Why is the nomological deductive model accused of ‘eschewing any account of causality’ when the general laws on which it is based are (apparently) causal ones (‘when the thread is loaded with a weight exceeding that which characterizes the tensile strength of the thread, then it will break’)?
Answer by Craig Skinner
Good question. Gets us into deep waters quickly.
The nomological-deductive model (covering-law model, DN model) of explanation (Hempel and Oppenheim, 1948) is so ‘accused’ because the model was formulated so as to deliberately avoid invoking causality. Hempel was influenced by Hume’s account of causation as constant conjunction without any evident connexion. In Hume’s view, we think we see causation in the world but all we really see is one event (the cause) followed by another (the effect), repeatedly, so that we come to expect the second whenever we see the first; but we never see a connexion between the two; or see how the first necessitates the second. So, in the absence of a good account of causation or causal necessity, Hempel preferred to stick with natural laws, construed as exceptionless regularities, which, along with the particular circumstances, allowed deduction of an outcome, thereby explaining it.
A problem immediately arises. The DN model is insensitive to the direction of explanation. Classical example: we explain the length of the pole’s shadow by reference to the height of the pole (plus the angle of the sun, and the law that light moves in straight lines). But we could equally well explain the pole’s height by reference to the shadow’s length. But of course this latter is no explanation of why a pole is, say, 20 feet tall (for that, we need to look into the habits and intentions of pole makers and buyers, and uses of poles). Ah yes, you say, but the shadow’s length is caused by the height of the pole, whereas the pole’s height is not caused by the shadow’s length. True, but now you have invoked causality, which is disallowed.
As well as this directional symmetry problem, the DN model suffers from:
* the explanatory irrelevance problem. Example: people taking contraceptive pills dont get pregnant; Mr Jones takes contraceptive pills; so Mr Jones doesn’t get pregnant.
* inability to deal with generalizations which are not exceptionless and/or cant plausibly be taken as natural laws (eg smoking causes lung cancer).
* inability to deal with everyday occurrences. Example: a caveman dropping his mammoth bone, shattering his wife’s prized shell, instantly understands her explanation that her ornament is ruined because of him and his carelessness, although neither has the notion of a natural law, far less of the DN model, nor could they or anybody else fit the cavewoman’s explanation into the model, without positing ‘hidden structure’ or ‘implied laws and circumstances’ somehow contained within this everyday explanation.
In view of the problems with the DN model, causal explanatory models were duly formulated. Notably Salmon’s causal-mechanical (mark transmission) model, Lewis’s counterfactual model, Woodward’s manipulative model. All have problems and limitations.
I incline to the manipulative model (A causes B if altering A, leaving all else unchanged, alters B).
First, we dont need the prior notion of a law.
Secondly, it’s in line with scientific and medical practice and research.
Thus, to know if drug A causes blood pressure (BP) to fall, the effect on BP of taking A is compared with the effect of taking a placebo in a patient group (patients and researchers ‘blinded’ as to who’s on what, and order of treatments randomized within the group, so as to factor out confounding variables, leaving only A and not-A as the comparison). We find BP falls when people take A, but not on placebo. Ergo, A causes a fall in BP.
Another example. ‘Smoking causes lung cancer’ is impossible for the D-N model to handle, difficult for some other models, smoking being neither necessary nor sufficient as a cause of lung cancer (most smokers dont get cancer, some people with lung cancer have never smoked). But the manipulationist view says that if we can reduce smoking in the population then lung cancer rates will change. And indeed, reduced smoking has been followed by falling lung cancer rates in Western Europe (but increased smoking has led to soaring lung cancer rates in developing countries)
Thirdly, manipulation is one way in which we acquire the notion of causation. The baby learns that her hand movement caused his drink to spill, the child learns that his shove caused the bully to fall to the ground. So that, whereas Hume (correctly) held that we get the idea of causation by observing succession, he omitted to say that we also learn it by experiencing success.
Note that the ‘manipulation’ doesn’t have to be due to human action or intention. It can occur naturally. Thus a big asteroid impact ‘manipulated’ the earth’s climate 600 million years ago causing the accelerated extinction of the dinosaurs.
Whether we ‘explain’ laws with reference to causality (as you suggest) or ‘explain’ causality with reference to laws, we try to explain the obscure with the equally obscure. So, Hume reduces causation to exceptionless regularity, Hempel adopts this as his notion of a law only to find that he cant come up with criteria which are necessary and sufficient to distinguish a law from an accidental generalization (this latter problem is still with us). On the other hand, we have no agreed account of causality, and often find that in answering a causal question (what causes an apple to fall) we fall back on invoking laws (gravity, spatial curvature).
Understanding causality is very much a work in progress.