The source of a scientific realist's insomnia can usually be traced to one of two arguments: the pessimistic induction from the history of science or the underdetermination of theories by empirical evidence. Kyle Stanford, however, argues in Exceeding Our Grasp that a more persuasive case can be made for relinquishing realist sympathies. Stanford's argument originates in his problem of unconceived alternatives: past generations of scientists have systematically failed to recognize alternative theories that were at least as well supported by the available evidence as those theories actually endorsed. An induction from such cases implies that current theories are also being endorsed, despite the availability of alternative, equally well confirmed, and as yet unconceived alternatives. A realist attitude towards contemporary theories thus appears unjustified.
Stanford's central argument is innovative and the book should be required reading for anyone concerned with the realism debate. Those chapters (principally 1 and 6) that examine recent work on the realism debate are particularly strong and would make an excellent resource for an undergraduate class seeking an overview of such arguments. Discussions of the developmental theories of Darwin, Galton, and Weismann are illuminating and a refreshing change from the antirealist's usual fascination with phlogiston, caloric and ether (although Stanford is not shy about also adding these latter cases to his stock of examples designed to promote his inductive argument). Some might quibble that these developmental theories were significantly more speculative than those offered by de Vries and Morgan at the turn of the twentieth century. The latter's more empirical approach to broadly developmental issues might open space for the realist to deny that the lessons Stanford draws from the older generation of theoreticians transfer straightforwardly to later work. But Stanford's general point must surely be conceded: past scientific commitments have been dogmatically championed, yet ultimately replaced by theories that wholly abandoned those very assumptions.
As Stanford explains, his argument from unconceived alternatives is a consequence of our apparent tendency to apply eliminative methods of reasoning in situations where they are not reliable. Hypotheses are often endorsed on the grounds that all competing hypotheses have been discredited and thus can be eliminated from further consideration. Stanford's historical analyses reveal that, at least insofar as we apply eliminate reasoning to fundamental accounts of nature, such inferences have proved worryingly fallible. The failure to conceive of viable alternative hypotheses is simply a poor reason for thinking there are no such alternatives. Stanford's attention to the limitations of inferential practices marks an important departure from alternative, prominent antirealist theses, notably van Fraassen's constructive empiricism. Stanford's thesis allows, for example, the possibility of acquiring knowledge of unobservable entities.
Admitting knowledge of unobservables might seem a large concession to the realist, but Stanford makes clear that his argument is directed against claims to the effect that “the central and fundamental claims of our scientific theories must be at least roughly accurate” (6). As an objection to realism, however, this strategy seems to me rather too narrow. Selective realist strategies distance themselves from recommending the approximate truth of entire theories, restricting realist attitudes to only certain parts of theories. Admittedly Stanford subjects prominent, selective realist theses to careful scrutiny and I think correctly finds such accounts wanting. However, I do not share his pessimism concerning the general strategy nor accept his suggestion that selective realist strategies must be “prospectively applicable” (168). Stanford argues convincingly against Kitcher that selective realists cannot use current theories, first, to assess which parts of past theories were responsible for success and, second, to determine which parts of past theories are true. But from the failure of such retrospective reconstructions it does not follow that the selective realist's strategy must be prospectively applicable. All the selective realist needs are distinct vantages for establishing whether parts of past theories were involved in success and whether those same parts were retained across all subsequent cases of theory change. The latter must surely rely on the perspective of our current theories; it is less clear that judgments concerning the source of success cannot be made from a distinctive viewpoint. For example, when scientists introduce a new idea, we might reasonably suppose that it is endorsed because it improves upon previous theories. From the perspective of the replaced theory we can surely evaluate which parts of the new theory make it more successful than the old. The retention of those parts across subsequent changes of theory would seem to satisfy the selective realists’ hopes. Insofar as Stanford's argument relies on a rejection of the selective strategy, I am unconvinced. A further reason for persisting with realist attitudes stems from Stanford's positive views on scientific inquiry.
In the final chapter Stanford discusses and defends an instrumentalist attitude towards scientific theories. The details, however, suggest the distinction between Stanford and many realists is better conceived not as a question about which objects are epistemically accessible but instead as an issue of relative optimism and pessimism. Both sides agree that knowledge of unobservable phenomena is possible. The realist supposes that such knowledge will expand and thus our understanding of even the most fundamental aspects of reality will grow. Stanford's instrumentalist thesis appears to leave room for such growth. Simultaneously, Stanford is deeply impressed by the historical record as evidence that our scientific theories might be subject to radical changes in the future. Those concerns, however, are consistent with the realist's belief in a pattern of increasingly truthlike theories. The realist has certainly not answered all challenges in making this case. Nevertheless, Stanford writes as if he is defending the most potent argument yet for abandoning scientific realism. I suspect in the end he will prove more successful at convincing realists to be cautious in just how far they extend their realist commitments.
