---
title: "PHI403 Lecture 04 — What's in a Correlation?"
type: source
tags: [philosophy-of-science, correlation, statistics, hume, bradford-hill]
authors: [Anjum, Rani Lill; Rocca, Elena]
year: 2023
venue: "PHI403 Causation in Science, NMBU"
kind: handout
raw_path: "raw/Philosophy of Science/PHI302 04 What's in a Correlation.pdf"
created: 2026-04-20
updated: 2026-04-20
key_claims:
  - A correlation between A and B admits at least five causal interpretations: A→B, B→A, common cause, mutual causation, no causal link.
  - Spurious correlations can be near-perfect (divorce rate in Maine × margarine consumption) — correlation is clearly not causation.
  - Bradford Hill's nine criteria (temporal priority, consistency, strength, specificity, plausibility, mechanism) mix statistical and ontological considerations.
  - Modern statistics' hard work to separate genuine from accidental regularities implicitly concedes that causation is more than regularity.
---

# PHI403 Lecture 04 — What's in a Correlation?

A diagnostic of the [[concepts/regularity-theory-of-causation|regularity theory]] that anchors correlational science. Hume (*Treatise*, 1739) identified causation with constant conjunction — a view "still alive and well" in evidence-based decision-making.

The lecture distinguishes **five types of correlation** (co-variance, plain regularity, stable proportion, invariance, constants) and **five possible causal conclusions** from a correlation (A→B, B→A, common cause, mutual causation, no causal link — the last exemplified by *divorce in Maine × margarine in US*, r = 0.997).

**Bradford Hill's nine criteria** — temporal priority, consistency, strength, specificity, plausibility (mechanism), etc. — are introduced as the standard statistical toolkit. The lecture notes that some criteria are statistical and some ontological, and that statisticians' effort to separate genuine from accidental regularities tacitly concedes that causation is *more* than regularity — a tension between the *methodological* "correlations first" and the *ontological* "nothing over and beyond correlations".

## Connections
Back-link: [[sources/2023-anjum-rocca-phi403-causation-in-science]]. Concepts: [[concepts/regularity-theory-of-causation]] · [[concepts/causation]] · [[concepts/mechanisms-in-science]].