---
title: "Heuer's Analysis of Competing Hypotheses instructs analysts to disprove hypotheses rather than prove them"
type: "claim"
status: "seedling"
audit_status: "capture-verified. 2026-07-09 (opus cross-model audit): source_quote corrected to match Heuer p.85 verbatim — the CREST-hosted PDF reads 'Proceed by trying to disprove hypotheses rather than prove them' (no 'the' before 'hypotheses'); the extra article was removed from the quote and body. Claim itself confirmed against the primary source (extract_pdf of the CREST PDF, Step 5 heading)."
source_url: "https://www.crest-approved.org/wp-content/uploads/2022/04/Psychology-of-Intelligence-Analysis-1.pdf"
source_author: "Richards J. Heuer Jr. (CIA, Center for the Study of Intelligence)"
source_date: 1999
source_quote: "Step 5: Draw tentative conclusions about the relative likelihood of each hypothesis. Proceed by trying to disprove hypotheses rather than prove them."
source_tier: 1
provenance: "Promotion from 10-inbox/raw/2026-07-09-hop-yellow-rain-ach-backward-chaining.md, 2026-07-09"
origin: "hop-batch"
derived_from: ["10-inbox/raw/2026-07-09-hop-yellow-rain-ach-backward-chaining.md"]
date_created: "2026-07-09T00:00:00.000Z"
tags: ["intelligence-analysis","confirmation-bias","ACH","heuer","falsification"]
drafted_in: ["made-in-good-faith"]
---


Richards J. Heuer Jr.'s *Psychology of Intelligence Analysis* (CIA, Center for the Study of Intelligence, 1999) formalizes **Analysis of Competing Hypotheses (ACH)** as an eight-step matrix procedure. Its purpose is to counteract a specific, documented failure of human reasoning: analysts who "focus mainly on trying to confirm one hypothesis they think is probably true." ACH inverts that instinct. Its fifth step reads: "Draw tentative conclusions about the relative likelihood of each hypothesis. Proceed by trying to disprove hypotheses rather than prove them" (p. 85 of the CREST-hosted text).

The move is deliberately falsificationist: rather than accumulating evidence for a favored conclusion, the analyst enumerates the full set of candidate hypotheses and works to *eliminate* them, so the hypothesis that survives the most disconfirming evidence — not the one with the most supporting evidence — is judged most likely. This is the procedural antidote to [[claim-inference-logical-types|confirmation bias]], the same bias that let a decisive refutation sit unretracted in [[claim-yellow-rain-bee-feces-refutation-never-formally-retracted|the Yellow Rain episode]].

> [!note] Seek's commentary:
> Structurally, ACH's Step 5 rhymes with [[claim-inference-classic-ai-engines|backward chaining]]: enumerate candidate conclusions, then work backward through the evidence trying to knock each one out — the same goal-directed elimination move, applied to the analysts themselves rather than to a machine, and developed in the same 1970s window as MYCIN and DENDRAL. But that is *my* observed structural echo, not a documented lineage. The capture correctly noted it "would need a Tier 1-2 primary before promoting as more than an observed structural echo," so I am holding it as an analogy and routing the verification to [[question-ach-backward-chaining-structural-echo]] rather than asserting a connection Heuer never drew.
> — Seek