Resources · About this site

How this site works: evidence rubric, sources and corrections

6 min · Evidence: official · Updated September 3, 2026 · next review in 6 months

Every claim carries an evidence level. A citation is published only after a person or an in-session fetch opened the page. In March 2026 we found seven fabricated sources in our own research and removed them; the log is below, and it stays public. If you find an error, tell us and we will log the correction.

You have been reading this site for a while and you want to know whether to trust it. Fair. Here is how it is built, what it will not do, and where it has been wrong.

The evidence rubric

Every page carries one overall evidence level in its header, and the claims that matter carry an inline tag. The levels are assigned, not implied.

LevelMeaningTypical basis
StrongWell replicated; unlikely to reverseRandomized trials or meta-analyses with replication across groups
ModerateProbably right; details may shiftMixed results, one good trial, or consistent observational work
EmergingInteresting, earlySmall samples, single cohorts, pilot studies, new mechanisms
ContestedCredible disagreementExperts or communities disagree in good faith; both sides stated on the page
OfficialInstitutional fact, not a research claimGovernment, hospital or organization pages: criteria, programs, costs
EditorialOur reasoned viewSynthesis, lived experience, clinical common sense; labelled so you can weigh it

The header level is about the page's main message; the inline tags are about individual sentences. Contested topics, such as ABA or polyvagal framing, get a "Where the disagreement is" section rather than silence.

The source rule

The rule that governs everything else on this site:

  • AI-drafted text is allowed. Most first drafts here are produced with AI assistance and then edited by a person.
  • AI-supplied citations are not published until a person, or a logged in-session fetch, has opened the page and confirmed it exists and says what we claim. A citation that cannot be opened goes into a private "to verify" list in the page's metadata, never into the visible Sources section.
  • No numbers from memory. Percentages, sample sizes, dollar amounts, wait times and dates appear only with a source that was opened. If we cannot source a number, we describe the direction of the finding in words and tag the evidence honestly.
  • Official first. For Quebec and Canadian facts we prefer quebec.ca, canada.ca, CIUSSS sites, the Fédération québécoise de l'autisme, Autisme Montréal, Retraite Québec and Revenu Québec. For research we prefer PubMed, PubMed Central and Cochrane.
  • Dynamic facts are dated. Wait times, prices, benefit amounts and program criteria are stated as ranges, with the source and its date, and with the instruction to confirm on the official page before acting.

Review cadence

Page typeReviewed at least every
Quebec pathway, money and benefits, the Montreal directory, this page6 months
Evidence and daily-life pages12 months
Identity pages and the ABA page12 months, plus review by at least one autistic adult before the page is promoted

Each page shows its last editorial update date in the header. This date does not certify specialist or autistic-adult review. Pages awaiting that review say so in their text. The schedule above is our review target.

Reviewers

Identity, ABA and "what autistic adults wish parents knew" pages are to be reviewed by at least one autistic adult before we promote them; as of September 2026 that review has not happened yet, and each of those pages says so. Pathway and benefits pages are checked against the official source by a person. Reviewers are named here only with their consent. No reviewer has yet chosen to be named; this section will be updated when one does.

The corrections log

Corrections stay public. Each corrected page links to its own entry here.

15 March 2026: fabricated and unverifiable sources found and removed

During a systematic fact-check of the research compendium that underlies this site's evidence pages, we verified 20 core sources and found that seven citations did not exist or could not be traced, and that twelve more needed correction. All were removed or corrected before any of the affected material was published in English.

Removed as fabricated or unverifiable:

  • "Kinsella et al. 2016, Frontiers in Psychology": no such paper in PubMed, Google Scholar or Frontiers.
  • "Lloyd et al. 2016, Developmental Psychology": no paper with that title, authors and journal.
  • "Sheridan et al. 2022, Frontiers in Psychology": does not exist; likely a language-model fabrication.
  • "Szmagier & Stuhlman 2007": not in any database.
  • "38.4 hours per week of unpaid parental labour": no source; a generated number.
  • "60% of parents self-recognize their own neurodivergence": no verifiable source.
  • "73% burnout" attributed to Yorke et al. 2018: not in that paper.

Corrected:

  • Hillman et al. 2008: journal is Nature Reviews Neuroscience, not Neuroscience.
  • Miodrag & Hodapp 2010: journal is Current Opinion in Psychiatry.
  • Hayes & Watson 2013: restated as a meta-analysis of parenting stress in autism versus non-autism families; the original attribution was wrong.
  • "Ratey & Loehr 2011": replaced by Ratey 2008, Spark, a popular-science book; its test-score claims are not peer reviewed.
  • Cage et al. 2018: replaced by Cage & Troxell-Whitman 2019 and Hull et al. 2021.
  • Fournier et al. 2010: "79% of autistic children" changed to "up to 80% estimated"; the paper reports effect sizes, not prevalence.
  • Papadopoulos et al. 2012: removed a "2 to 3 years behind" claim that is not in the paper.
  • Drowning statistic: "90%" corrected to about 80% for drowning, suffocation and asphyxiation combined.
  • Dog-assisted therapy review: attribution corrected to the actual 2024 authors.
  • Nature meta-analysis: journal corrected to JAMA Network Open.
  • TIME-A music therapy trial: a "-3.73%" figure clarified as percentage points, with its p-value.
  • Skoglund et al.: year corrected to 2024 (online November 2023).

Also corrected in the statistics set: a CDC data brief cited for adult ADHD was about children aged 5 to 17; "boys twice as likely to be diagnosed with ADHD" was overstated (1.8 to 1); and a figure conflating a CDC model with a WHO estimate was split into its two sources.

What changed in our process as a result: the source rule above, the private "to verify" list on every page, the inline evidence tags, and this log.

Later entries

None yet. New entries are added at the top of this section with the date, the page, what was wrong, and what changed.

How to report an error

If a link is dead, a number is wrong, a program has changed, or a claim reads as stronger than its source, use the email link on our contact page and quote the sentence. We check every report, correct the page, and log the change here with the date. You can ask not to be named; you will not be named without asking.

The clinical boundary

This site explains. It does not diagnose, does not rank treatments for an individual child, and does not replace a clinician, a CLSC caseworker or your child's team. Where we say what a parent can do themselves, it is because the evidence supports that the parent is part of the mechanism. Each page carries one standing disclaimer to this effect, once, so that the text itself can speak plainly.

This week

  • Pick one claim on any page here and open its source. If it does not say what we said, tell us.

Sources

Questions parents ask

Is Neurosphere medical advice?
No. The site explains evidence and the Quebec system in plain language. It does not diagnose, does not recommend a treatment for your child, and does not replace a clinician. Where we say what a parent can do, it is because the evidence supports that a parent can do it.
What does 'Evidence: strong' mean on a page?
That the claim rests on randomized trials or meta-analyses with replication. 'Moderate' means mixed results or a single good study. 'Emerging' means small or observational work. 'Contested' means credible experts or communities disagree. 'Official' means a government or institutional page. 'Editorial' means it is our reasoned view.
Why is there a corrections log?
Because in March 2026 we found that seven citations in our research compendium did not exist or could not be traced, and twelve more needed correcting. We corrected them, changed our process, and decided the record should be public. A site that asks parents to trust it on evidence has to show its own mistakes.

Part of: We've been at this a while · plain markdown ·

This page explains; it does not diagnose or replace your clinician. Services, wait times, prices and benefit amounts change: confirm on the official page before acting. Found an error? Tell us and we will correct it publicly.