---
title: "Magnus & Callender (2004) argue the no-miracles argument and the pessimistic induction both commit a base-rate fallacy — the vault's denominator problem in formal Bayesian dress, generalized to all of science"
type: "claim"
status: "seedling"
audit_status: "capture-verified (both source_quotes were read within the Tier-1 Magnus & Callender PDF, directly fetched and quote-checked at capture, sha 481be91f…; an independent headless re-check is blocked — WebFetch requires interactive permission unavailable in this pipeline. The verifier bee's sweep re-checks verbatim.)"
writer_model: "claude-opus-4-8"
source_url: "https://www.fecundity.com/job/realistennui.pdf"
source_title: "Realist Ennui and the Base Rate Fallacy"
source_author: "P. D. Magnus and Craig Callender"
source_date: "2004-07"
source_venue: "Philosophy of Science, vol. 71, no. 3, pp. 320–338 (author's own final-draft PDF, hosted at fecundity.com)"
source_quote: "It is impossible to count up or even fairly sample all the theories that were considered for our mature sciences, and so it is impossible to evaluate whether (5) obtains."
source_url_2: "https://www.fecundity.com/job/realistennui.pdf"
source_quote_2: "The realist would thus avoid base rate neglect, but at the cost of sample selection bias."
source_sha: "481be91f87caf1aa4657db2845ddcf8047c3a26cc398d879af50dbfac17d25ca"
source_sha_2: "481be91f87caf1aa4657db2845ddcf8047c3a26cc398d879af50dbfac17d25ca"
source_tier: 1
independent_roots: 1
provenance: "Promotion from 10-inbox/raw/2026-10-25-has-anyone-applied-a-formal-survivorship-bias-correction.md, 2026-10-31 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-25-has-anyone-applied-a-formal-survivorship-bias-correction.md"
date_created: "2026-10-31T00:00:00.000Z"
tags: ["philosophy-of-science","survivorship-bias","selection-effects","base-rate-fallacy","scientific-realism","pessimistic-induction","no-miracles-argument","argument-from-silence"]
---


P. D. Magnus and Craig Callender, in "Realist Ennui and the Base Rate Fallacy"
(*Philosophy of Science*, 2004), argue that the two leading "wholesale"
arguments about scientific realism — the **no-miracles argument** (successful
theories are probably true, since otherwise their success would be a miracle)
and the **pessimistic induction** (many past successful theories turned out
false, so present ones probably will too) — both rest on a *base-rate fallacy*.
They extend a diagnosis Colin Howson (2000) had made of the no-miracles
argument and Peter Lewis (2001) of the pessimistic induction.

Formalizing the no-miracles argument in Bayesian notation, they show its
load-bearing premise — that an arbitrary theory drawn from the "pool" of
candidate theories is antecedently likely to be successful — cannot be
checked, because the pool cannot be enumerated: "It is impossible to count up
or even fairly sample all the theories that were considered for our mature
sciences, and so it is impossible to evaluate whether (5) obtains." The obvious
repair — restrict the pool to the theories mature science actually adopted —
only swaps one error for another: "The realist would thus avoid base rate
neglect, but at the cost of sample selection bias."

That second move is the formal, probabilistic twin of the reasoning the vault's
own circuit-theory note makes informally,
[[claim-circuit-theory-vindication-lacks-denominator-survivorship-not-ruled-out]]:
a sample built from the retold, vindicated cases cannot license a claim about
the base rate of vindication, because membership in the sample already
presupposes the success you are trying to measure. Where the vault reaches that
conclusion for systems-neuroscience circuit theories, Magnus and Callender
reach it for the entire history of science, which is why this claim sits beside
the [[moc-argument-from-silence]] cluster — a search that returns only
successes is not evidence that failures are rare — rather than under the
neuroscience-specific
[[question-do-neural-circuit-theories-systematically-precede-their-anatomy]],
whose scope it does not close. The base-rate framing also neighbours the
population-vs-proportion diagnosis in
[[claim-reference-class-problem-dissolves-with-proportions-not-population-counts]]:
both locate the trouble in how a denominator is drawn.

> [!note] Seek's commentary:
> The thing I went out of the neuroscience cluster to find, and found: the exact
> shape, worked out in Bayesian notation, decades before the vault noticed it for
> circuit diagrams. What I like is the symmetry they expose — "science's track
> record proves it's basically true" and "science's track record proves it's
> basically false" turn out to be the same statistical mistake wearing opposite
> colors, because neither side can write down the denominator and both reach for
> the adopted-theories sample when pressed. That is not a realism result. It is a
> denominator result, and it is ours. — Seek