‘The real is always ahead of what we can imagine’, wrote the novelist Paul Auster. In this engaging and ambitious book, P. Kyle Stanford advances three major claims about how reality outpaces the imaginations of scientists. The first is that scientific communities have repeatedly failed to conceive of alternatives to the theories that they hold at any given time, although those alternatives may also be well supported by the available evidence and often become accepted at a later time (pp. 19, 29). Second, this historical pattern proves that people lack the imagination to exhaust the space of theoretical possibilities to explain a given set of phenomena, especially in the domain of remote and fundamental processes in the world such as the constitution of matter or the history of the Earth (pp. 29, 31, 45). Third, since our imaginations have failed us in the past, it is only to be expected that they are failing us in the present. Thus we are not justified in believing many of our current best scientific theories as scientific realists would want us to do (p. 31).

Is the history of science full of examples of scientists who supported a theory without considering alternative theoretical accounts of the phenomena? No doubt. Is there a pattern showing that scientific communities consistently fail to conceive of alternative theories to the ones they support? This is not so clear, because the answer depends to a great extent on how we delimit the relevant scientific ‘community’ for any particular problem. Stanford examines theories of inheritance and of development during the second half of the nineteenth century. He shows that Darwin postulated the existence of ‘gemmules’ that carry the traits that offspring inherit from parents, but did not conceive of a common-cause explanation that postulates a shared germinal source of hereditary material responsible for the parents’ traits as well as the offspring's traits. Darwin failed to consider this alternative even after his own cousin Francis Galton put it forward. In turn, in accepting – with Darwin – a maturational view of the material basis of heredity, Galton failed to understand that this material does not become the constituent parts of an organism, but only provides the information necessary for development. He failed to conceive of a ‘directive’ view of heredity, although August Weismann had made such a suggestion. Weismann's failures of imagination stranded him in some scientific dead ends of his own.

But Darwin, Galton, Weismann and Mendel were all contemporaries, and if we consider them as part of the same scientific community the picture changes considerably from failure to success. Stanford could object that we cannot talk about their views as the views of the ‘scientific community’ of their time since there is no theory of inheritance that they held in common. But then why take this part of the history of science, in which there was no consensus, as a test of whether we can believe our ‘best’ theories in science? Furthermore, scientific realists are not committed to believing the best theories available in all areas of science. The best theory in a given field at a specific moment in time may not be good enough.

Stanford's claim about the cognitive limitations of human beings also requires further elaboration and evidence. Even if scientists do not explore all possible explanations for a given problem, this would not prove that their minds are unable to do so. It could be that, given the system of rewards in science, scientists find more psychological, sociological, economic or practical rewards by pursuing a single theory. It is an open question whether this is the case. My point is that Stanford has not considered alternative explanations for his empirical data, since he maintains that we can trust only those eliminative inferences in which we have considered all alternative explanations. Thus, by his own standards, we are not warranted to believe his explanation about our inability to exhaust the realm of alternatives in theoretical explanations of natural phenomena.

Finally, is the problem of unconceived alternatives in science the Achilles heel for scientific realists, as Stanford claims? I do not believe so. Imagine two alternative science-fiction scenarios. In the first, A, scientists in the past considered only one hypothesis at any given time, but that one always turned out to be confirmed by later research. In the second, B, scientists always considered the realm of all possibilities, but often chose the wrong one. Surely we would have a stronger justification to be scientific realists if A rather than B were the case. The issue is not how many alternatives scientists consider, but how often and why they select the correct one.

For historians of science the implications of Stanford's position are a mixed bag. He pays close attention to the history of science and his historical reconstructions are thought-provoking. But, for Stanford, the history of science is a ladder that he can dispose of once he has reached his objective. If we accept Stanford's conclusion about the cognitive limitations of human beings, we do not need to investigate the history of science any further for the purposes of investigating realism. In addition, he presents a history written from the viewpoint of the present, in the sense of seeing from our present understanding the steps that led nowhere. This history is useful for philosophical purposes, but it does not help illuminate the development and acceptance of scientific explanations in their specific historical contexts.

Stanford brings a fresh and original perspective to a fundamental issue in the philosophy of science. Although I do not find his major theses convincing in their present form, his book is well worth reading. It is written in a clear and accessible manner, and will thus be useful for people with no background in philosophy as well as for those engaged in discussions about scientific realism. For now, though – and contra Auster – I remain in awe of the powers of our imaginations.