---
title: "Chapter 10. From Evidence to Investigation"
book: "Why Minds Break: An Evolutionary and Developmental Programme for Psychopathology"
author: Ian Pitchford
url: https://evidencepress.org/why-minds-break/chapter-10/
---

# Chapter 10. From Evidence to Investigation

The final chapter proposes a research registry of separately evidenced claims, tests it on premenstrual dysphoric disorder and psychosis, and states how the programme could fail.

The closing chapter asks researchers to pose a different question before a study starts: which observations would separate a proposed explanation from its closest rivals, and what practical choice would turn on the answer. It develops the research classification promised in Chapter 1, called a registry. Each entry is a bounded claim with its own evidence, so a phenotype, an association, an experimentally shown dependency and a computational interpretation are logged separately. Evidence supports a claim only to the extent that the study ruled out the alternatives it actually tested. A computation therefore cannot borrow the strength of a dependency, and a person may need care when neither has been identified.

The chapter keeps Murphy and Stich's distinction between malfunction and mismatch but declines to treat such categories as mutually exclusive. It asks instead where a causal contribution lies, what keeps it going, in which settings it produces which outcomes, and what shows defence, dysfunction or trade-off. Overlap between diagnoses, seen in Danish registers, the Dunedin cohort and a fourteen-disorder genomic study that found five factors alongside a broader shared component, cannot by itself pick out its cause. It lists six rival structures, among them shared causes, cascades, reciprocal loops and measurement artefacts, which may coexist. Calling a finding transdiagnostic describes its scope, not a common mechanism.

The proposal borrows almost everything from RDoC, HiTOP, network theory, the Hierarchically Mechanistic Mind, Del Giudice's fast–slow–defence model and Hunt and Jaeggi's DCIDE. What it contributes is a rule for how results may pass from one claim to another. In premenstrual dysphoric disorder, hormone suppression and addback experiments support a dependency on hormonal change in susceptible women, though change versus level is not settled and no computation has been measured. In psychosis, raised striatal dopamine on average and moderately effective drugs, some acting away from D2 receptors, are supported, though a drug's benefit does not show where illness began; meanwhile a 2026 meta-analysis saw little average group difference in how strongly induced perceptual expectations were used, which undercuts a general computational deficit.

A first research programme should be modest enough that failure would be unmistakable. Taking the idea that some people with psychosis overweight unreliable evidence, it would name rival models first, check that the parameter can be recovered and measured reliably, test whether it improves prediction beyond strong comparisons, test causation with an intervention, and fix in advance what to retire. The framework should lose support if its distinctions keep failing to improve decisions, and the registry itself can be tested against careful existing methods. The book ends with David Hull's description of organisms as genealogical actors in ecological roles, and calls its proposal modest and answerable to evidence.

## Questions the chapter asks

- If several diagnoses tend to occur together, does that mean they share one underlying cause?
- What would it take for a computational theory of psychosis to earn a place in classification?
- How would anyone know whether this way of organising research had failed?

This is a summary of the chapter, written for this site; it is not an extract from the book. Coming soon from Amazon in Kindle, paperback and hardback editions.

*Why Minds Break* and these summaries are © 2026 Ian Pitchford, all rights reserved. The cover drawing is by Santiago Ramón y Cajal (1915), public domain.