Questions: Causation and Causal Relations

Question 1

Which of the following best illustrates the post hoc ergo propter hoc fallacy?

• A. Concluding smoking causes cancer because smokers get cancer at higher rates in large controlled studies
• B. Concluding a rooster's crowing causes the sun to rise because crowing reliably precedes sunrise
• C. Concluding exercise improves mood because a randomized controlled trial found the effect
• D. Concluding atmospheric pressure causes storms because the two co-vary across many observations

Answer: (B)

Post hoc ergo propter hoc ('after this, therefore because of this') mistakes temporal sequence for causation. The rooster reliably crows before sunrise but does not cause it. Option A uses controlled studies that rule out confounders; option C uses an RCT designed precisely to establish causation; option D describes a correlation that does reflect a real causal mechanism.

Question 2

According to Hume, repeated observation of one event following another gives us direct perception of the necessary connection between them.

• T. True
• F. False

Answer: False

This is Hume's central negative claim: we never perceive necessary connection. We observe constant conjunction (A is always followed by B) and temporal priority (A comes before B), but no 'power' or 'necessitation' is observed in the events themselves. Our idea of necessary connection arises from a 'determination of the mind' — a habit of expectation — not from observation. This is the core of Hume's skeptical analysis of causation.

Question 3

Why is it philosophically difficult to distinguish causation from mere correlation?

Model answer: Correlation is a statistical regularity between two variables, but causation seems to require a real dependence or mechanism that explains why A brings about B. It is difficult to specify what that 'something more' is without using causal concepts, and the observable data alone underdetermines which causal structure produced it.

A correlation between A and B is compatible with A causing B, B causing A, a third variable C causing both, or pure coincidence. Identifying which is operating requires more than statistical data — it requires causal assumptions about what would happen under interventions, or knowledge of the underlying mechanism. This is why causal inference in both philosophy and statistics is technically and conceptually demanding.