Paul Dicken

Recent literature in the scientific realism debate has been concerned with a particular species of statistical fallacy concerning base-rates, and the worry that no matter how predictively successful our contemporary scientific theories may be, this will tell us absolutely nothing about the likelihood of their truth if our overall sample space contains enough empirically adequate theories that are nevertheless false. In response, both realists and anti-realists have switched their focus from general arguments concerning the reliability and historical track-records of our scientific methodology, to a series of specific arguments and case-studies concerning our reasons to believe individual scientific theories to be true. Such a development however sits in tension with the usual understanding of the scientific realism debate as offering a second-order assessment of our first-order scientific practices, and threatens to undermine the possibility of a distinctive philosophical debate over the approximate truth of our scientific theories. I illustrate this concern with three recent attempts to offer a more localised understanding of the scientific realism debate—due to Stathis Psillos, Juha Saatsi, and Kyle Stanford—and argue that none of these alternatives offer a satisfactory response to the problem.

There is a natural objection to the epistemic coherence of Bas van Fraassen’s use of a distinction between the observable and unobservable in his constructive empiricism, an objection that has been raised with particular clarity by Alan Musgrave. We outline Musgrave’s objection, and then consider how one might interpret and evaluate van Fraassen’s response. According to the constructive empiricist, observability for us is measured with respect to the epistemic limits of human beings qua measuring devices, limitations ‘which will be described in detail in the final physics and biology’ (van Fraassen 1980: 17). In order for the constructive empiricist to determine what counts as observable, he will have to appeal to our best scientific theories of light, human physiology, and so forth. To put the same point in a slightly more abstract way, in order to draw a distinction between observable and unobservable entities, the constructive empiricist needs to use his best scientific theory of observability – call it T* – to tell him the identity of the observable entities. This raises an interesting difficulty. Constructive empiricism is the view that ‘science aims to give us theories that are empirically adequate; and acceptance of a theory involves as belief only that it is empirically adequate’ (van Fraassen 1980:12). When he accepts a theory, the constructive empiricist only believes the statements of his scientific theories that are about observable entities. Thus, in order to know which statements of a scientific theory to believe, the constructive empiricist needs to know which statements of that theory are about observable entities. In particular then, the constructive empiricist only believes the statements of his theory of observability T* that are about observable entities. Therefore, in order to know which statements of T* he can believe, the constructive empiricist needs to know which statements of T* are about observable entities. However, it is T* that tells the constructive empiricist what counts as an observable entity: the constructive empiricist therefore needs to use T* to tell him which statements of T* he can believe. The fact that the distinction drawn by T* must also apply to itself is not an immediate cause for alarm..

In this paper, I explore Rosen’s ‘transcendental’ objection to constructive empiricism—the argument that in order to be a constructive empiricist, one must be ontologically committed to just the sort of abstract, mathematical objects constructive empiricism seems committed to denying. In particular, I assess Bueno’s ‘partial structures’ response to Rosen, and argue that such a strategy cannot succeed, on the grounds that it cannot provide an adequate metalogic for our scientific discourse. I conclude by arguing that this result provides some interesting consequences in general for anti-realist programmes in the philosophy of science.Keywords: Constructive empiricism; Mathematical fictionalism; Partial structures; Modality

As part of his wider critique of the credibility of miraculous testimony, Hume also offers a rather curious argument as to the mutual detriment of conflicting testimony for the miracles of contrary religious worldviews. Scholarship on this aspect of Hume’s reasoning has debated whether or not the considerations are to be understood as essentially probabilistic, and as to whether or not a probabilistic interpretation of the argument is logically valid. The consensus would appear to offer a positive answer to the first question and a negative answer to the second. In this paper I expose a deeper fallacy in Hume’s reasoning that undermines both probabilistic and non-probabilistic readings. My critique is closely based upon analogous considerations in the philosophy of science, and the equally intriguing issue as to the epistemological relevance of conflicting scientific theories throughout the history of science

Carnap's mature philosophy of science is an attempt to dissolve the scientific realism debate altogether as a philosophical pseudo-question. His argument depends upon a logico-semantic thesis regarding the structure of a scientific theory, and more importantly, a meta-ontological thesis regarding the explication of existence claims. The latter commits Carnap to a distinction between the analytic and the synthetic, which was allegedly refuted by Quine. The contemporary philosophy of science has therefore sought to distance itself from logico-semantic considerations, and has pursued the scientific realism debate as an essentially epistemological thesis. I show however that one of the most prominent positions in this recent debate—van Fraassen's constructive empiricism—not only ends up in very close proximity to Carnap's attempted dissolution, but even provides the resources for extending and refining his programme. Rather than a historical footnote, Carnap's mature philosophy of science offers a live-option in the current debate.

In this journal, Peter Lipton and I discussed Musgrave's objection that the constructive empiricist cannot consistently maintain his own distinction between the observable and the unobservable, and van Fraassen's initial reply. We considered several possible interpretations of van Fraassen, and expressed misgivings about each. Muller and van Fraassen have consequently clarified the official constructive empiricist response to Musgrave, although some issues still remain.According to Muller and van Fraassen, Musgrave's objection assumes that constructive empiricism is to be understood in line with a syntactic account of our scientific theories – whereby a theory consists of a set of sentences formulated in some suitable language – whereas constructive empiricism has always been wedded to an explicitly semantic account of our scientific theories, which understands a scientific theory in terms of a class of models. Once this oversight is amended, the objection is seen to be toothless. As Muller and van Fraassen write: " Indeed, the only proper response, and the one that turns the table on Musgrave and his followers, is to argue that his incoherence argument does not work in the context of the semantic view as here elaborated, while the bad consequence he draws is an inevitable corollary to the older view within which he presents it. "The force of Musgrave's objection therefore turns on nothing more than implicitly saddling the constructive empiricist with an independently problematic account of what a scientific theory is.This move should cause some surprise however, for Musgrave's objection – albeit usually presented in conjunction with an explicitly syntactic construal of our scientific theories – is in fact quite neutral between these competing accounts. As Musgrave originally put it, in order for the constructive empiricist to draw his distinction between observable and unobservable …