Jones had streptococcal infection and took penicillin; the statistical law says penicillin cures strep with 90% probability; Jones recovered. An IS explanation is constructed. However, Jones also wore a red hat during treatment, and patients with strep who took penicillin and wore red hats also have a 90% recovery rate. What does this reveal about the IS model?
AThe IS explanation is strengthened because two factors contributed to Jones's recovery
BThe irrelevance problem: the IS model cannot exclude causally irrelevant factors, since any premise set yielding high probability qualifies
CThe IS explanation is weakened because the red hat introduces a confound
DThis shows the IS model is robust — the same probability in both reference classes confirms the explanation
The irrelevance problem: the IS model accepts any premise set that makes the explanandum highly probable, regardless of causal relevance. The red hat is causally irrelevant to strep recovery — including it neither raises nor lowers probability. But the IS model has no mechanism for excluding it; any high-probability argument counts as an explanation. This shows the IS model captures statistical association, not causal explanation. Salmon developed his causal-statistical account precisely to require that explanatory factors be causally, not just statistically, connected to the outcome.
Question 2 Multiple Choice
According to the IS model, which of the following constitutes a scientific explanation of an event?
AA deductively valid argument showing the event was the only possible outcome given the laws
BAn inductive argument from a statistical law and initial conditions that renders the event highly probable
CAn argument that identifies the causal mechanisms responsible for the event
DA narrative description of the sequence of events leading to the outcome
The IS model extends Hempel's covering-law framework to probabilistic cases. An IS explanation has the structure: statistical law L + initial conditions C → (with high probability) → event E. The 'high probability' requirement distinguishes it from the DN model's deductive entailment. Option A describes the deductive-nomological model. Option C describes Salmon's causal-statistical model, developed *in response to* the IS model's failures. Option D is a narrative, not an explanation in Hempel's sense.
Question 3 True / False
Under the IS model, a single event can simultaneously receive a high-probability IS explanation and a low-probability IS 'explanation,' depending on which reference class is chosen.
TTrue
FFalse
Answer: True
This is Salmon's problem of ambiguity. Under 'patients who received penicillin,' John's recovery has probability 0.9 — a successful IS explanation. Under 'patients who received penicillin and had a drug-resistant strain,' recovery has probability 0.1 — no IS explanation, and the event is surprising. Same event, same person, contradictory IS verdicts. The IS model provides no principled basis for choosing the 'correct' reference class, which Salmon regarded as a fatal flaw showing that statistical association alone cannot underwrite genuine explanation.
Question 4 True / False
The IS model improves on the DN model primarily by requiring that explanatory laws be universal and exceptionless.
TTrue
FFalse
Answer: False
The IS model improves on the DN model by *relaxing* the requirement for universal laws, allowing statistical laws to figure in explanations. The DN model already required universal, exceptionless laws. The IS model expands the framework by accepting probabilistic laws as legitimate explanatory premises, substituting high-probability inductive support for deductive entailment. The challenge for the IS model is not finding stricter laws but handling problems created by probabilistic laws: the vagueness of 'high' probability, the irrelevance problem, and the ambiguity problem.
Question 5 Short Answer
What is the problem of ambiguity in the IS model, and why did it lead Salmon to propose a causal-statistical account of explanation instead?
Think about your answer, then reveal below.
Model answer: The problem of ambiguity: the IS model's verdict on whether an event is explained depends on the reference class used to describe it, but the model provides no principled method for selecting the correct class. The same event can be rendered highly probable (a successful IS explanation) by one reference class and improbable (no IS explanation) by another. Salmon argued this shows IS explanations are not objective — they depend on how we describe the explanandum. His causal-statistical account responds by requiring that explanations cite actual causal mechanisms, making the causally homogeneous reference class the one that picks out all and only the causally relevant factors.
The deeper issue is that statistical regularities can be causally spurious. The IS model, because it only requires high probability, is blind to the difference between correlation and causation. Causal explanation requires identifying the mechanism by which C brought about E, not just showing that events like C are usually followed by events like E. Salmon's insight was that the reference class problem is not a technical glitch but a symptom of a fundamental inadequacy: statistical association is not the same as causal explanation.