Interview and statement templates
Workplace Investigation Interview Questions by Allegation Type
Last updated 2026-07-30
Question lists are cheap. Every HR vendor publishes one, the counts escalate — thirty-five, forty-eight, fifty-five — and none of them tells you the only thing that matters, which is what to ask second. A question asked out of sequence does not produce a worse answer; it produces a contaminated one, and no later question repairs it.
This page is not another party-by-party list. The complainant, respondent and witness sets already have their own pages in this library and are summarized here in a paragraph each. What has no home elsewhere is the layer on top: the four-stage funnel every interview runs through, and the element-shaped overlays that change with the allegation. A harassment finding has to reach unwelcomeness, the protected-characteristic link, severity or pervasiveness, and the harasser's authority. A retaliation finding turns on knowledge and sequence. A policy-breach finding turns on the version of the policy the employee actually acknowledged. Those are different interviews wearing the same room layout.
Run the party set first and the overlay after it, in that order and never the reverse. Overlay questions are element-shaped, and element-shaped questions asked early teach the interviewee what you are collecting. Asked after the account is on the record they test it; asked before, they write it.
An investigator reading forty numbered questions top to bottom regardless of the answers is producing a transcript. The sequencing logic in each section is what you hold onto when the interview goes somewhere the list did not anticipate.
Jurisdiction
Written primarily against US federal standards — EEOC guidance, NLRB decisions and Title VII case law — and against general investigative practice. State law, sector rules (public employers, education, regulated industries) and collective agreements impose additional requirements this page does not cover. Where a page borrows a concept from another jurisdiction it says so explicitly. Outside the US, treat the structure as transferable and the legal references as not.
Workplace Investigation Interview Questions by Allegation Type template
Free Word download, no email required. Or open it in Professional Drafter — it lands in your template library, ready to draft against.
When to use this document
- Preparing an interview plan for a specific matter and needing a checked set rather than a remembered one.
- Running an allegation type you handle rarely — retaliation, disparate treatment, a code-of-conduct breach — where the elements are easy to under-cover.
- Training or supervising a first-time internal investigator who needs the reasoning, not just the questions.
- Auditing a completed investigation for what was never asked, before opposing counsel or an agency investigator identifies it for you.
- Building a house question set that multiple investigators will apply consistently across matters.
Worked example
Fictional scenario, continuing the dispatch-team matter (RW-2026-038). Complaint dated 29 May 2026, received 30 May; investigator R. Hall, HR Business Partner, internal; complainant Nadia Okafor, Dispatch Coordinator; respondent Sean Whelan, Team Lead; report 14 July 2026. The 9 June itemization and the 12 June witness interview are worked through on the complainant-interview and witness-interview pages. This example takes the layer those pages do not cover: how the harassment overlay was actually run on Allegation 2 — the supervisor-status questions and where their answer came from, one leading question caught before it was asked, the 19 June exchange that made a follow-up necessary, and the narrow complainant follow-up of 1 July.
The overlay question the party set could not answer: is Whelan a supervisor?
Employer exposure on a substantiated harassment finding depends on whether the harasser is a supervisor in the Vance sense. That is not a fact a narrative about comments produces, and it is not a fact the complainant is in a position to establish.
Asked of Ms Okafor, 9 June:
INV: Does Mr Whelan decide anything about your work?
NO: He does the rota. And he did my mid-year performance review.
That answer is useful and it is not an answer to the question. Under Vance v. Ball State Univ., 570 U.S. 421 (2013) a supervisor is someone the employer has empowered to effect a tangible employment action — hiring, firing, failing to promote, reassignment with significantly different responsibilities, a decision causing a significant change in benefits. Building a rota and writing a review are not those things; a review may feed a decision that somebody else makes. The pre-Vance formulation — authority to direct daily work — would have made him a supervisor on this answer alone, and it is the formulation most published question banks still carry.
So the question went to the people who could answer it.
Asked of Mr Whelan, 19 June:
INV: If you wanted a coordinator moved off dispatch permanently, what would you have to do?
SW: Put it to the Ops Manager. It is not mine to do.
And to the record, 22 June: Delegation of Authority schedule v4 — Team Leads hold no hiring, termination, promotion or grade authority; recommendations only.
Recorded conclusion: on the current record Mr Whelan is not a supervisor in the Vance sense, notwithstanding the title. That reopens rather than closes the interview plan. Co-worker harassment is assessed on employer negligence, so what management knew and when now carries the analysis — which sent R. Hall back to the prior-reports questions with two further managers to interview.
The leading question that was caught, and the three that replaced it
At stage three on 9 June, working on the continuation point, the question the investigator was about to ask was:
✗ 'And he carried on making comments after you asked him to stop, didn't he?'
That supplies the fact, embeds the conclusion, and invites a one-word agreement. Had it been asked, the notes would read 'Q: he carried on after you asked him to stop, didn't he? A: yes' — and a reviewer would be entitled to say the continuation came from the investigator. Allegation 2 is one of the two allegations ultimately substantiated; resting it on that exchange would have been the weakest available foundation.
What was asked instead:
✓ 'What happened after that conversation?'
✓ 'Over the next few weeks, was there anything further?'
✓ 'Tell me about the next occasion you remember.'
NO: There were a couple more. One was about two weeks later, something about the dress I had on. And another one, I could not tell you the date.
The answer is hers. It carries a count, an approximate interval and a subject, none of which the investigator supplied. Note what the rewrite also declines to do: it does not ask whether the comments were harassment, and it does not ask how they made her feel. Both come later and both are worth less.
19 June — the flat denial, and the two questions that unbundled it
INV: The second allegation is that in early April Ms Okafor asked you to stop making comments about her appearance, and that comments continued afterwards. What is your response?
SW: She never said anything to me.
INV: Do you say that conversation did not happen?
SW: I do not remember it.
INV: Do you deny it happened?
SW: I cannot say it did not. I do not remember it.
Two separate questions, deliberately. 'Are you denying it or saying you do not recall?' is a compound question and the answer to it would have been unattributable — the same defect this page's contaminating-forms section is about, and it is easiest to commit when you are trying to be fair.
What the exchange produced is two distinct records, and both were kept. First, the substantive position: he does not recall the request and does not deny it was made. That is not a conflict of accounts at all, which is what let the finding rest on the contemporaneous evidence rather than on a credibility contest. Second, the movement: his opening answer was a flat denial and he retreated from it inside thirty seconds. That retreat was recorded verbatim in the notes, because an answer that changes under a neutral follow-up is a credibility datum whether or not you end up relying on it.
R. Hall did not rely on it, and said so in the report. A general 'she never said anything to me' offered before the specifics are fully in mind is at least as consistent with imprecision as with untruthfulness, and treating a first-pass denial as a lie would have been the same error as treating the complainant's certainty as proof. Recording the movement and declining to weigh it is a different act from not noticing it — and only the first survives a reviewer reading the notes.
The counterfactual: what running the overlay first would have cost
Suppose the 9 June interview had opened with the harassment overlay instead of closing with it — unwelcomeness, frequency, first and last occasion, authority.
Ms Okafor would have understood inside ninety seconds that counting occasions and identifying who held power was what the investigator wanted. Every answer after that would have been shaped by it: a number offered with more confidence than the recollection supported, an emphasis on the rota and the performance review because those were plainly the answers being fished for. The account would have looked stronger and been worth less, and nothing in the notes would have shown why.
Run in the correct order, the free narrative produced 'about February', 'it kept happening', and the April request — none of it counted, none of it framed. The counts and the authority questions arrived afterwards and tested material that already existed. That is the entire difference between a set of questions and a sequence of them, and it is invisible on any published list of thirty-five essential questions.
1 July — the follow-up, and what it deliberately did not do
Mr Whelan's non-recollection had to be put to Ms Okafor before it appeared in a finding. It was, on 1 July, in an interview that ran eleven minutes.
INV: Since we spoke I have taken an account that differs from yours on one point, and I want to put it to you. Another account of the early April conversation is that it did not take place, and that the person concerned has no recollection of it. What is your response to that?
NO: It happened. I know what I said.
INV: Has anything come back to you since we spoke — anything further about that day, or anything in writing?
NO: No. Just the text to Priya.
That is the whole re-interview. What it did not do is re-run the February-to-May chronology, re-take the narrative, or revisit Allegation 1. Nothing in her account had been challenged on those and reopening them would have generated a second telling differing from the first in immaterial ways — discrepancies manufactured by the investigator, which then have to be explained in the report.
What it also did not do is name a witness. 'Another account of this is [X]' keeps the source out of the room; 'Mr Whelan says' would have been permissible here since he is a party, but the same neutral form is used for every re-interview in the matter so that the one time it protects a witness does not stand out as different.
Three questions across two interviews decided Allegation 2: a stage-three question on 9 June that did not tell her the answer, an unbundled clarification on 19 June, and a neutral put-back on 1 July. The finding of 14 July rests on those and on a text message dated 8 April.
Section-by-section guide
What each section is for, what a good one contains, and where they typically go wrong.
1.Matter details, allegation types and interview sequence
Record the sequence before you record the questions, because the sequence is the part that gets challenged. The default order is complainant, then the documents the complainant identifies, then witnesses, then the respondent, then any re-interviews. Each stage exists to make the next one answerable: you cannot itemize allegations before hearing the complainant, you cannot put a complete case to a respondent before the witnesses have been taken independently, and you cannot re-interview usefully until you know which conflicts are material. Departures happen — a witness resigns, a respondent is out on PTO, an urgent safety issue forces an early conversation. Write down the departure and the reason. An investigation that interviewed the respondent second will be asked why, and 'the complainant was out until 22 June' is an answer; silence is a finding against you. Record the standard too, because it governs how long you keep asking. Every question here is asked to a preponderance-of-the-evidence standard — more likely than not. A question asked to remove all doubt is a question asked past the standard, and the twenty minutes it consumes come out of an allegation you never reached.
2.The four-stage funnel — applies to every interview in the matter
Every set below runs through the same four stages, and the stages exist because information degrades when they are taken out of order. Stage one is free narrative: one open invitation, then silence. Interrupting in the first two minutes is the most expensive habit in investigative interviewing, because it converts an account into a response to your assumptions. Stage two is expansion: take each thing they raised and open it out using their words, not yours — 'you said he was being a bit much; tell me what that looked like.' Stage three is specifics: date, time, location, exact words, who was within earshot, what happened in the ten minutes afterwards. Stage four is testing and closure: put the inconsistencies and the contrary evidence, then ask what you have not asked. The funnel runs one way. Once you have asked a specific question about a topic, the free narrative on that topic is gone and cannot be recovered — which is why a specific question asked in minute two costs you the whole account, not just that answer.
3.Universal opening block — run identically for everyone
The same opening block for every interviewee in the matter, so that variance in the answers carries meaning. Consistency here is what lets you say later that the complainant and the respondent were treated the same way. Three of these earn their place for reasons that are not obvious. The collusion screen — whether anyone has spoken to them about this interview beforehand — belongs at the front, before an interviewee has any reason to conceal the answer, and it is asked neutrally of everyone rather than pointedly of one person. The records question is the highest-yield question in the bank: people keep diaries, message threads and calendar notes they will never think to mention unless asked directly. And the representation question belongs in the block rather than in the respondent set alone, because you cannot always predict who ends up facing discipline. In a unionized workplace, NLRB v. J. Weingarten, Inc., 420 U.S. 251 (1975) gives an employee the right to request union representation at an investigatory interview they reasonably believe may lead to discipline — the employee must invoke it, and you are not obliged to volunteer it, but asking and recording the answer costs nothing and closes a line of attack. Public-sector questioning that could lead to criminal exposure raises separate Garrity issues; take advice rather than improvising.
4.The three party sets — what each one is for
Three interviews doing three different jobs, summarized here because each has its own page with the full question set and a worked example. The complainant interview converts a grievance into numbered, provable allegations. Its internal order is narrative, itemization, contemporaneous records, prior reports, impact, resolution sought — and the records questions sit after itemization because they are keyed to specific dates. 'Do you have any evidence?' reliably produces 'no'. 'What did you do the evening of 8 April, and did you tell anyone?' produces a text message. See complainant-interview. The respondent interview discharges a fairness obligation: each allegation goes to them separately, in full, with dates, before any finding is made. It opens with background rather than allegations, because context is easier to obtain candidly before the room turns adversarial. Where a respondent accepts an act but disputes its character — 'I said it, but it was a compliment' — record the acceptance and the characterization under separate headings; the first is a fact established on his own account, the second is a position going to a different question. See respondent-interview. The witness interview harvests recollection that has to be independent to be worth anything. Narrow by time and place rather than by content, and record the point at which you had to disclose a specific in order to reach a question — a reviewer otherwise cannot tell which half of the interview was uncontaminated. See witness-interview.
5.Overlay — harassment allegations
Under Title VII the questions that decide a harassment matter are unwelcomeness, whether the conduct was based on a protected characteristic, and whether it was severe or pervasive enough to alter the conditions of employment — Meritor Savings Bank v. Vinson, 477 U.S. 57 (1986). Harris v. Forklift Systems, Inc., 510 U.S. 17 (1993) adds a double test most question sets miss: the environment must be one a reasonable person would find hostile or abusive, and one the complainant actually perceived that way. Both halves need questions. Ask about frequency, duration and audience separately — 'how often', 'over what period' and 'who was around' are three answers, and pervasiveness is built from all three. Employer liability is a separate line of questioning and it turns on the harasser's status. Under Faragher v. City of Boca Raton, 524 U.S. 775 (1998) and Burlington Industries, Inc. v. Ellerth, 524 U.S. 742 (1998) an employer is vicariously liable for a hostile environment created by a supervisor, subject to an affirmative defense — reasonable care to prevent and correct, plus an unreasonable failure by the employee to use the complaint channels — which is unavailable where a tangible employment action followed. That defense is built or lost on questions about the policy and the reporting route, which is why they belong in the interview rather than in the report. Who counts as a supervisor is where question sets go wrong. Vance v. Ball State Univ., 570 U.S. 421 (2013) held that a supervisor is someone the employer empowered to effect a tangible employment action — hiring, firing, failure to promote, reassignment with significantly different responsibilities, or a decision causing a significant change in benefits. Directing someone's daily work is expressly not enough. Ask the authority questions in those terms, and verify the answer against the delegation-of-authority record rather than the complainant's impression. Where the harasser is a co-worker, liability turns on negligence — what management knew or should have known, and what it did. A note on the EEOC materials: the agency's 1999 vicarious liability guidance is still online but is flagged there as superseded by the 2024 Enforcement Guidance on Harassment in the Workplace, which was itself rescinded on 22 January 2026. Its value now is the Vance notice it carries on its own first screen. The framework you are questioning to is the caselaw, which is unaffected. Finally, do not ask 'was it unwelcome?' and treat the yes as your finding. Unwelcomeness is established by what the complainant did and said at the time, which is why the behavioral questions below outrank the conclusory one.
6.Overlay — discrimination allegations
Most of what follows is shaped by McDonnell Douglas Corp. v. Green, 411 U.S. 792 (1973) — the burden-shifting structure US courts use for circumstantial Title VII claims: prima facie case, a legitimate non-discriminatory reason from the employer, then an opportunity to show that reason is pretext. It is a litigation evidentiary framework, not a standard you apply, and comparator evidence is one route to an inference rather than an element you must satisfy. Question to it anyway, because testing the stated reason for documentation, consistency and stability is exactly the analysis a later fact-finder will perform. Establish who actually made the decision — not who signed the letter — and what they knew about the complainant's protected characteristic at the time. Then establish the reason given, and when it was first given. A reason articulated at the time of the decision and one articulated after the complaint was filed carry different weight, and a reason that has changed since is more probative than most of what else you will collect. Comparator questions have to be concrete rather than categorical: same supervisor, same rule, comparable conduct, comparable record. A comparator who reported to a different manager under a different policy version is not a comparator, and an analysis built on one is a defect respondent's counsel finds in a single reading.
7.Overlay — retaliation allegations
Retaliation turns almost entirely on knowledge and sequence, so the questions are chronological rather than thematic. Three things have to be established: protected activity, knowledge of it by the person who took the action, and an adverse action that followed. The knowledge question is the one investigations skip and the one that decides them — a manager who genuinely did not know about the complaint cannot have retaliated for it, and 'who told you, when, and what were you told' is worth more than any amount of inference from timing. On the adverse-action side the controlling standard is Burlington Northern & Santa Fe Railway Co. v. White, 548 U.S. 53 (2006): the action must be 'materially adverse,' meaning it 'well might have dissuaded a reasonable worker from making or supporting a charge of discrimination.' That is deliberately broader than the standard for a substantive discrimination claim and is not limited to hiring, firing or pay — the Court framed it as separating significant harms from trivial ones, not as requiring economic detriment. The EEOC's 2016 retaliation guidance adopts that standard; it does not create it. Cast the questions wide accordingly: exclusion from meetings, loss of desirable assignments, schedule changes, a marked change in supervision. Always ask about pre-existing plans and intervening events, and always ask for the documentary trace of when a decision was actually made. A performance plan drafted before the complaint and delivered after it looks like retaliation and is not — and the only thing that separates the two is a date on a document.
8.Overlay — bullying and hostile conduct with no protected characteristic
There is no federal statute making generic workplace bullying unlawful, so a bullying finding rests entirely on the organization's own policy — which makes the policy wording the first question rather than an afterthought. (Several other jurisdictions do legislate directly on this: Australian and some Canadian provincial frameworks impose statutory duties around workplace bullying and psychological safety. If you are working from those materials in a US matter, the concepts transfer but the legal hook does not.) The characteristic defense is that the conduct was legitimate management action — performance feedback delivered firmly. That defense is tested with pattern, proportion and selectivity questions, not with questions about how the conduct felt. Ask what the same manager does with other reports, ask whether the conduct occurred in front of an audience, and ask what the stated performance concern was and whether it was ever documented. Firm feedback documented in a one-to-one and delivered to everyone is management. The same words delivered only to one person, in front of the team, with no record, is not.
9.Overlay — policy and code-of-conduct breaches
Policy-breach investigations fail on notice far more often than they fail on facts. Establish the version of the policy in force on the date of the conduct, whether the employee acknowledged it, and what the provision actually says — not what everyone assumes it says. An investigation that measures conduct against a policy the employee was never given, or a version that post-dates the conduct, has a problem no amount of careful fact-finding repairs. The second failure point is inconsistent enforcement: a rule ignored for two years is difficult to enforce against one person without an explanation for why now. Ask what the practice actually was, not what the document says it should have been. Separate the rule breach from the intent behind it. Most conduct policies are breached by the act regardless of motive, and letting the interview drift into intent early loses the factual acceptance that the act occurred — which is usually the only thing the interview was going to produce that you could not get elsewhere.
10.Question forms that contaminate the record
A badly formed question damages evidence permanently, and the damage is invisible in the notes unless the question itself was written down. Record questions verbatim alongside answers, always. Five forms cause almost all of the problem. Leading questions supply the fact and invite agreement. Compound questions bundle two propositions so that a single 'yes' cannot be attributed to either — and note that this catches well-intentioned clarifying questions too: 'are you saying it did not happen, or that you do not recall it?' is compound, and the answer to it is unusable. Ask the two halves separately. Assumptive questions embed a contested fact as a premise, and the interviewee adopts it without noticing. Conclusory questions ask for the finding rather than the evidence for it — 'did you feel harassed?' asks the witness to do your job, and the answer proves nothing either way. Judgmental questions cost you the interviewee for the rest of the session and buy nothing the neutral form would not have produced.
11.Re-interview and follow-up questions
Re-interviews are not a sign the first interview failed; refusing to run one when the evidence has moved is a sign the investigation did. The trigger is a material conflict or a new allegation, and the rule is that the person affected answers it before it appears in a finding. Keep it narrow. Put the specific new material, take the response, stop. A re-interview that revisits the whole account invites a fresh narrative that will differ from the first in immaterial ways, and you will then spend the report explaining discrepancies you manufactured. When you put contrary evidence, put the proposition and not the person. 'Another account of this is [X]; what is your response?' is fair and keeps the source out of the room. 'Priya says she heard it' identifies a witness, converts the response into a reaction to a colleague, and hands the respondent a name he did not have. Record how you put it, in the words you used.
12.Closing block and note verification
The same closing questions in every interview, and the verification record after them. The 'anything I have not asked' question is not a courtesy — it routinely produces the most useful material in the interview, because interviewees withhold things they assume are irrelevant until they are invited to judge relevance themselves. Record the review step even when nothing changed. 'Notes sent 21 June, read back, no corrections' is a record. A blank field is not, and it is indistinguishable at review from a verification that never happened.
Common mistakes
The failure modes that show up most often when this document is reviewed later.
Treating 'he assigns my work' as establishing supervisor status
That is the pre-Vance test, and it still sits in most published question banks. After Vance v. Ball State Univ. a supervisor is someone empowered to effect a tangible employment action, so directing daily work settles nothing. Get it wrong in the interview and the report either overstates the employer's exposure or applies the wrong liability framework entirely — and the fix requires a delegation-of-authority document nobody thought to request.
Using a note template that has a column for answers and none for questions
The format causes the defect. Nobody decides to stop recording their own questions; they use a two-column form headed 'Topic / Response' and three interviews later there is no way to tell whether a fact came from the witness or from the investigator. A leading question is undetectable in notes that never held it, which is why opposing counsel asks for the questions before the answers.
Asking the comparator question categorically — 'was anyone else treated differently?'
It reliably produces a name and nothing else. You then discover at report stage that the comparator reported to a different manager, under a different policy version, with a different disciplinary record — and the analysis you built on them collapses. Establish the shared manager, rule, conduct and record first; ask what happened to them second.
Attributing contrary evidence to a named person when you put it
'Priya says she heard it' hands the respondent a name he did not have, and in a team of three coordinators he did not need much. It also converts his response into a reaction to a colleague rather than to a proposition, which is a worse answer. Put the account, never the mouth it came from.
Reading dates off the complaint document to everyone else
'On 17 April, did you see anything unusual?' imports the complainant's date into every subsequent account, and the timeline that looks independently corroborated at report stage is the same timeline three times over. Establish the witness's own anchor — a shift, a delivery, a meeting — and let the date come back to you.
Asking for the gist before asking for the words
Once someone has summarized an exchange in their own gloss, they reproduce the gloss on every retelling and the original words are gone. 'What was the general tone?' asked before 'what exactly did he say?' costs you the only version of the sentence that a policy provision can be applied to.
Never asking anyone when they first learned a complaint had been made
In a matter with no retaliation allegation the question feels irrelevant, so it gets skipped. Then a retaliation allegation arrives in August, the knowledge chain is the whole case, and everyone's answer about June has been reconstructed with the stakes known. The date costs one question to capture and cannot be recovered afterwards.
Questions
- In what order should I interview the complainant, witnesses and respondent?
- Complainant first, then the documents and messages the complainant identifies, then witnesses, then the respondent, then any re-interviews. The respondent goes last so the complete case can be put once rather than in installments — returning with fresh allegations looks like a shifting case and gives an unnecessary opportunity to tailor an account. Where you have to depart from the order, record the departure and the reason.
- How many questions should a workplace investigation interview have?
- Wrong unit of measurement. A single-incident matter may be fully covered in twelve questions; a four-month pattern complaint with five incidents needs the same twelve applied five times. What determines adequacy is whether every numbered allegation has been reached at the specifics stage — date, words, location, who was present, what followed — and whether the elements of the allegation type have all been asked about. Count coverage, not questions.
- Can I ask leading questions in an investigation interview?
- Sparingly, late, and on the record. There is a legitimate use at stage four: putting contrary evidence to someone requires stating it, and 'another account of this is X, what is your response?' is unavoidable and fair. What is not defensible is a leading question that establishes a fact in the first instance. If you lead, record the question verbatim and why it was necessary, so a reviewer can weigh what came after it separately from what came before.
- Should I ask every witness exactly the same questions?
- Ask the same opening block of everyone, because consistency there is what makes variance in the answers meaningful and demonstrates even-handed treatment. Beyond that, identical questions to differently placed witnesses waste the interview — someone who was on shift on the relevant afternoon and someone who transferred in a month later need different questions. Keep the core constant, vary the specifics, and record which set you used for each person.
- How do I ask about a protected characteristic without leading the witness?
- Ask about what was said and done before asking about the connection. 'What were the words used?' precedes 'was the comment connected to X?' If the connection is in the words, you now have it from the witness. If it is not, you can ask the connecting question knowing the answer is an inference the witness is drawing rather than an account you supplied — and you can record it as such.
- Can I ask the respondent whether the complainant has a reason to lie?
- Yes, and you should. The respondent is the person most likely to know of any motive to fabricate, and a credibility assessment that never sought one is incomplete. The obligation the question creates is that you then test whatever comes back. If the answer is 'she is angry about a shift swap I refused in March', that is checkable against a date and a system record. Left in the report unchecked, it is an untested allegation against the complainant published under your signature.
- What standard am I asking these questions to satisfy?
- The preponderance of the evidence — whether each allegation is more likely than not to have occurred. That matters at the interview stage, not just at the writing stage, because it tells you when to stop. Investigators who keep questioning toward certainty run out of time before they reach the third allegation, then record it as inconclusive for want of evidence they never went looking for. Ask to the point where the question is answered on the balance of the material, and move on.
- How do I question supervisor status without taking the complainant's word for it?
- Ask in the terms the test actually uses. Under Vance v. Ball State Univ., 570 U.S. 421 (2013) the question is whether the employer empowered that person to hire, fire, fail to promote, reassign to significantly different responsibilities, or make a decision significantly changing pay or benefits — not whether they direct daily work or hold the title. Ask the respondent what they would have to do to move someone off the team permanently, ask the complainant who signs off on those decisions, and then verify both against the delegation-of-authority schedule or the job description. A job title answers nothing.
The blank template
Exactly what is in the Word download, so you can see the structure before deciding whether you want it.
Matter details, allegation types and interview sequence
The four-stage funnel — applies to every interview in the matter
- Stage 1 — Free narrative: [Tell me in your own words what has been happening. Take as long as you need; I will ask questions afterwards.]
- Stage 2 — Expansion: [You mentioned X. Tell me more about X. What did that look like? Then what happened?]
- Stage 3 — Specifics: [When exactly? Where? What were the words used, as close as you can get? Who else was there or within earshot? What did you do immediately afterwards?]
- Stage 4 — Testing and closure: [Put inconsistencies and contrary evidence. Is there anything I have not asked you about that you think I should know?]
Universal opening block — run identically for everyone
- [Role, tenure, base, shift pattern and reporting line — confirmed on the record, not assumed from the org chart.]
- [How do you know each of the parties? Have you ever reported to either of them, or they to you?]
- [Is there anything about your relationship with either party — friendship, conflict, a shared history — that you think I should know?]
- [Has anyone spoken to you about this interview before today? Who, and what was said?]
- [Do you keep any record of your working day — a diary, a notebook, calendar entries, messages? Anything covering this period?]
- [Do you want a representative or support person present today?]
- [Confirm the non-retaliation position was explained.]
- [Do you have any concerns about repercussions for speaking to me today?]
The three party sets — what each one is for
- Complainant — [Does every numbered allegation now have a date, words, location, witnesses and an aftermath?]
- Respondent — [Was every allegation put separately and in full, and is every response classified as accepted, characterized or denied?]
- Witness — [Was anything reached only after allegation specifics were disclosed, and is the disclosure point recorded?]
Overlay — harassment allegations
- [How did you respond at the time — what did you say, what did you do, did you move away, did you stop attending something?]
- [Did you ever ask them to stop, or indicate you were uncomfortable? When, in what words, and was anyone else there?]
- [Did the conduct change after that? How, and for how long?]
- [How many times, over what period? Can you place the first and last occasions?]
- [Was anyone else present when it happened? Did anyone comment on it afterwards?]
- [Was the comment or conduct connected to [protected characteristic]? What was said that connects it?]
- [Subjective effect: what did it change about how you worked — what did you avoid, stop doing, or route around?]
- [Authority, in Vance terms: is the respondent empowered to hire, fire, promote, reassign you to significantly different work, or make a decision that significantly changes your pay or benefits? Who actually signs that off?]
- [Verification: what does the delegation-of-authority schedule or job description say the respondent can decide? Do not rest supervisor status on a job title.]
- [Did anything change in your job after this started — shifts, duties, overtime, training, a reference?]
- [Affirmative-defense line: were you aware of the anti-harassment policy and the reporting routes? Did you use them? If not, what stopped you?]
- [Did anyone in management see or hear any of it? Was it raised with the respondent, by whom, and what happened next?]
Overlay — discrimination allegations
- [Who made the decision? Who else was involved in it, and who was consulted?]
- [What reason were you given at the time? Was it in writing? Has the reason you were given changed since?]
- [What is the normal process for this decision? Was it followed here? What was different?]
- [Comparability first: who reported to the same manager, under the same rule, with a comparable record? Only then — what happened to them?]
- [What did the decision-maker know about [protected characteristic], and how did they know it?]
- [Were any remarks made about [protected characteristic]? By whom, when, and how close in time to the decision?]
- [Were you told about the opportunity? Who else was told, and how?]
- [Is there anything in writing — notes, criteria, scores, a shortlist, an email thread — that shows how the decision was made?]
- [Respondent-side: talk me through the decision as you made it. What were the criteria, and when were they set?]
- [Respondent-side: how have you handled the same situation with other employees in the last two years?]
Overlay — retaliation allegations
- [What did you do that you believe prompted this — a complaint, supporting someone else's complaint, a request for accommodation, a report to an agency? When exactly?]
- [Who did you tell, and in what form? Is there a record of it?]
- [Who do you believe knew about it? How would they have found out?]
- [Respondent-side: when did you first become aware of the complaint? Who told you, and what were you told?]
- [What changed afterwards? Be specific — duties, shifts, overtime, meetings you stopped being invited to, tone, supervision level, references.]
- [When did each change happen, in relation to the protected activity?]
- [Materiality: would that change have made you think twice about raising the complaint in the first place? What did it cost you in practice?]
- [Respondent-side: when was this decision made? Show me the record of it. Who else was affected by the same decision?]
- [Respondent-side: was anything already planned or underway before the complaint? What is the documentary trace of that?]
- [Was there any intervening event — a performance issue, a restructure, a client complaint — between the two?]
Overlay — bullying and hostile conduct with no protected characteristic
- [Describe the pattern — how often, over what period, in what settings?]
- [Was it in private or in front of others? Who was usually present?]
- [Does the same person behave this way toward others in the team, or only toward you?]
- [What is the reporting relationship? What can they decide about your work, hours or pay?]
- [Was there a stated performance concern? When was it first raised, and is it documented anywhere?]
- [Respondent-side: what was the performance issue, when did you first raise it, and where is it recorded?]
- [Respondent-side: how do you deliver the same feedback to other members of the team?]
- [Witness-side: how does the respondent address other people in the team? Is anything different about how they address the complainant?]
- [What effect has this had on your work — what have you stopped doing, avoided, or changed?]
- [Which provision of the policy do you say this engages, and what does that provision actually require?]
Overlay — policy and code-of-conduct breaches
- [Which policy, which version, effective from when? Where is the acknowledgment record?]
- [Were you given this policy? When, in what form? Did you receive training on it?]
- [Read the provision aloud and ask: what did you understand this to require?]
- [Walk me through what you did on [date], step by step.]
- [Do you accept doing [act]? Setting aside why — did it happen?]
- [What was the reason? Were you instructed to, or did you believe you had authorization?]
- [Has this been done this way before, by you or by others? How commonly?]
- [Respondent-side or manager-side: how has the organization handled the same breach in the last two years?]
- [What system records exist — access logs, expense records, timestamps, approvals, email trails?]
- [Was there any harm or loss? Has it been quantified, and by whom?]
Question forms that contaminate the record
- Leading — ["He blocked your path, didn't he?"] → rewrite as ["What happened when you got to the corridor?"]
- Compound — ["Did he say it, and did anyone else hear it?"] → split: ["What did he say?"] then ["Who else was there?"]
- Compound clarifier — ["Are you denying it, or saying you do not recall?"] → split: ["Do you say the conversation did not happen?"] then ["Do you deny it happened?"]
- Assumptive — ["When he made these comments, how did you react?"] → rewrite as ["You mentioned comments. Tell me about one you remember clearly."]
- Conclusory — ["Did you feel harassed?"] → rewrite as ["What did you do after it happened?"] and ["Did you say anything to him about it?"]
- Judgmental — ["Why did you wait three months to report it?"] → rewrite as ["Talk me through what happened between February and when you made the complaint."]
Re-interview and follow-up questions
- [Since we spoke I have received information that differs from your account on one point. I want to put it to you and give you the chance to respond.]
- [Another account of this is [X]. What is your response to that?]
- [Have you remembered anything further since we spoke, or found anything in writing?]
- [Have you discussed the matter with anyone since our last conversation?]
- [A new allegation has been raised: [allegation, verbatim, with date]. What is your response?]
- [Does anything in your earlier account need correcting?]
Closing block and note verification
- [Is there anything I have not asked you about that you think I should know?]
- [Is there anyone else you think I should speak to?]
- [Is there anything in writing I have not seen?]
- [Confirm next steps, the non-retaliation position, and how to reach me if they remember something later.]
Download this template
Free Word download, no email required. Or open it in Professional Drafter — it lands in your template library, ready to draft against.
Sources
- Meritor Savings Bank v. Vinson, 477 U.S. 57 (1986) — hostile environment under Title VII — Cornell LII
- Harris v. Forklift Systems, Inc., 510 U.S. 17 (1993) — objective and subjective test — Cornell LII
- Faragher v. City of Boca Raton, 524 U.S. 775 (1998) — Cornell LII
- Burlington Industries, Inc. v. Ellerth, 524 U.S. 742 (1998) — the affirmative defense — Cornell LII
- Vance v. Ball State University, 570 U.S. 421 (2013) — who counts as a supervisor — Cornell LII
- Burlington Northern & Santa Fe Railway Co. v. White, 548 U.S. 53 (2006) — materially adverse action — Cornell LII
- McDonnell Douglas Corp. v. Green, 411 U.S. 792 (1973) — Cornell LII
- NLRB v. J. Weingarten, Inc., 420 U.S. 251 (1975) — representation at investigatory interviews — Cornell LII
- NLRB — Weingarten Rights
- EEOC — Enforcement Guidance: Vicarious Liability for Unlawful Harassment by Supervisors (1999) — flagged on the EEOC site as superseded in 2024 by guidance since rescinded; carries the Vance notice
- EEOC — Enforcement Guidance on Retaliation and Related Issues (2016) — adopts the Burlington Northern standard
- EEOC — Harassment overview
Related templates
Complainant Interview Questions & Template
A complainant interview template with opening script, question sequence and note format — structured to itemize allegations rather than collect a narrative.
Investigation Interview Questions for the Accused Employee
A respondent interview template covering notice, the right to respond to each allegation, and the questioning sequence that makes a finding defensible.
Witness Interview Questions & Template
A witness interview template with an opening script, non-leading question structure, and a record of what the witness could and could not speak to.
Harassment Investigation Report Template & Example
A harassment investigation report template covering unwelcomeness, severity or pervasiveness, the protected characteristic, and employer notice — with a worked findings example.
This page is a documentation aid, not legal advice. Workplace investigation requirements vary by jurisdiction, sector and policy, and a template cannot tell you what your matter requires. The worked example describes a fictional scenario.