Thunder Bay Police at Ontario's Human Rights Tribunal

The shape of the record

Published decisions per year, 2009–2026. The 2025 spike (orange): 15 decisions — more than 2016–2021 combined — including seven employment matters decided across two days in June.

The funnel, by application

How they end

Coding: full texts with verified corrections; "no final outcome" includes continuing matters and endings that never produced a published decision. Settlements are otherwise invisible in this record — the one shown surfaces only through the Burns contravention litigation.

Not one of the 57 decisions orders a remedy. Every matter that produced a final decision ended in dismissal or deemed abandonment; the rest end off the record. Whether any claim had merit was never publicly decided in seventeen years.

Internal versus external: the defining split

A civilian alleging discrimination in policing services and an officer alleging discrimination in their employment are different legal worlds — different Code provisions, different defences, different power dynamics. This record is unusual for how much of it is the second kind.

Nearly two in three applications against Thunder Bay police came from inside the institution — and only one in three from the public. The 26 employment applications are not 26 people: four members account for 19 of them (six, five, four and four), and the remaining seven filed one apiece — multiple filings being typical where allegations span years or respondents, each file disposed of separately, which is why June 2025 panels dealt with four, five and six applications per applicant at once.

The chair's five applications sit on their own row because of what the Tribunal did with them, not because she stood outside the institution: it held she was neither an employee nor a service recipient — the jurisdictional hole her case exposed. In every ordinary sense she was an insider, and she is counted as one. At Durham, employment is about 14% of matters; this record is the inverse.

Why the unit matters

Counted as applications, employment leads 26 to 17. Counted as published decisions, it inverts — employment produces 15 of 57, civilian matters 35 — because one civilian matter, Rosario, generated eleven decisions alone. Applications are the honest unit for "how many claims were brought"; decisions measure how much paper a claim generated. We publish both and name the unit every time.

What people plead: grounds, by application

Unit: applications (of 50) whose decision texts engage each ground — the union across a file's decisions. Text-derived and conservative; pleading-level verification is the next coding pass.

Reprisal appears in 26 of 50 applications — more than half. In a record where 62% of applications come from inside the institution, punishment-for-complaining outrunning every identity ground except disability is not a statistical curiosity; it is the institution's self-description. The Indigenous-identity cluster appears in 19 of 50 applications (38%) — and not one of those nineteen has reached a merits hearing.

The quiet-years inversion

2016–2020 produced five decisions; 2019 — the year after Broken Trust and the Sinclair report — produced zero. The oversight system at its loudest coincided with the claims channel at its quietest; the wave arrived from 2021 onward, filed by the service's own members. Oversight and adjudication are not the same pipe, and the record shows them running on opposite clocks.

One adjudicator, one docket

The modern cluster has been case-managed and decided substantially by a single Vice-Chair, sitting alone or presiding on panels. We report this in aggregate and without a name, as a structural fact about a small adjudicative ecosystem — not an insinuation: concentration is an institutional design choice with consequences for consistency and for perception, in a city where every institutional actor is known to every other.

The June 2025 batch

Seven decisions across two days in June 2025 — and an eighth four days later. Eight member-applicants, all employment, all against the Board, whose application files span 2021 to 2023. The pattern across them: preliminary/summary hearings testing timeliness under s.34 and reasonable prospect of success — the paper gates, applied to the service's own members. Where final outcomes have since issued in this group, they are dismissals: Hanchuck (PTSD developed during employment; allegations before May 2021 untimely, the rest found to be "mere speculation" with no reasonable prospect — dismissed by a two-member panel, July 2026) and Dunning and companions proceeding through the same gate.

The cluster's origin is now a matter of criminal record: two of these officers reported a sergeant's warrantless entry in November 2020 — he told them it was "probably a Charter breach but not a big deal" — watched the matter cleared internally, and filed at the Tribunal in 2021 alleging that reporting it brought reprisal. In May 2026 that sergeant was convicted and sentenced to three years for obstruction and breach of trust. The officers who reported him are still litigating what reporting him cost them. Attrition since the batch: Nordal's reprisal claims dismissed June 2025 for lacking factual basis; McKay deemed abandoned June 2026; Hanchuck dismissed July 2026; five of Dunning's six applications dismissed. All eleven individually named command respondents: removed (the pattern).

Walmark shows the employment-side trap most clearly: his application was dismissed under s.45.1 because the WSIB process had "appropriately dealt with" the substance — the workplace-injury system displacing the human-rights forum. For injured officers, the road often ends before it begins.

Why the Tribunal, when officers have a union?

Police officers in Ontario are not short of forums. A serving member with a workplace grievance can normally go to grievance arbitration under the collective agreement; a workplace injury or psychological diagnosis goes to WSIB; misconduct by a colleague goes to the chief, then LECA; and discipline is dealt with under the policing statute. The Human Rights Tribunal is the slowest and least certain of these routes. So the fact that eleven members brought twenty-six applications to it is itself a finding — and the record suggests three reasons.

• Some claims exist nowhere else. Reprisal for asserting Code rights is a free-standing wrong under s. 8 of the Code. A collective agreement does not create it, and no grievance can award a Code remedy for it. Where the alleged harm is the punishment for complaining, the Tribunal is the forum — which is why "reprisal" is the most-pleaded ground in this entire record, at 203 mentions.
• The internal routes were the thing that failed. The cluster's documented origin is two officers reporting a superior's warrantless entry up their own chain of command and watching it cleared internally — the sergeant was later criminally convicted of obstruction and breach of trust. When the complaint mechanism is the alleged problem, using it again is not a remedy.
• And where another forum did act, the Tribunal stepped back. Walmark was dismissed under s. 45.1 precisely because the WSIB process had "appropriately dealt with" the substance. The Tribunal defers to those other routes when they work — which makes its own record of never reaching a merits hearing here harder to explain away, not easier.

How 48 matters ended

Decisions per matter

Sixty-three percent of matters produce a single published decision. One civilian matter produced ten — which is why counting decisions rather than matters inverts the whole picture of this service. By matters, employment leads. By decisions, civilian matters do.

Who decided these matters

Twenty-three adjudicators appear across 48 matters, in 58 assignments (panels count more than once). The distribution is unlike any other corpus in this project.

We name adjudicators because they sign every decision and those decisions are public; naming who decided what is reporting the record. What we do not do is characterise anyone. We report assignment alongside outcome and draw no inference about a person from a docket, because in a system where the overwhelming majority of matters end at procedural gates, the individual is not the variable. Our reasoning, in full.

Unit: adjudicator assignments (58) across 48 matters, from the matter index. Panels of three count as three assignments. Coding source: text-verified.

The settlement wars: the Burns arc

The only settlement visible in this record is visible because both sides litigated its aftermath. Everything about the visibility gap is in this one arc: resolution happens off the record, and the record only reopens when resolution fails.

The applicant who ran the board

Georjann Morriseau — former Fort William First Nation chief, appointed board chair in 2019 to implement the Sinclair recommendations — filed five applications after the service she oversaw investigated her. The Tribunal dismissed her discrimination claims for want of jurisdiction: a board appointee is neither an employee of the board nor a recipient of its services. Her reprisal allegations were allowed to continue, with a genuinely significant ruling attached: s. 8's protection covers instituting proceedings, even where the underlying claims fail — "the fact that the Tribunal thereafter determined it does not have jurisdiction... does not mean she was not subject to reprisal for instituting the proceedings." Then, June 2, 2026: the day before the scheduled hearing she asked opposing counsel's consent to adjourn but did not copy the Tribunal; counsel refused, copying the Tribunal; when the hearing convened she was not in attendance, and a three-member panel deemed the applications abandoned. No reconsideration or judicial review appears in reporting since; the 30-day windows for both have passed, leaving only discretionary extension.

The media record

This is among the most journalistically documented policing stories in Canada — the Globe and Mail on the charges and sentencing, CBC on the trials, APTN on Morriseau, TBNewswatch and the Chronicle-Journal on every hearing, Ricochet on the cluster's origins. This case study adds what journalism structurally cannot: the tribunal paper trail, complete, counted, and joined. Where our account and reporting diverge, we cite the decision text and say so.

Other decisions worth knowing

• Chookomolin (2025): a civilian sought to litigate "in the public interest." Dismissed — under s.35, only the Ontario Human Rights Commission may file public-interest applications. The OHRC has not filed one against this service, in either of the two eras of systemic-racism findings.
• Rosario (2012–2015): one services matter, at least nine published decisions — a demonstration of why decision counts overstate and understate at once.
• Nothing (2018): an Indigenous applicant's case deemed abandoned when he did not call into a teleconference. The entire dismissal is five paragraphs.

Time: every applicant, filing to last decision

Each bar is one applicant's proceedings — from their earliest application (file-number year; exact date where a decision states it) to their last published decision. Dots are decisions.

Settled matters are invisible in this colouring — settlement produces no published final decision. The one settlement we can verify (Burns, February 2011) appears only because its breach was litigated.

The exact dates the record supplies are stark. Hanchuck: application filed May 19, 2022; summary hearing June 2, 2026; dismissal July 2, 2026 — 4.1 years to a paper gate, never a merits hearing. DeBungie: filed August 18, 2020; dismissed February 2026 — 5.5 years. The June 2025 batch's files span 2021–2023; their summary hearings came three to four years later, together. And at the far end sits Ewing, filed September 2009 — the record's reminder that member complaints against this service are not new.

Unit: applicant-proceedings (consolidated files grouped). Filing approximated by file-number year except where decisions state the exact date (6 do). "Last decision" is not resolution for proceedings without a final outcome in the record.

What this record cannot show

Method: all 57 decisions machine-extracted and pattern-verified; critical decisions read in full; matter counts from file numbers within the texts; conservative classification. Dataset release to follow the same pipeline as Case Study 01.