Over the past 50 years, Peter Achinstein has earned a reputation in the philosophy of science for careful, thought-provoking, and methodologically significant conceptual analyses. This new book matches that reputation: he begins with a detailed analysis a particular notion, speculation, and the clarified concept serves as the glue to hold together the book’s wide-ranging discussions. Achinstein also investigates issues of interest for practicing scientists and the ideas of practicing scientists appear frequently in the book. Indeed, two of the most influential philosophers who discussed speculation were also scientists: John Stuart Mill (who was an economist and psychologist as well as a philosopher) and Isaac Newton. Within a crowded literature on the topic, Achinstein offers a book that is novel, rigorous, and helpful for both philosophers and scientists.

          Speculation, as Achinstein examines it, means to introduce assumptions into scientific reasoning aimed towards the truth, without adequate evidence for believing these assumptions. Speculations are normally introduced as part of explaining, unifying, predicting, or calculating. The speculator might regard them as true, approximately true, or just a possibility worthy of consideration. They are distinct from speculative fictions (like imagining a universe in which ideal gases were physically possible) and from assumptions that serve a purely instrumental function, such as generating an unrealistic but predictively powerful forecasting model. Although speculations might be made in either public or private reasoning, Achinstein focuses on the public speculations, because the literature is best understood (with a few exceptions) as chiefly concerning the permissibility of scientists speculating in articles, books, public engagement, expert testimony, and so on.

          Achinstein contrasts his own view with three broad alternatives in the literature: (1) “Do not speculate!” (2) “Speculate, but test!” and (3) “Speculate even if you cannot test!” The first was most influentially proposed by Newton: he thought that scientists should remain silent when there was inadequate evidence. The second is most strongly associated with Karl Popper: he thought that to be boldly speculative was a virtue in a scientist, provided that she subsequently made a serious effort to test her speculative theories. Finally, the third view was most (in)famously advanced by Paul Feyerabend: he thought that scientists should advance and advocate as many alternative hypotheses as possible, even if they were incompatible with the apparent evidence.     I thought that Achinstein’s characterization of Feyerabend’s views was too quick The latter would argue that developing a range of alternative explanations of a phenomenon and contrasting these alternatives against our existing theories is a way of testing those alternative explanations. Feyerabend would also argue that, historically, science has primarily progressed through this “proliferation”. In fact, he believed that no other methodological strategy was so well-evinced by the history of science. Thus, he would agree with (2), in a broad sense of ‘test’. However, Achinstein is essentially right to contrast Feyerabend’s views on speculation with those of Popper. The most important point is that Feyerabend denied Popper’s claim that severe empirical testing of these alternative explanations was a universal methodological requirement for legitimate speculations.

          Achinstein enters into this debate in Chapter 1. He argues that Newton’s view condemns too many examples of successful scientific practice, including some of Newton’s own best reasoning. Against Popper, he argues that the demand to “test!” is vague, since the timing and value (relative to other uses of resources) of the necessary tests is unclear. How long can we wait until testing the theory? What about speculations, such as atomic theory, that take many centuries before flourishing to bear empirical fruits? Meanwhile, Feyerabend’s view fails to consider how speculation can be counterproductive for scientific progress. Achinstein argues against all three views that contextual and pragmatic factors determine the rationality of a particular scientist’s speculation. According to Achinstein, these contextual pragmatic factors will depend on the speculator’s knowledge and interests.

          Feyerabend and many other advocates of (3) would agree that scientists can speculate irresponsibly. The place of science within society was a great concern for Feyerabend, particularly in his later years. For example, speculations about evolutionary psychology can bring particular claims about sex and personality to public salience without much evidence behind them. Arguably, the social harms that can be caused by this salience outweigh the epistemic value of a scientist’s public speculations about them. However, Achinstein’s principal originality is that he notes that there can be epistemic reasons to speculate or not to speculate (p. 63). I would have liked to read more of his thoughts on these epistemic reasons not to speculate, since the idea has some plausibility but nonetheless is somewhat underdeveloped in this book. A connection with decision theory and the social epistemological literature on the responsibilities might be helpful. These are both booming areas within contemporary philosophy of science and they naturally lend themselves to developing some of Achinstein’s ideas in this book. Fleshing out Achinstein’s rather brief remarks on the constraints on speculation would certainly be an excellent doctoral or postdoctoral project.

          Recommending or forbidding speculations is only meaningful insofar as we have a good grasp of the distinction between speculations and other sorts of assumptions. Since the concept of evidence (and its absence) is crucial to distinguishing speculation from non-speculation, much of Chapter 1 is about different analyses of evidence. The presentation of Achinstein’s various definitions of different evidential concepts (and Bayesian epistemology) is quick and spread piecemeal over the course of a chapter that is also discussing the methodology of speculation. Consequently, the discussion of evidence in Chapter 1 is not suitable for those unfamiliar with the relevant theories of evidence. In particular, Achinstein’s presentation of his own analyses leaves many elements undermotivated: someone unfamiliar with his work would not appreciate his reasons for focusing on explanatory relations between a hypothesis and its relevant evidence, as opposed to Bayesianism’s focus on their probabilistic relations. I recommend that readers without such background knowledge make use of Achinstein’s own The Book of Evidence (2001) and also read the preface and first chapter of Rudolf Carnap’s The Logical Foundations of Probability (1962). It would also help to at least peruse some of Colin Howson and Peter Urbach’s Scientific Reasoning: The Bayesian Approach (2006) for an excellent example of the Bayesian analysis of evidence. Having discussed speculation in general, Achinstein considers a variety of issues in the philosophy of science and science itself, with a unifying theme of asking whether some major theories in these disciplines are speculative. Chapters 2 and 3 tackle the difficult issue of simplicity. Achinstein argues that many philosophical theses about simplicity in science are speculative; claims like “We know that nature is simple” and “Science presupposes that nature is simple” have been asserted without sufficient evidential warrant. Additionally, he outlines a strategic theory of simplicity’s role in science. According to Achinstein, Newton’s influential rules for inductive reasoning (which favour simple hypotheses) should be interpreted as advocating strategies for debating causal hypotheses, rather than as a take-it-or-leave-it endorsement of a particular theory of nature. While favouring simpler theories is fallible, but Achinstein sees this preference as a potentially useful rule-of-thumb. Newton thought that the justification of this strategic role for simplicity was that we know that nature is simple. However, alternative justifications are possible. As Achinstein says, one could also justify simplicity’s role in the Newtonian rules on the grounds of non-epistemic criteria, such as cognitive economy and ease of testing. If simple causal hypotheses are easier to use in our reasoning or easier to test, then it might make sense to focus on debating them instead of more complex alternatives.

          In Chapter 4, Achinstein examines a long-running issue in the philosophy of science: the particularism vs. holism debate. According to “particularist” philosophers of science like Mill, scientific theories are tested one theory at a time. According to “holist” philosophers of science like Pierre Duhem and W. V. O. Quine, scientific theories can only be tested together as a group. Achinstein argues that both particularism and holism are speculative. Yet it was not always clear to me whether he is arguing that the two approaches are merely speculative or outright mistaken. For instance, he spends a significant portion of the chapter arguing that particularists often oversimplify the evidential relations between theories and their tests, but this point is distinct from whether particularism as such is a speculative theory. However, the chapter as a whole is extremely good.

          I agreed with Achinstein that holists make overgeneralisations about scientific evidence, but this claim raises the question of how holism has become so influential among so many good philosophers. My own impression is that holists have tended to sit with legs on both sides on a fence with respect to whether their principal claim is descriptive (“Scientists generally reason as though theories can only be tested a group”) or normative (“Scientists should reason as though theories can only be tested a group”). When pressure is put on the normative side by some philosopher of science, holists have leant on the descriptive leg via the sweeping descriptive generalizations that Achinstein rightly characterizes as speculative. When pressure is put on the descriptive side by some historians of science, holists have leant on the normative leg via a priori arguments, like those of Quine. I also was persuaded by Achinstein that at least some particularists, like Mill, have used oversimplified analyses of evidence. As with so many of Achinstein’s arguments, his reasoning here can be appreciated even if one disagrees with his own theory of evidence, which he applies in much of the chapter.

          In the final chapter, Achinstein discusses “TOEs”, theories of everything. Despite the scope of the topic, in just a single chapter he is able to contribute meaningfully to the vast literature on such theories. A TOE would explain all phenomena in terms of a set of irreducible entities (such as strings) and a set of irreducible laws (such as a unifying system of laws for both relativity and quantum fields). Achinstein examines string theory, in addition to some inchoate proposals in the general direction of a TOE by Thomas Nagel and David Chalmers. Achinstein argues that all of the TOEs proposed thus far are speculations. More surprisingly, he makes the case that even the desirability of a TOE has not been established. Achinstein also persuasively contends that a TOE’s value will depend on factors specific to that theory, not just the very broad arguments that have hitherto been offered for the desirability of a TOE. This is because a TOE might have very little explanatory utility for humans, due to our cognitive limitations. In principle, using the particular TOE to explain all fundamental phenomena might be possible, but in practice it might be beyond our capabilities. Achinstein’s reasoning in this chapter is extremely sharp and wide-ranging. I was particularly interested by his arguments that the history of science does not support any sort of trend towards a TOE. At times the discussion is a little vague, but this is due to the imprecision of some Achinstein’s interlocutors, rather than his own fault. It is also important to note that, while Achinstein believes that the existing TOEs are speculative, it does not follow that they are intrinsically defective.
          Throughout the book, his style is dense but clear. He makes ample use of the tools of contemporary philosophy of evidence and philosophical analysis, but the book is also well-informed by his extensive knowledge of contemporary physics and the history of science. Very few philosophers of science have Achinstein’s intimate familiarity with the ideas of William Whewell, James Clark Maxwell, and Newton. Consequently, one of the book’s strong points is that it provides an exciting opening into the works of these thinkers.

          Achinstein has once again given us a carefully written and thought-provoking book. Philosophers of science are the most natural audience, but the methodologically curious scientist will find much of value in Achinstein’s discussions of speculation, evidence, explanation, underdetermination, and other topics. Given some background reading on theories of evidence (particularly Bayesianism and Achinstein’s own theory) the book would be accessible to undergraduate philosophy students, as well as scientists with little prior reading in philosophy. This book is also an excellent example of how conceptual analysis can be useful for tackling methodological issues that make a substantial difference to what scientists do.

Bibliography

Achinstein, Peter (2001) The Book of Evidence Oxford: Oxford University Press.

Carnap, Rudolf (1962) The Logical Foundations of Probability Chicago: University of Chicago Press.

Howson, Colin and Urbach, Peter (2006) Scientific Reasoning: The Bayesian Approach Chicago and La Salle, Illinois: Open Court.

© William Peden 20190

William Peden, Department of Philosophy, Durham University