Thematic Affinities and Psychoanalysis

Dr. Lacewing’s paper is a very interesting one. We agree in part, but only in part. Lacewing (2012) rejects the general thesis that “causal inferences must always be justified on the basis of Mill’s canons” (p. 199). I agree, but so does his target, Adolf Grünbaum, as we shall see in a moment.

But first there is a question about Grünbaum’s alleged reliance on Mill’s Methods of Agreement and Difference. This interpretation may not make a difference to Lacewing’s arguments, but it is worth correcting, given that some philosophers criticize both methods and yet their criticisms have no direct bearing on any of Grünbaum’s arguments. When Grünbaum speaks of “Millian methods,” he is not necessarily talking about either the method of agreement or the method of difference. He means “modern” methods of controlled (or experimental) inquiry that are based on Mill’s four methods (1993, 163). The modern methods, such as the use of randomized clinical trials, were inspired by Mill’s methods, but neither their utility nor reliability requires that Mill’s methods be problem free.

Lacewing need not talk of Mill’s methods to make his points. Instead, he can focus on the principle he quotes from Grünbaum (2004, 146; see also, Salmon 1984, 185–6). Roughly, the principle says that if X is causally relevant to the occurrence of Y, then the incidence of Ys in the class of Xs and Ys will be different compared with the incidence of Ys in the class of non-Xs and Ys.

Contrary to what Lacewing says, this is not a method for inferring causes nor is it an epistemic principle. It is rather a statement of the meaning of “X is causally relevant to the occurrence of Y” (Grünbaum 2004, 146). Alternatively, one could treat it as an ontological principle about what constitutes causal relevance (Salmon 1984, 185–6).

I, unlike Grünbaum or Salmon, do try to give an epistemic interpretation of the principle but I note that there are problems requiring its modification (Erwin 1996, 77–84; 2009, 509–602). One obvious problem concerns cases of singular causation. We can sometimes determine the cause of just one behavior without getting statistical evidence, as in my case of the autistic child (Erwin 2010, 11–2).

A second problem concerns novel causation. Salmon gives the example of inferring that a baseball thrown with a certain velocity will shatter a window and says that we can point to many cases where a window shattered after being struck by a baseball. This is true, but most of us would draw the same inference about a watermelon despite knowing of no case of window breaking involving a watermelon.

Such cases are not counter examples to the Salmon-Grünbaum principle taken as a semantic or ontological principle (although cases involving “offsetting or countervailing” causal factors [Erwin 1996, 82–3] may require a modification of the principle). If throwing a watermelon at a window pane is causally relevant to its breaking, then it will make a difference to window breaking in other cases where the causally relevant conditions are the same. However, we do not need statistical evidence to warrant the causal conclusion in the first case we encounter. Nor do we need statistical evidence in the case of the autistic child. For this reason, it is false to say that we always need statistical or experimental evidence to warrant causal inferences. But Grünbaum agrees.

I raise the same issue as Lacewing in Erwin (1993, 433): “One question prompted by Hopkin’s account is this: Does Grünbaum (1984) really require that all causal hypotheses be confirmed by controlled experimentation? That does appear to be too high a standard.” I then give the autistic child example. Grünbaum (1984, 259–60) not only agrees, but provides his own counterexamples. These are also counterexamples to the claim that where we do not have experimental evidence, we at least need a statistical comparison.

Where we do not...