Ronald Giere's Science without Laws is a collection of essays that ranges over a wide variety of topics, including the role of visual models in science, the influence of gender on the content of scientific theories, the history of twentieth-century philosophy of science, science and technology studies, the nature of theory choice, and questions surrounding scientific realism. Many of the basic themes will be familiar from Giere's previous work. Giere advocates a “naturalistic”, “evolutionary” model of science, with the analogues to variation, selection, and transmission in biological evolution supplied by an account of the mechanisms underlying the cognitive capacities of individual scientists (here Giere looks to cognitive science) and an account of the “social mechanisms operative in the social and institutional environments in which scientists work and interact” (here Giere looks to “the social sciences, particularly the sociology of science”) (49). Like other advocates of the semantic view of theories, Giere regards scientific theories as families of models, but he adds to this the idea that theories of concept structure drawn from cognitive psychology have illuminating things to tell us about the structure of such models. For example, he suggests that concepts in classical mechanics exhibit a “graded” structure, with “central” (e.g., simple pendulums) and “peripheral” cases (e.g., coupled and damped pendulums) (106ff).

For Giere, a theoretical model such as a model of a simple harmonic oscillator is in itself neither true nor false. We make a statement which is true or false when we introduce what Giere calls a “theoretical hypothesis” which claims that some “designated real system is similar to the proposed model in specified respects and to specified degrees” (179, italics in original). Giere variously describes his view as “constructive” or (in more recent essays) as “perspectival” realism or, alternatively, as “realism without truth.” As I understand it, what he means by this is that no model will exactly resemble the real system it purports to describe in every respect. To construe a model realistically we need only the idea that it is an adequate representation in some specified respect of the system being modeled. The notion of such a representation is a more general idea than the notion of a statement's being “true,” with representation having to do with a qualitative notion of “fit” between model and world—a notion that admits of degrees. The notion of “fit” is explained by means of an analogy with maps. A map is not the sort of thing that is literally true or false but nonetheless a map can be a more or less good or useful representation for various purposes and hence can “fit” what it represents more or less well.

Of course even antirealists (as this position is usually understood) will agree that we are sometimes warranted in believing that there is some similarity between models and what is modeled. For example, constructive empiricists in the sense of Reference van Fraassenvan Fraassen (1980) hold that our models should aim at fitting (and at least sometimes succeed in fitting) what is observed (in the sense of perceived by humans)—thus that some empirical substructure of our models should be “similar” to what is observed. Realism as usually understood holds that at least in the case of some models what we are warranted in believing goes beyond the claim that the model is observationally adequate. Giere agrees. He contends, very plausibly, that what is important in science is not the notion of what is perceivable by human beings but rather the broader notion of what is detectable or measurable. Presumably, this means that among the respects of similarity of model to world that matter is some notion of approximate fit to what can be measured, rather than merely to what is observed. Giere is also a realist about physical modality. He holds that the respects of similarity that matter include fit with the physically possible histories and not just the actual history of the system being modeled: as he puts it at one point, the model should tell us (at least to some extent and to some degree of approximation) “what the system would do if one were to change the initial conditions in a specified way” (p.185). When a model does this it tells us something about the causal structure of the system being modeled.

Although I'm sympathetic to these ideas and find them quite suggestive, I'm sure that Giere would agree that it would be desirable to say more about how claims of similarity between the more abstract or theoretical parts of models and the world are to be understood and evaluated. Consider a gravitational theory with a fixed flat background space-time and a separate gravitational potential and compare this with a space-time theory like General Relativity (GR) according to which gravitation is built into the curvature tensor. Are both of these models equally “similar” to the phenomena they model? Does similarity in this context mean anything more than “fits all possible (or perhaps just some range of) measurements up to some specified degree of approximation”? If not, why do we need the notion of similarity? If so, what does similarity add?

In correspondence Giere has suggested that “similarity” should be understood to mean more than fit with all possible measurements; among other things, it includes the relationship between features of the model and features of the world that are not directly measurable, at least if those features are inferable from what we can measure. For example, even if we think that space-time curvature is not directly measurable, we should think of GR models as claiming a similarity between the structure of space-time and the curvature specified in those models. However, when a quantity in a model can take different values that do not correspond to any difference we can measure, we should not interpret that quantity realistically. Since in the case of models with a flat space-time background and a separate gravitational potential, neither quantity is separately measurable, we should not interpret either realistically. The real structure of space-time is more similar to the structure postulated in GR models than to the structure postulated by models with a flat space-time background.

Although this is an attractive picture, which will resonate with many philosophical readers, I think that it is in some tension with the pragmatic, theory-as-map idea that Giere wishes to defend. An alternative view, advocated by some general relativists (e.g., Kip Reference ThorneThorne 1994), which takes the map analogy more literally, is that the curved and flat space-time models do not really represent alternative ways the world might be. Which kind of model is more appropriate in a given context will depend upon the purposes for which it is to be used—and in particular on such nonrealist virtues as computational tractability. On this view, there is no question of one kind of model being more similar to the real structure of space-time than the other kind.

As Giere himself remarks at one point (93), if one takes the analogy between theories and maps seriously, it seems to follow that one should not think of “fit” as a two place relationship between model and world but rather as a three place relationship between model, world, and the purposes and capacities of the model user. While the usefulness of a map will depend in part on the relationship between the map and what it purports to represent, whether various sorts of inaccuracies or omissions will undermine the usefulness of a map will depend also very much on the uses to which it is put as well as facts about the interests, information processing abilities, and other capacities of its user. A 1:1 scale map the size of London, although in one sense highly accurate, will not be useful for getting around the city. The familiar map of the London underground is very useful for the purpose for which it is designed, in part because it provides information relevant to travel by underground in such an accessible and transparent format, even though (perhaps because) it does not provide an accurate representation of distance relationships between stops or other information one might like for surface navigation. If theories or models are like maps we may similarly need to acknowledge the possibility that different theories will be more or less useful for different purposes or problems, that usefulness for certain purposes may require inaccuracies and omissions in other respects (so that we should not expect there to be some single “map of everything”), and that the forms that useful theories take will not just be dictated by the “world” but will rather also reflect facts about us and our cognitive/technological capacities and interests. I don't take these observations to be criticisms of Giere's proposal, but rather to reflect the point that its implications are rather subversive of many of the traditional aspirations of realism.

Although Giere is a realist about causal structure, he is a skeptic or deflationist about laws of nature—hence the title of this collection. This is not to say that Giere finds no room at all for laws, properly understood, in his account of science. Laws serve as “general rules for the construction of models” (6) and also serve to characterize particular models, as, for example, a model of a Newtonian Gravitational system. However, in neither role do they function as true statements about the world. As I see it, there are at least two different strands to this portion of Giere's discussion. In part, his target seems to be a certain philosophical conception of what laws are—in particular, the familiar “Humean” picture of laws as universally quantified conditional claims that describe exceptionless regularities. Needless to say, rejection of this particular account of laws is compatible with believing that there is some other, more acceptable account of the notion, according to which it plays an important role in understanding some but perhaps not all areas of science (e.g., physics but not psychology). A second strand in Giere's discussion suggests a more radical skepticism about the notion of law—it is not just that philosophers have misdescribed a notion that figures importantly in some areas of science. Instead, the notion itself (or at least any reconstruction that retains its central features) is fundamentally flawed and ought to be discarded. It is this second line of thought that is suggested by Giere's “genealogy” of the concept of law, which traces its origins to the theological commitments of writers like Descartes and Newton—commitments that are indefensible in any naturalistic account of science. We also see this more radical skepticism in remarks like the following:

One of Giere's arguments for this deflationary treatment is that many paradigmatic candidates for laws such as the gravitational inverse square law are false whenever other forces or causal factors are operative, which is always or virtually always. If this argument were correct, it would seem to follow that such generalizations play at best a peripheral role both in science itself and our understanding of it. But the argument seems to depend on construing the gravitational law as a statement of a regularity connecting the total force experienced by some mass m ₁, to the quantities on the right hand side of that law, which include a second mass m ₂, rather than as a statement describing the component force experienced by m ₁ due to the gravitational attraction of m ₂. Although there are indeed circumstances—e.g., sufficiently strong gravitational fields or velocities that are large in comparison with the speed of light—in which the Newtonian gravitational law breaks down, no physicist that I know follows Giere in holding that the mere presence of a non-gravitational force falsifies the law. Instead the law is given the “component force” construal described above and this allows it to be applicable to a wide range of situations in which non-gravitational forces are present. Given the evident artificiality and implausibility of the conclusion to which the total force construal (and more broadly, the Humean conception of law with which it is associated) leads, why not take this to be a reason for replacing the Humean conception with a conception of law that better fits scientific practice rather than for aiming for an account of “science without laws”? More generally, it seems to me that one can agree with Giere both about the inadequacy of standard philosophical accounts of lawhood and the wrong-headedness of certain philosophical views about what laws do (e.g., that they are somehow required to serve as “truth makers” for causal claims) without concluding that the notion of law plays no role at all in science.

Giere suggests that the notion of law persists in our contemporary understanding of science simply through a kind of cultural inertia. But I take it that Giere would also agree that when, say, a researcher attempts to construct a Newtonian model of the motion of a collection of particles within a gravitational field, there are important differences among the claims that will figure in the model. “Initial conditions” like the position and velocity of each particle can take any one of a large number of different values and for each of these the gravitational force law as well as Newton's other laws will continue to hold, but there is no similar freedom to vary the exponent in the inverse square law. Why can't observations like this be used to construct a nontheological account of laws: laws are just those generalizations that are invariant (in the sense that they are stable or would continue to hold) across some sufficiently large and otherwise “suitable” range of changes in conditions and circumstances (“initial conditions”). Such an account ought to be acceptable to a modal realist like Giere. It has the advantage of connecting up with contemporary physicists’ understanding of laws (which relates laws to generalizations that obey certain symmetry principles and are invariant under certain transformations—see, e.g., Reference FeynmanFeynman (1965) and Reference WignerWigner (1967)). It also allows us to avoid some of what Giere rightly objects to in standard philosophical treatments of laws such as the commitment to exceptionlessness and universality, since a generalization can be invariant under some range of changes in initial conditions without being invariant under all such changes, as the example of the gravitational inverse square law illustrates.

As this short summary perhaps suggests, there is much in this book to interest, stimulate, and provoke both philosophers of science and the science studies community.