SF-0024Michigan Files
The textbook example of nature in balance is a study whose own researchers say it isn't
Isle Royale's wolves and moose have been counted every year since 1958. The Michigan Tech project that counts them writes that what guides nature is "rich, dynamic variation, not 'balance of nature'."
A bull moose browsing on Isle Royale. Moose reached about 2,000 on the island in 2019; the 2026 winter study counted 524.
Bull moose (Alces alces) browsing within Isle Royale National Park.
- Image
- Nate De Jager
- Taken
- 2017
- Location
- Isle Royale National Park, Keweenaw County, Michigan
- Archive ref.
- USGS media library — DeJager_Moose_ISRO
- Licence
- Public domain as a work of the US federal government, corroborated by the item's Sources/Usage statement (licence record)
- The image itself
- Authentic, unaltered
Bull moose, Isle Royale National Park, Michigan. Nate De Jager, US Geological Survey Upper Midwest Environmental Sciences Center, 2017. Public domain. Cropped.
Verdict
"Isle Royale is the textbook example — wolves and moose, perfectly balanced, going up and down together in a cycle."
The individual facts may check out, but the way they are framed leads to a false conclusion.
Nothing in the claim is invented. What fails is that it reports an early reading of the data as though it were the finding, and the effect on a reader is a belief the researchers explicitly disown.
ESTABLISHED, and not in dispute here: the study exists, it is extraordinary, and it is the reason anyone can adjudicate this at all. Annual observation of wolves and moose on Isle Royale began in 1958, initiated by Durward Allen, and has run continuously since — the longest study of its kind.
ESTABLISHED, and fatal to the claim: the project rejects the balance reading in its own words. "Rich, dynamic variation, not 'balance of nature' seems to be the force that guides nature." Its researchers record that early ecologists did conclude the populations "seemed to exhibit some kind of 'balance of nature'", and that longer observation did not sustain it.
The sharpest line is about prediction, and it is theirs: "The accuracy of our predictions for Isle Royale wolf and moose populations is comparable to those for long-term weather and financial markets." A system in a stable cycle is forecastable. This one is not, and the people with sixty-eight years of data say so.
The current count is the claim failing in real time rather than in principle. The 2026 winter study found an estimated 37 wolves and 524 moose — a 75% fall in moose since their 2019 high of about 2,000, with wolves near a historic high, and for the first time in sixty-eight years no calves at all during the moose count. Two populations moving hard in opposite directions is not what "in balance" describes.
That number carries a qualifier this file states rather than buries: the wolf population fell to two through inbreeding depression and was restocked with nineteen wolves by the National Park Service in 2018 and 2019. So "wolves near a historic high" is partly a management outcome. It is also why the verdict does not rest on the 2026 counts, and rests instead on what the project says about its own sixty-eight years.
Misleading rather than False, and the distinction is the point. The study is real, the island is real, the early conclusion was real and honestly reached. Only the tense is wrong: a finding that was superseded is still being taught as current.
Misleading requires 1 primary or authoritative source. This file cites 4.
In short
Isle Royale is taught as the clean case: an island, wolves, moose, and two populations rising and falling in step. The project that has run the count since 1958 says the opposite. Its researchers record that early ecologists did read a "balance of nature" into the data, and that longer observation dissolved it — that their forecasting accuracy is comparable to long-term weather and financial markets. The 2026 count found 37 wolves and 524 moose, a 75% fall in moose since 2019. The study is real and the island is real. The balance is the part that was added.
The study is real
An island in Lake Superior. One predator, one main prey, and no easy way in or out for either. Annual observation of wolves and moose began there in 1958, initiated by Durward Allen, and has never stopped. The project describes itself as “the longest continuous study of any predator-prey system in the world.”
That is precisely why it ended up in textbooks. If you wanted to design a natural laboratory for the question, you would design Isle Royale.
The balance was added
Early on, the numbers did look like equilibrium, and the project says so in its own history: ecologists “concluded that they seemed to exhibit some kind of ‘balance of nature.’”
Then the count kept running, and the same history records the moment it stopped fitting:
By 1969 the moose population had doubled, and one now had to admit there had been a major shift in the balance.
The project’s own summary of what it found is not the textbook version:
Rich, dynamic variation, not “balance of nature” seems to be the force that guides nature.
And the line that ought to end the argument, because it is a testable claim about their own work:
The accuracy of our predictions for Isle Royale wolf and moose populations is comparable to those for long-term weather and financial markets.
A system in a stable, repeating cycle is forecastable. That is what a cycle is. Sixty-eight years of data, gathered by the people making the statement, say this one is not.
What it looks like right now
The 2026 winter study, run from 22 January to 3 March, estimated 37 wolves and 524 moose.
Moose peaked at about 2,000 in 2019. That is a 75% fall in seven years, while the wolf count sits near a historic high and predation runs close to the highest rate ever observed — roughly a quarter of the moose population taken in a single year.
And a first, in a study old enough to have very few of them left: for the first time in 68 years, the count found no calves at all. Sarah Hoy, who co-leads it: “In the past year, I have only seen one moose calf.”
Rolf Peterson’s description is of a system at an edge, not at rest: the populations are “approaching the edge of where they have been in the past, with moose low and wolves high.”
The wolf number has a qualifier attached
It has to be said in the same breath, because Michigan Tech says it in the same release. A decade ago there were two wolves left on Isle Royale, “due to inbreeding depression”. In 2018 and 2019 the National Park Service translocated nineteen more to stop the population disappearing.
So the thirty-seven are not a natural rebound, and anyone told “wolves at a near-record high” without that sentence has been handed the wrong picture. It is also why this file does not hang its verdict on the 2026 numbers. They illustrate; the sixty-eight years of the project’s own conclusions carry the weight.
What this file does not claim
That the study failed. It did the opposite of failing: it kept counting until its own early conclusion stopped fitting, and then said so. Revising an answer after sixty-eight years of observation is the method working.
That predator and prey are unrelated here. Wolves are a major cause of moose mortality on Isle Royale, and in the most recent year that pressure was near a record. What is being rejected is stability and predictability — not the relationship.
And it does not claim that one year’s numbers refute a cycle. They cannot; cycles have troughs. The weight is carried by the researchers’ own general statements about variation and forecast accuracy. The 2026 counts are an illustration of the point, not the proof of it.
The evidence
The study is real: "Researchers began annual observations of wolves and moose on Isle Royale in 1958", initiated by Durward Allen.
Cited sources:S1
And it is the longest of its kind, which the project states on its own front page rather than in the history cited above: "This research project is the longest continuous study of any predator-prey system in the world."
Cited sources:S4
The early balance reading was genuinely made, and the project records it: ecologists "concluded that they seemed to exhibit some kind of 'balance of nature.'"
Cited sources:S1
The project also records when the reading broke, in its own words rather than a summary of them: "By 1969 the moose population had doubled, and one now had to admit there had been a major shift in the balance."
Cited sources:S1
And the project rejects it, in its own words: "Rich, dynamic variation, not 'balance of nature' seems to be the force that guides nature."
Cited sources:S1
A balanced cycle would be forecastable. The researchers say theirs is not: "The accuracy of our predictions for Isle Royale wolf and moose populations is comparable to those for long-term weather and financial markets."
Cited sources:S1
The current counts are not a balance. The 2026 winter study, conducted 22 January to 3 March, estimated 37 wolves and 524 moose.
Cited sources:S2
The moose trajectory is a collapse, not an oscillation around a mean: a 75% drop since 2019, when moose on Isle Royale reached a high of about 2,000.
Cited sources:S2
Predation is running at an extreme rather than a steady state: wolf predation rates were "close to the highest level ever observed", with about 24% of the moose population killed by wolves over the past year.
Cited sources:S2
The researchers describe the system as at an edge rather than at equilibrium. Rolf Peterson: "The wolf and moose populations are now approaching the edge of where they have been in the past, with moose low and wolves high."
Cited sources:S2
The count produced a first in the study's whole history: "For the first time in 68 years, the winter study research team observed no calves during the annual moose count." Sarah Hoy: "In the past year, I have only seen one moose calf."
Cited sources:S2
Evidence that cuts the other way
Findings that complicate or argue against the verdict above. They are listed because leaving them out would make the file look stronger than it is.
The wolf number is not a purely natural outcome, and the source this file relies on for it says so on the same page. A decade ago the population "had dwindled to just two wolves due to inbreeding depression", and "In 2018 and 2019, 19 wolves were translocated to Isle Royale by the National Park Service to prevent the island's wolf population from going extinct." "Wolves near a historic high" is therefore partly a management result, and a reader who takes 37 as a natural rebound has been given the wrong picture. It weakens the 2026 counts as an illustration — which is one more reason the verdict does not rest on them.
Cited sources:S2
This file does not claim the study failed or that its early researchers were careless. The opposite: the balance reading was a reasonable conclusion from the data available, and it was revised because the count kept running. A study that changes its answer after sixty-eight years of observation is the system working, not failing.
Cited sources:S1
Nor does it claim predator and prey are unrelated on Isle Royale. Wolf predation is described by the researchers as a major cause of moose mortality, and the 2026 figures put it near a record. Rejecting "balance" is rejecting a particular claim about stability and predictability, not the existence of a predator-prey relationship.
Cited sources:S2
A single year's counts cannot by themselves refute a cycle, since a cycle has low points. What does the work here is the project's own general statement about variation and forecast accuracy; the 2026 numbers illustrate it rather than prove it.
What we still don’t know
- Whether the populations cycle on a longer period than the record can resolve. Sixty-eight years is long for a field study and short for an ecological cycle, and the project's own position is that variation has been continuous and driven by contingent events rather than periodic. This file follows that position; it does not establish that no cycle of any length exists.
- What is driving the moose collapse. S2 does name a driver — "The sharp decline in the moose population has been driven by an increase in wolf predation" — and this file reports that. What it does not have is the relative weight of that against the other pressures the project discusses elsewhere, such as winter ticks, summer heat and balsam fir availability. The single cause is the source's; the apportionment is nobody's, here.
- Why the wolf population is where it is. The inbreeding crash and the 2018–2019 translocation of nineteen wolves are recorded in the contradicting evidence, from S2, because they qualify the number this file quotes. What is not established here is how much of the 2026 count and the predation rate follows from that intervention rather than from the system left alone. No source read for this file apportions it, and this file does not guess.
Sources
Every source below was read in full before it was cited. Archive snapshots are recorded at research time, because a file whose sources have evaporated is indistinguishable from one that never had any.
S1Tier 1Read in full
The Study — Wolf-Moose Project
Wolves and Moose of Isle Royale, Michigan Technological University
The spine of this file, and the source that makes the verdict possible: the project's own account of what it has and has not found. Establishes the 1958 start and Durward Allen's initiation; that early ecologists concluded the populations "seemed to exhibit some kind of 'balance of nature'"; that the project's own reading is "rich, dynamic variation, not 'balance of nature'"; and the comparison of its forecast accuracy to long-term weather and financial markets.
S2Tier 1Read in full
Isle Royale Wolf Population Nears Historic High as Moose Numbers Plummet
Cyndi Perkins · Michigan Technological University · 2026-04-27
The current state of the system, from the university that runs the count. Establishes the 2026 winter study dates of 22 January to 3 March; the estimates of 37 wolves and 524 moose; the 75% moose decline since the 2019 high of about 2,000; predation close to the highest rate ever observed at about 24% of the moose population; no calves observed for the first time in 68 years; and Rolf Peterson's description of the populations approaching the edge of their historical range. It is also the source for the fact that qualifies its own headline number: the wolf population fell to two through inbreeding depression, and nineteen wolves were translocated by the National Park Service in 2018 and 2019. That is in this release, not in outside material, and this file records it in the contradicting evidence for that reason. Not independent of S1: the same institution and the same research team, reporting on the same project. Treated as one voice reporting different things, not as corroboration.
S3Tier 1Read in full
Bull moose at Isle Royale National Park — USGS media library item
Nate De Jager · US Geological Survey, Upper Midwest Environmental Sciences Center · 2017
This file's photograph and its rights record: the item-level "Sources/Usage: Public Domain." statement, and the credit naming the photographer's USGS centre, which is what establishes the us-federal-work basis under the rule in EDITORIAL_MANUAL.md §4. Independent of the Michigan Tech sources — a different agency, cited only for the image.
S4Tier 1Read in full
Wolves and Moose of Isle Royale — project home page
Wolves and Moose of Isle Royale, Michigan Technological University
One sentence, cited for one claim: "This research project is the longest continuous study of any predator-prey system in the world." The word "longest" appears nowhere on S1, the page this file otherwise relies on, so the superlative is sourced here rather than asserted. Not independent of S1 or S2 — the same project and the same institution, a different page.
Checked for circularity: S1 and S3; S2 and S3; S4 and S3 were confirmed not to derive from one another. Two sources that trace back to the same origin are one source, and counting them twice is the most common way a false claim looks well-evidenced.
How we checked this
What we went looking for, and what we found
Three checks, one of which set the verdict and one of which pulled the file back from overreach.
First, on False versus Misleading. False was tempting because the project contradicts the balance reading directly. It was rejected because the claim's components are all true — the study exists, the island is isolated, the early conclusion was genuinely reached and honestly published. What is wrong is the tense: a superseded finding taught as current. Misleading is the verdict that describes that, and the reasoning says so rather than leaving the choice implicit.
Second, against the correction overshooting into "predator and prey are unrelated here", which would be worse than the claim. The 2026 report puts wolf predation near the highest rate ever observed, at roughly a quarter of the moose population in a year. Rejecting "balance" rejects stability and predictability, not the relationship, and the contradicting evidence says that explicitly.
Third, against resting the verdict on one year's numbers. A single low count cannot refute a cycle, because cycles have troughs. The load is therefore carried by the project's own general statements about variation and forecast accuracy, with the 2026 counts illustrating rather than proving. That was a deliberate demotion of the most dramatic material in the file.
A fourth line was checked: Isle Royale's wolves went through severe inbreeding and were augmented by translocation from 2018. The first draft of this file set it aside as out of scope on the grounds that the primary material had not been read. The cooling-off pass established that this was wrong — the inbreeding crash and the nineteen translocated wolves are both stated in S2, the release this file read in full and quotes for its headline numbers. It is now reported in the contradicting evidence and in the body, where it belongs, because it qualifies the number this file quotes rather than sitting outside it.
- We searched for evidence against our own conclusion
- We traced where the claim started
- The claim is stated in its strongest form, not a weak version
- The photograph was verified independently of the story
- We checked whether this was settled after the popular version froze
- Image rights were traced to a stated licence
- We checked what we were leaving out
- The verdict sat before publication, and was re-read cold
- Confidence in the verdict
- high
- Verdict assigned
- Most recent authoritative source
- Researched by
- The Strange File
RESEARCH AND IMAGE COMPLETE. TWO GATES OUTSTANDING.
This is the first file in `michigan-files`, a cornerstone category that has held zero files. The backlog recorded the reason as image rights rather than weak subjects, and that is what was tested first here: the rights were established before any research was done, on the principle that a file which cannot be illustrated should not be written. The category test — a Michigan subject with Michigan sources — is met by Michigan Technological University, which has run the count since 1958 and publishes both sources.
Source independence is stated rather than implied. S1, S2 and S4 are the same institution and research team; they are treated as one voice reporting different things. Only S3, the USGS image item, is independent, and it is cited for the photograph rather than for the claim.
COOLING-OFF HAS NOW RUN, by an independent session on 2026-08-05, and it found four defects. They are recorded in full under coolingOffNotes rather than summarised away. Three of them were this file misdescribing what its own sources contained, which is the failure mode a fresh-eyes pass exists to catch: the author remembers what they concluded, not what they read.
OUTSTANDING — the owner has not approved publication. DRAFT is not permission to publish, and the calendar side of the cooling-off gate does not clear until the day after the verdict was assigned.
ARCHIVE SNAPSHOTS — all sources carry an archiveUrl. Corrected 2026-08-05: an earlier pass recorded some of them as archiveCheck: none-exists, on the strength of the Internet Archive's availability API, which had answered that no capture existed. It was wrong, and so was the note that called that "a fact about the archive's coverage". The CDX API lists captures for every one of them. The availability endpoint returns an empty answer for URLs the archive demonstrably holds, and an empty answer is indistinguishable from an authoritative no. `npm run archive` now queries CDX for exactly this reason.
No sensitivity flag. The subject is an ecological claim. The only individuals named are the researchers and the photographer, in each case because their own institutions name them.
Part of a collection
The Correction OvershootsDebunking has a second error, and almost nobody guards against it.
The Qualifier Is The StoryThe interesting content is in the clause everyone skims.
Also related
Yellowstone is not overdue, and the recurrence arithmetic gives it another 100,000 years
The USGS computes the average interval between Yellowstone's eruptions — 725,000 years — calls it meaningless because it rests on two intervals, and notes that even if you trust it, there are about 100,000 years to go.
Nobody has settled how Devils Tower formed, and the Park Service says so
The rock is phonolite porphyry and the columns are cooling cracks — that much is agreed. How the magma got there is not. The NPS lists four origin hypotheses and says discrepancies remain for all of them.