, witnesses and complainants, people who are generally cooperative and who have information police need. Lesson 8 covers taking formal statements, and later courses cover interviewing suspects, which raises additional legal issues. The aim throughout is to obtain as much accurate information as possible while treating people fairly and avoiding the memory contamination described in lesson 3.

Interviews, not interrogations

For much of the twentieth century, police interviewing in North America was dominated by an accusatory style, most associated with the Reid technique, which was designed to obtain confessions from suspects the interviewer already believed guilty. Research over the past thirty years has shown that accusatory methods increase the risk of false confessions and contaminated accounts and are no better at producing accurate information than non-accusatory methods. They are also entirely unsuitable for victims and witnesses, who are trying to help.

Many police services have moved to an investigative interviewing approach, which treats the interview as a search for information rather than a search for confirmation. The best-known model is PEACE, developed in England and Wales in the early 1990s after a series of wrongful convictions and now used, in various forms, by many services in Canada, Australia, New Zealand and elsewhere. Some US agencies and the largest private interrogation training firm in the United States moved away from confrontational methods in the late 2010s. PEACE is an acronym for the five stages of an interview.

• Planning and preparation: decide what the interview needs to achieve, what is already known, where to hold it, whether an interpreter or support person is needed and how it will be recorded. Even a two-minute interview at the door benefits from thirty seconds of planning.
• Engage and explain: introduce yourself, explain the purpose of the interview and what will happen, and build rapport. Rapport is not friendliness for its own sake; it is the working relationship that lets the person concentrate and speak freely.
• Account, clarification and challenge: obtain the person's account in their own words, then clarify gaps and, where necessary, raise inconsistencies with other evidence, calmly and without accusation.
• Closure: summarize what the person said so they can correct it, explain what happens next, answer their questions and tell them how to contact you if they remember more.
• Evaluation: afterward, review what was learned, how it fits with other evidence, what needs to be checked and how the interviewer could do better.

The cognitive interview

In the 1980s two American psychologists, Ronald Fisher and Edward Geiselman, developed the cognitive interview, a set of techniques based on what psychology had learned about memory. Laboratory and field studies have consistently found that it produces substantially more correct information than standard police interviews, with little or no increase in errors. Its core techniques can be used in a full interview at the station and, in shortened form, at the scene.

The enhanced cognitive interview adds techniques for building rapport, transferring control of the interview to the witness, who knows more about the event than the interviewer, and using open questions and pauses. A key instruction is to tell the witness that it is fine to say 'I don't know' or 'I'm not sure', which reduces guessing. Interviewers are also taught not to interrupt. Studies of real police interviews found that officers interrupted witnesses within a few seconds of their starting to speak, cutting off details before they emerged.

Types of questions

The way a question is asked shapes the answer. Interviewers distinguish several types of question, and good interviews use them in a deliberate order, starting broad and narrowing down.

• Open questions invite a full account in the person's own words. A useful memory aid is TED: 'Tell me...', 'Explain...', 'Describe...'. 'Tell me everything that happened from the moment you heard the glass' is an open question.
• Probing questions follow up on what the person has said, usually beginning with who, what, where, when, why or how: 'What did you see when you looked out of the window?' They are still fairly open.
• Closed questions can be answered in a word: 'Was the streetlight on?' They are useful for confirming specific details late in the interview, but if used too early they limit what the person says.
• Leading questions suggest an answer: 'He was carrying a crowbar, wasn't he?' They should be avoided with witnesses and victims, because they can introduce misinformation and make the account less reliable in court.
• Multiple questions ask several things at once: 'Was he tall, and did he go left or right, and was anyone with him?' The person answers one part and the rest are lost. Ask one question at a time.

A good interview usually begins with a free recall phase, in which the person tells their account with as few interruptions as possible, followed by a questioning phase that probes each part of the account in turn, moving from open to closed questions. The interviewer listens far more than he or she talks. Researchers who analyze recorded interviews often measure the share of the interviewer's questions that are open and the share of words spoken by the witness; in good interviews, the witness does most of the talking.

Analyzing an interview

Linh Tran's part-time job at the communications center has given her an interest in how conversations are structured. For a class exercise, she takes the transcript of the probationary officer's first conversation with Mira Osei, recorded on the body-worn camera, and classifies each question by its opening words. The program below is deliberately simple: it uses the first words of each question as a guide, which is how training officers often do a first pass, and then counts how much each person spoke.

# Classify interviewer questions by type and measure who did the talking
turns = [
    ("officer", "Can you tell me what happened?"),
    ("witness", "I heard glass smash, really loud, and I went to the window and saw a guy "
                "run out of the store and go toward Pine. He had a white hoodie with the hood up."),
    ("officer", "Was he carrying anything?"),
    ("witness", "I think so. Maybe a bag."),
    ("officer", "He had a crowbar, didn't he?"),
    ("witness", "Maybe. I'm not sure. It was dark."),
    ("officer", "Describe the man for me as best you can."),
    ("witness", "Young, maybe twenties. Thin. He ran a bit funny, like his leg hurt. Tall, I think."),
    ("officer", "How long could you see him for?"),
    ("witness", "Only a few seconds, until he got past the bakery."),
    ("officer", "Was he tall or short, and did anyone go with him?"),
    ("witness", "Tall. No one else."),
]
def kind(q):
    q = q.lower()
    if q.endswith(("didn't he?", "wasn't he?", "right?", "wasn't it?")):
        return "leading"
    if " and " in q and q.count("?") == 1 and q.startswith(("was", "did", "is")):
        return "multiple"
    if q.startswith(("tell", "explain", "describe", "can you tell")):
        return "open"
    if q.startswith(("what", "how", "where", "when", "who", "why")):
        return "probing"
    return "closed"
counts = {}
words = {"officer": 0, "witness": 0}
for who, text in turns:
    words[who] += len(text.split())
    if who == "officer":
        k = kind(text)
        counts[k] = counts.get(k, 0) + 1
        print(f"{k:9s} {text}")
asked = sum(counts.values())
print(counts)
print(f"open or probing: {(counts.get('open', 0) + counts.get('probing', 0)) / asked:.0%}")
print(f"witness share of words: {words['witness'] / sum(words.values()):.0%}")

open      Can you tell me what happened?
closed    Was he carrying anything?
leading   He had a crowbar, didn't he?
open      Describe the man for me as best you can.
probing   How long could you see him for?
multiple  Was he tall or short, and did anyone go with him?
{'open': 2, 'closed': 1, 'leading': 1, 'probing': 1, 'multiple': 1}
open or probing: 50%
witness share of words: 64%

Half of the officer's questions were open or probing, and the witness spoke nearly two-thirds of the words, both reasonable for a short doorstep conversation. But the analysis also flags two problems. The question 'He had a crowbar, didn't he?' is leading, and it introduced the idea of a crowbar, which the officer had heard about on the radio. Ms. Osei's answer, 'Maybe. I'm not sure,' shows she did not see one, but if she later 'remembers' a crowbar, the defense will point to this question. The multiple question at the end got an answer to its first part only, and 'tall' contradicts what the man's measured height later showed. The probationary officer should note both problems honestly, and the investigator should conduct a fuller interview using the cognitive interview techniques.

A simple program like this cannot judge the quality of an interview. It misses tone, rapport, pauses and whether the right topics were covered, and some questions do not fit neatly into one type. Training officers use it only as a starting point for a conversation about technique, together with the recording itself.

Interviewing victims

Victims are not simply sources of information. They have been harmed, and the way police treat them affects their recovery, their willingness to report future crimes and their confidence in the justice system. Victims of violent and sexual crimes may be in shock, afraid, ashamed or angry, and their accounts may come out in fragments, out of order or with details that change. These are normal responses to trauma, not signs of dishonesty.

Trauma-informed interviewing is an approach that takes these effects into account. It involves giving the victim as much control as possible over the setting, the pace and breaks; explaining every step; avoiding questions that suggest blame, such as 'Why didn't you scream?' or 'Why were you walking there alone?'; allowing the victim to tell the account in their own way and order; and recognizing that memory for a traumatic event may be clear for some details and absent for others. In the United States, federal grants and state laws have encouraged trauma-informed training for officers who interview victims of sexual assault, and many agencies have adopted it alongside reforms to the testing of sexual assault evidence kits. Victims should also be told about victim services, protective measures and how to stay informed about the case.

Interviewing witnesses at the scene

At a busy scene, the first priority is safety and the second is often to identify and separate witnesses before they talk to each other or leave. The officer should get each witness's name and contact details at once, then a brief account, often called a first account, that captures the key facts needed to act: what happened, who was involved, descriptions and directions of travel. A full interview can follow later, in a quieter setting, ideally recorded. The first account should be noted carefully, including the questions asked, because it is the witness's earliest and least contaminated version.

Witnesses are often reluctant. Some fear retaliation, some distrust police, some are busy or do not want to go to court. Officers can reduce reluctance by explaining why the witness's information matters, what will happen next and what protections exist, and by treating the witness with patience and respect. Pressuring or threatening a witness is wrong and counterproductive: it damages the relationship and the reliability of what the witness says. Some witnesses are themselves vulnerable, such as children, people with intellectual disabilities and people in crisis. Children in particular should be interviewed by specially trained officers or forensic interviewers, often at a child advocacy center, using structured protocols designed to avoid suggestion.

Interviews across language and culture

When a witness or victim is not fluent in English, or is deaf or hard of hearing, a qualified interpreter should be used for any interview that matters. Family members, especially children, and bystanders should not be used to interpret, except in an emergency, because they may change or omit what is said, may have an interest in the outcome and may expose the person to embarrassment. Telephone and video interpretation services make professional interpreters available in many languages at any hour. When using an interpreter, the officer speaks directly to the witness, not to the interpreter, in short segments, and records that an interpreter was used, their name or identification number and the language.

Cultural differences can also affect interviews. Norms about eye contact, personal space, how directly to speak and how to address authority figures vary, and an officer who misreads them may wrongly judge a witness evasive or hostile. Jesse Morningstar explains that, in his community, elders may answer a question with a story rather than a direct reply, and that interrupting to 'get to the point' would be rude and would end the conversation. Listening patiently is almost always the right approach.