Realism ‘naturalized’?

Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) Muhammad Ali Khalidi Natural Categories and Human Kinds: Classification in the Natural and Social Sciences Hardcover: 264 pages Publisher: Cambridge University Press (July 15, 2013) Language: English ISBN-13: 9781107012745 Price: £55.00 (US$ 90.00) Realism ‘naturalized’? A review of Muhammad Ali Khalidi, Natural Categories and Human Kinds By Georg Theiner (Villanova University) When we use language to describe and categorize the world, are we ‘carving nature at its joints,’ or are we rather ‘carving joints into nature’ 1? Within the contemporary debate, scientific realists believe the former. They are committed to the view that there is a privileged ontological stratification of things into natural kinds (NKs) that is metaphysically independent of how we choose to represent them, and the taxonomies employed by our best scientific theories are on target in our quest to discover what those are. Because of its association with the stronger doctrine of metaphysical Realism (with a capital ‘R’), scientific realism has had a somewhat curious fate. Inside the analytic camp, prominent metaphysical Realists have argued that modern science vindicates the Aristotelean view that NKs have ‘real essences.’ For less Realistically minded philosophers inside the analytic camp, as well as the majority of those outside looking in, the more zeitgeistig position has been to embrace some form of anti-realism (e.g., social constructivism) which rejects the idea of a mind-independent (or representation-independent) reality. Through ‘guilt by association’ with anything having the ‘fixity’ of Aristotelean essences, anti-realists generally repudiate the notion of NKs. In his rich and empirically astute study of classification in the natural and social sciences, Muhammad Ali Khalidi, Professor of Philosophy at York University, presents a ‘naturalized’ account of NKs that aims to liberate scientific realism from its essentialist shackles. As the curious juxtaposition of terms in his book title (‘natural categories’ and ‘human kinds’) adumbrates, Khalidi intends to subvert the traditional opposition between the ‘natural’ and the ‘humanly constructed.’ His goal is to show that “there is often a close connection between the kinds that are present in the world and the categories that we invent to understand the world, and […] defend the position that some of the types into which humans are divided can also be 1 I borrow this delightful phrase from Lupyan (2005). Thanks to Christopher Drain and an anonymous referee for their valuable comments on an earlier version of this review. Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) considered natural kinds” (xi). He proposes an account of NKs that is ‘naturalist’ insofar as it is developed under the guidance of a ‘reflective equilibrium’ methodology (cf. Goodman 1954/1983), by making mutual adjustments between the empirical deliverances of science and general philosophical principles until a good ‘fit’ is achieved (pp. 3-4, 201-204). At the same time, it is firmly ‘realist’ (with a small ‘r’) insofar as NKs are said to correspond to all and only the categories that would appear in an ideally completed scientific edifice (pp. 42-43, 79). The bulk of Khalidi’s book is concerned with the exposition and justification of his biconditional thesis. In Chapter 1, Khalidi begins his assault on metaphysical Realism, understood here as the view that kinds (e.g., proton) are ontologically distinct entities over and above the particulars which belong to them, similar to property universals (e.g., having positive charge of 1.6 × 10-19 C). He raises the following conundrum for kind Realism (pp. 8-10). If kinds are held to be primitive, and categorically distinct from the collection of property universals ‘associated’ with it, the Realist owes us an account of the metaphysically necessary connection between these two types of entities. On the other hand, if kinds are identified with conjunctions of property universals, this leaves unexplained why only some such conjunctions (but not others) constitute natural kinds. It would also suggest, in turn, that NKs can be reduced to a smaller set of ‘perfectly natural’ (Lewis) properties. Khalidi does not offer a straightforward resolution of this tension in his book. He more or less takes it for granted that kinds cannot be identified with their actual members (pp. 10-11), yet opposes the identification of kinds with properties because of a certain “looseness of fit” (p. 207). Without prejudging this thorny issue, Khalidi may be right that the whole issue of ontological ‘placeholders’ does not conform to the pristine picture promulgated by ‘armchair’ metaphysics. Khalidi’s real ‘beef’ with metaphysical Realism, though, is over its liaison with essentialism. The Aristotelean idea that the possession of ‘real essences’ provides a mark of the ‘natural’ (and thus, the ‘real’) has been articulated in a number of ways, all of which Khalidi rejects as incoherent or problematic. I won’t go into the details of his criticisms here, some of which are familiar from the literature, but they certainly leave the essentialist with options. For example, hard-nosed Realists are wont to limit their essentialist view of NKs to a ‘sparse’ ontology of micro-level entities such as kinds of elementary particles (e.g., quarks, leptons), their properties, and a few fundamental relations. Since they readily concede that macro-level entities are ‘nothing but’ complex arrangements of those micro-entities, they won’t be easily moved by the supposedly ‘counterintuitive’ implication that there are no genuine causal relations among macro-entities. That said, without an independent account of the causal-explanatory role played by nonfundamental scientific categories, the revisionist metaphysics of ‘micro-physicalism’ flies in the face of scientific practice, and scores rather low on the conservatism scale when judged by the ‘reflective equilibrium’ method. Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) In Chapter 2, Khalidi lays the groundwork for his brand of scientific realism. He opens with an epistemic gambit, building on the criterion that NKs should be discoverable by science (rather than invented or stipulated). He takes the familiar process of scientific theory construction as the sign of genuine discovery, provided that it converges on ‘projectible’ categories that have inductive and explanatory value. To bolster his case for realism, Khalidi takes on board Mill’s idea that scientific categories divide things into groups whose members share “a large number of important properties” (p. 49), and that those properties should preferably “be those which are the causes of many other properties” (p. 52), while rejecting Mill’s postulate that there must be “inexhaustibly many” of them (p. 51). But what’s so special about the scientific method? People have relied on the use of folk categories (e.g., fish, air) to make sense of the world long before the advent of modern science, and – as Khalidi admits – those categories afford at least moderate amounts of projectibility (62f). In response, Khalidi puts his realist cards on the table, and endorses a strict demarcation between ‘epistemic’ and ‘non-epistemic’ purposes. Science is privileged because its classifications are introduced (primarily) for epistemic purposes, with the “aim to discover divisions that exist in nature” (p. 63). This seals the fate of folk categories. If they are likewise introduced to serve epistemic purposes, we should expect them to coincide with or be superseded by scientific categories; alternatively, their persistence can only mean that they are retained for strictly non-epistemic (e.g., instrumental or evaluative) purposes (p. 64). For Khalidi, science is the measure of all things. Midway through Chapter 2, Khalidi willfully exposes his epistemic gambit. Facing a version of the ‘Euthyphro’ problem (p. 79), he asks, rhetorically, whether the ‘discoverability by science’ criterion “puts the epistemic cart before the metaphysical horse” (p. 64). His answer is that “projectibility is the epistemic marker for the metaphysical relation of causality” (p. 80). Using Boyd’s ‘homeostatic property cluster’ theory of NKs as a stalking horse for his own causal account, Khalidi criticizes Boyd’s notion as too restrictive, and thus unable to run the gamut of causal-explanatory practices found in the sciences. Instead, he adopts a liberalized version of Craver’s (2009) ‘simple causal theory’ (p. 78), arguing that “there is no single causal template that fits all instances of natural kinds or related natural kinds to their associated properties” (p. 80). Below, I will say more about the details of Khalidi’s account which are spelled out in Chapters 3-5; but the upshot of his discussion is that “[a]ssociated with each natural kind there is often a network of causal properties arranged in a hierarchy of causal priority, in which the causal relationships are nonstrict and are subtly contingent on contextual factors” (p. 207). Before moving forward, let me note a worry concerning the nature of vagueness, which is never explicitly discussed in the book. In a section titled “fuzzy kinds” (Chapter 2.4), Khalidi rejects Mill’s condition that there has to be an “impassable barrier” (p. 65) between any two NKs, and describes two types of cases in which kinds can have fuzzy membership conditions (p. 66). The Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) first type arises when we have partly disjunctive (‘cluster’) kinds with a graded membership structure (e.g., organisms in the course of speciation events). The second type arises when kinds are associated with properties that vary continuously along some dimension (e.g., isomeric compounds). Khalidi argues that realism about kinds is compatible with their fuzziness, since “chasms are more prominent than ditches […] but no less real” (p. 68). This has the effect of rendering the naturalness of a kind a matter of degree (p. 81), proportional to the epistemic measure of projectibility. Khalidi suggests that we can represent fuzzy NKs on a graded scale on which focal members score very high, but on which NK-ness “trails off” (p. 170) towards more marginal members, down to borderline cases that lie somewhere in between neighboring kinds. But this leaves open the question of whether vagueness is ultimately matter of linguistic (or representational) indecision, epistemic uncertainty, or whether there really are such things as vague objects ‘in nature’ (cf. Evans 1978). If we follow Khalidi in resisting the identification of kinds with (sets of) individual members (p. 11), it would seem that fuzzy kinds exist as ontologically vague or indeterminate entities in the world. Perhaps that is something realists (unlike Realists?) simply need to accept. The aim of Chapter 3 is to clear the ground for an ‘abundant’ conception of NKs, by criticizing three influential arguments against the possibility of NKs uncovered by the special sciences (such as biology, geology, neuroscience, psychology, or sociology). Khalidi’s rebuttal is anchored in a discussion of fluid mechanics – a sub-branch of physics that nevertheless ought to count as a special science since it deals with macroscopic principles governing the flow and diffusion of liquids that have no counterparts at the microlevel. Those include the substrateneutral property of viscosity or the resistance to flow, which is “the result of a complex array of intrinsic and extrinsic factors” (p. 94), the kind Newtonian fluid (a fluid that has constant viscosity, for a given pressure and temperature, regardless of the magnitude of the force applied to it 2), and Fick’s first law of diffusion. Khalidi uses the reference to fluid mechanics to dispel three ‘myths’ about special science NKs. First, he argues that unlike purely disjunctive kinds (e.g., jade), which yoke together properties without any real causal links, polythetic NKs like viscosity or Newtonian fluid afford novel predictions and explanatory unification across separate and micro-structurally distinct domains (e.g., fluids, gases, and solids) by allowing us to tap into an entire network of causal relations. Second, he rejects Kim’s argument that macrolevel entities must be causally inert by opposing the ‘causal exclusion’ principle on which it rests, for it depends on a metaphysically loaded ‘production’ or ‘generation’ model of causation that is out of sync with many of our causal-explanatory practices. Khalidi doesn’t explicitly endorse any particular analysis of causation, but his ‘liberalized’ causal theory of NKs is most plausibly premised on an interventionist or counterfactual understanding (pp. 97-98). Third, Khalidi defends a ‘non-dichotomous’ view of natural laws, arguing that whatever features are taken to be constitutive of lawfulness can either be had by special science kinds, or be equally absent in 2 Khalidi gives a few intuitive examples of non-Newtonian fluids to illustrate this property (p. 86). For example, non-drip paint thins out (more flow) when you stir it harder, whereas pudding stiffens (less flow). Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) basic science kinds. The distinction between so-called ‘universal’ and ‘restricted’ laws is merely taken to reflect the relative abundance of different kinds in the universe (e.g., protons are simply more plentiful in the universe than economic markets), and the background conditions upon which basic science kinds depend are typically stable relative to those that figure in the special sciences. What stands out from Khalidi’s discussion is his insightful analysis of ‘crosscutting’ kinds (Chapter 3.6). Scientific kinds cross-cut one another if they apply to the same things, but in virtue of picking out different aspects of the underlying phenomena, pertaining to different types of causal processes (p. 114). In some cases, cross-cutting descriptions are pitched at the same spatiotemporal scale, such as when atoms are cross-classified by their atomic number or their mass number, or stars according to their spectral type or their luminosity. In other cases, such as the cross-classification of fluids according to their chemical composition or their fluid mechanical kinds, the phenomena which they subsume are separated by several orders of spatial magnitude (pp. 118-19). The latter cases prove to be particularly instructive. Over and above standard cases of multiple realizability, where we have one-to-many mappings from higher to lower levels, here we have the opposite: e.g., two fluid samples of the same microkind (say, having molecular composition X) that belong to two different macrokinds (Newtonian vs. NonNewtonian fluids). In general, systems of cross-cutting kinds typically stand in a many-to-many relationship between higher and lower levels of description. This is consistent with the global supervenience of the macro on the micro, provided that the relevant macrolevel differences are due to extrinsic or relational factors (such as temperature differences in the fluid mechanics example). Khalidi goes on to suggest – as others have before (cf. Wilson 2001) – that systems of crosscutting kinds are fairly ubiquitous in the special sciences. For example, in the philosophy of mind, content externalism is the thesis that two people can have different kinds of intentional mental states despite the fact that their brains are in the exact same type of neural state. This claim is based on the argument that intentional mental states are individuated in terms of their semantic content, and semantic content is categorized extrinsically in terms of certain causal or historical relationships which a person bears to her physical (Putnam) or social (Burge) environment. If Khalidi is right, then content externalism – far from being a philosophical eccentricity – reveals a deeper insight into what we might call the ‘aspect-relativity’ of ontology. Importantly, talk of ‘aspects’ should not be mistaken as referring simply to the observerrelativity or epistemic context-sensitivity of our taxonomic practices. The metaphysical content of Khalidi’s claim is rather that the special sciences refer to kinds and properties that are objectively invisible at different (esp. lower) spatiotemporal scales, enter into causal processes that are autonomous vis-à-vis their lower-level descriptions, and often cross-cut one another; but as such, they still track “real causal patterns” (p. 121). Khalidi eventually proposes that we Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) jettison talk of ‘levels’ and replace it by talk of ‘domains,’ to avoid the misleading implication that scientific domains can be ranked along a single ontological hierarchy. In Chapter 4, Khalidi expands his account to fund a limited defense of NKs in the biological and social sciences. Again, his main argumentative strategy is to undermine specific claims to the contrary. For example, he rejects Papineau’s (2010) thesis that natural (or artificial) selection is unique in giving rise to autonomous higher-level kinds that are implemented by a diverse set of causal mechanisms (Chapter 4.2). An example would be (human) pain, which is realized by a diverse set of causal mechanisms that mediate between bodily damage and its avoidance. According to Papineau, the causal unity of selected (or designed) kinds is due to the fact that selection (or intelligent design) explains why divergent causes happen to converge on a single effect. Khalidi argues that Papineau gets things backwards: if members of the higher-level kind (human) pain did not participate in robust causal patterns that allow organisms to avoid bodily damage, they would not have been selected for in the first place. In other words, selection works because it is sensitive to causal patterns at the macrolevel, but essentially ignores their realization at the microlevel. Another challenge for Khalidi seemingly stems from the fact that many biological and social kinds are individuated ‘etiologically’ – i.e., by taking into account the fact that their members share a certain causal history (Chapter 4.3). For example, cladists taxonomize biological species (red fox) or traits (radius bone) strictly in terms of diachronic features such as phylogeny or descent, while setting aside similarities among synchronic causal powers. Thus, the red fox is classified more closely with the raccoon dog than with the gray fox despite their superficial resemblances (p. 131). Non-biological examples include the geological classification of objects as meteorites, or of terrestrial rocks as being igneous, sedimentary, or metamorphic, depending on their causal origin. (Davidson’s famous example of sunburn belongs here as well.). The worry here is that if causal history and synchronic causal powers (‘hic-et-nunc’) can come apart, etiological kinds can’t be NKs if NKs are individuated strictly in terms of their causal powers. Khalidi offers an interesting response to defuse this objection (pp. 134-35). First, by subsuming an object under an etiological kind, we can still make inferences about its causal history (“retrodiction is also a type of projection,” p. 135). For example, we often make generalizations about the kinds of historical processes that generated particular members of etiological kinds. And most importantly, there is no reason why science ought to restrict its attention to synchronic causal patterns at the expense of diachronic ones. As the evolutionary distinction between analogous and homologies traits shows, taxonomic schemes that classify things by their synchronic causal powers may cross-cut classifications of the same individuals in terms of their causal history. In the second half of Chapter 4, Khalidi wades deeper into the muddy waters of social and human kinds – a territory that has often been taken to defy a realist understanding, thus Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) prompting a constructivist stance. This is because social and human kinds are closely tied to the symbolic powers of the human mind, including its ability to ‘carve joints into nature.’ However, Khalidi argues that it is far from clear what type of ‘mind-dependence’ is required to undermine scientific realism. Consider Hacking’s (1999) claim that many human kinds (e.g., child television viewer, multiple personality, woman refugee) are subject to the so-called ‘looping effect,’ insofar as the categories that are used to describe, and interact with human subjects tend to induce changes in the phenomena (kinds) under description, which in turn lead to a modification of the attitudes on the basis of which the relevant categorizations were initially made, so as to conform to the shifting social realities (Chapter 4.5). Khalidi notes that there are plenty of examples of ‘interactive’ nonhuman kinds in which deliberate human intervention has also played a significant causal role in shaping up the properties that are associated with being an individual of that kind, such as the creation of domesticated animal species or synthetically engineered chemical compounds. As long as we accurately represent the reflexive dynamics of ‘loopy’ kinds in our scientific theories, what our ‘self-reflective’ representations track are still real causal patterns. Khalidi thus abolishes ‘mind-independence’ as a criterion of NKs (p. 165), because it would falsely rule out the objective reality of kinds that depend for their existence on human beings, their minds, and their actions that are equally “part of the natural world and […] feature in causal processes” (p. 221). As he puts it, “the aim [of scientific inquiry] should not be to guarantee mind-independence but to ensure world-dependence” (p. 165). In Chapter 4.6, Khalidi pushes this thought one step further to deal with ‘institutional’ kinds. Those are social kinds that come into existence, and have the properties they do, only in virtue of social conventions, rules, or laws that specify the relevant conditions of membership (e.g., $10 bill, touchdown). Unlike Searle (1995), whose narrow focus on institutional kinds may have led him to overplay the constitutive mind-dependence of many other social kinds (cf. Wilson 2007), Khalidi places social kinds alongside a continuum in terms of their degree of conventionality (pp. 153-56) 3. That continuum is defined by (i) whether the existence of a social kind K is directly dependent on appropriate mental attitudes towards K, and (ii) whether this is true not just for type K as such, but for each token of K. Khalidi argues that some social kinds (recession, racism, or poverty) satisfy neither (i) nor (ii); others (cocktail party, war) satisfy (i) but not (ii), and only a small subset of social kinds (permanent resident, provost) satisfy both (i) and (ii). Since the first two types of kinds are individuated in terms of processes that are partly causal in nature, Khalidi concludes that their ‘mind-dependence’ poses no threat to a realist construal (p. 154). Only the third types does, because “[a]ny relations between these properties [associated 3 Searle (2010, pp. 116-17) has recently acknowledged that there can be ‘third-person fallout facts’ from institutional facts, such as recession, that do not depend on collective recognition for their existence. Contrary to Thomasson (2003) though, on which feature (i) of Khalidi’s ‘continuum hypothesis’ is based, Searle does not think that racism belongs into the ‘fallout’ category. This is because the fact that the members of a racist society may not be consciously aware of their racist practices, and not describe them as racist, is irrelevant. As long as people of different races are, in fact, treated as having different deontic status (i.e., implicitly recognized as having different rights and responsibilities), racism would still constitute an institutional kind for Searle, whether or not this fact is accurately reflected in the beliefs that are transparently held by racists (ibid, p. 118). Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) with the third kind] are not the outcome of causal processes but the result of a conventional linkage between them” (p. 156). His example of a purely conventional linkage is the property of not having a criminal record, which is associated by law with the institutional kind permanent resident. Khalidi grants that conventional kinds can turn into NKs if their members come to participate in novel causal processes under certain boundary conditions (e.g., when we find out that most permanent residents in a certain state are urban dwellers; p. 157). I am a bit puzzled by Khalidi’s anti-realism about the third type of social kinds. To see why, consider Searle’s (1995) formula for the collective imposition of a symbolic status function, “X counts as Y (in context C)” In Khalidi’s example, the absence of a criminal record is part of the X-term, and thus a necessary condition for a person to be a permanent resident (Y-term). Provided that the existence of the social kind permanent resident is maintained by appropriately functioning institutional mechanisms (that are part of context C), why shouldn’t we consider the deontic powers that people have qua (not) being permanent residents as causal determinants of their actions? For example, why does the deportation of a person who has lost her permanent resident status (= negative deontic power) not count as the outcome of a causal process? Is it because the causal powers of law enforcement agencies depend for their continued functioning on the collective acceptance by the very people who are subject to those laws? While that is arguably true, I don’t see why this type of mind-dependence would be categorically different from that of other interactive kinds, as it is claimed by Khalidi (p. 156). Perhaps institutional kinds are simply exceptionally persistent interactive kinds for which the ‘looping effect’ has been largely neutralized, or at least made invisible on shorter time scales, due to the stabilizing influence of long-lasting societal structures, cultural norms, and other mechanisms of social control. Thus, the historically stable character of institutional kinds, especially if they are codified in terms of formal rules or laws, should make them more amenable to Khalidi’s realist analysis, not less. Khalidi later returns to a related issue, namely the debate between ‘naturalism’ and ‘social constructionism’ (Chapter 6.5). Having rejected ‘mind-independence’ as a criterion of NKs, the question then arises what it means to say that an entity that we purportedly ‘discover’ is in fact ‘socially constructed.’ Khalidi argues that we have to disambiguate social constructionism as a thesis about psychological or linguistic categories (or kind-concepts) vs. a metaphysical thesis about kinds (pp. 224-27). For example, consider the assertion that urban schooling is socially constructed. Understood as a claim about categories, it means that our social-scientific concept ‘urban schooling’ – or any other scientific concept, for that matter – has been significantly shaped by non-epistemic interests such as the intrusion of political or moral considerations. This is a substantive claim which warrants empirical investigation. Understood as an ontological claim about social kinds, however, the above claim is almost trivially true, akin to the claim that cells are biologically constructed. Constructivist claims about social kinds – e.g., the claim that gender or race are socially constructed – are substantive and revealing mostly when they refer to Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) kinds that were traditionally regarded to be (purely) biological, but, upon closer inspection, turn out to bear the marks of a social kind. The inevitable intrusion of non-epistemic purposes into scientific practice hangs like the sword of Damocles over Khalidi’s ‘naturalistic’ account of NKs. As he candidly acknowledges, the categories employed in the human sciences (which, for him, include the social and psychological sciences but also medicine and at least some areas of human biology) do not only play epistemic roles but also have a normative or evaluative dimension (Chapter 4.7). Even as scientists, we not only use categories to describe and explain what human behavior is like, but we also envision or prescribe certain ways in which human behavior could be or should be like, all the while condemning (or praising) it for failing to conform to the projected norms and standards. Khalidi approvingly refers to a study of how the value-driven extension of the category child abuse to include fetal abuse was initiated by law enforcement officers and related agencies who sought the legal means to prosecute mothers who drink heavily or take certain drugs during pregnancy, while pursuing a political agenda (‘crack-as-a-social-problem;’ p. 161). Another example is how the definition of the psychiatric category mild cognitive impairment was strongly shaped by ethical imperatives in favor of taking a more proactive stance towards improving quality of life for the elderly (p. 162). Both examples demonstrate how easily the value-ladenness of human categories encroaches on the supposed objectivity of our ‘best efforts’ at representing the world. Moreover, as critical social theorists have long argued, a strict separation of epistemic and evaluative purposes is not only practically unfeasible but would also be undesirable, if the goal of the social sciences is to emancipate and empower human beings rather than just to study them (p. 163). This point is forcefully made by Fay (1983/1994, p. 108), who is quoted by Khalidi (p. 164) as saying that “[A] critical social scientist actually desires to see his causal generalizations [e.g., about domination and oppression] made otiose by a group of actors who, having learned them, alter the way they live.” Khalidi retorts that Fay’s recommendation in fact presupposes a principled distinction between ‘positive’ (epistemic) and ‘critical’ (emancipatory) knowledge interests, and that the emancipatory project requires that we first understand the causal processes that give rise to, but can in turn be used to prevent or redress manifestations of social injustice. I agree that Khalidi’s causal-realist understanding of social kinds, armed with an interventionist analysis of causation, is compatible with the idea of a critical social science. However, I suspect that social constructionists are going to refuse the bait, and insist that the separability of epistemic versus non-epistemic purposes is a metaphysical relic which does not hold up to scrutiny once science is considered as a socially, historically, and materially situated practice. This would undercut Khalidi’s case for scientific realism, which relies on the ideal of science as the locus of “methods that serve our epistemic purposes” and thus “enable us to identify natural kinds” (p. 216). The inverse claim, which has been central to the ‘strong programme’ in the sociology of scientific knowledge, is the constructivist thesis that “the classification of things Penultimate Draft (7/17/2015) Forthcoming in General Philosophy of Science 47(1) reproduces the classification of men” (Bloor 1982, p. 267). Khalidi anticipates this anti-realist move, citing the Foucauldian slogan that “knowledge is power” (p. 225n.11). His response is to dig in his heels, insisting that the goals of knowledge are distinct from the goals of power. Social constructionists will beg to differ. They might even point to Khalidi’s own characterization of realism to rebrand their own position: if the aim of realism is “to ensure world-dependence” (p. 165), and epistemic and non-epistemic purposes are inexorably intertwined in the causal fabric of our world (which includes human beings and their purposes) shouldn’t a fully ‘naturalized’ version of realism aim to reflect their actual interdependence? Philosophers and scientists on either side of the fence – whether of naturalist or socialconstructionist persuasion – will find plenty of material to sink their teeth into in Khalidi’s delightfully argued and wide-ranging treatment of NKs. I especially encourage the reader to take a close look at Chapter 5, where Khalidi adroitly brings his refined causal account of NKs to bear on a number of case studies in the special sciences, ranging from lithium, polymers, viruses, and cancer cells all the way to ADHD. It is precisely through a close, ‘naturalistic’ examination of such cases that we can get the ‘reflective equilibrium’ methodology to pay its greatest dividends. We can thus only commend Khalidi for his valiant attempt at restoring the notion of NKs to the central place within philosophy where it belongs. References Bloor, D. (1982). Durkheim and Mauss revisited: Classification and the sociology of knowledge. Studies in History and Philosophy of Science, 13(4), 267–97. Craver, C. F. (2009). Mechanisms and natural kinds. Philosophical Psychology, 22(5), 575–594. Evans, G. (1978). Can there be vague objects? Analysis, 38(4), 208. Fay, B. (1994). General laws and explaining human behavior. In M. Martin & L. C. McIntyre (Eds.), Readings in the philosophy of social science. Cambridge, MA: MIT Press. Goodman, N. (1983). Fact, fiction, and forecast (4 edition.). Cambridge, MA: Harvard University Press. Hacking, I. (1999). The social construction of what? Cambridge, MA: Harvard University Press. Lupyan, G. (2005). Carving nature at its joints and carving joints into nature: How labels augment category representations. In A. Cangelosi, G. Bugmann, & R. Borisyuk (Eds.), Modelling language, cognition and action: Proceedings of the 9th Neural Computation and Psychology Workshop. Singapore: World Scientific. Papineau, D. (2010). Can any sciences be special? In G. Macdonald & C. Macdonald (Eds.), Emergence in mind (pp. 179–197). Oxford University Press. Searle, J. R. (1995). The construction of social reality. New York: Free Press. Searle, J. R. (2010). Making the social world: The structure of human civilization. New York: Oxford University Press. Thomasson, A. (2003). Foundations for a Social Ontology. ProtoSociology, 18-19, 269–90. Wilson, R. A. (2001). Two views of realization. Philosophical Studies, 104(1), 1–31. Wilson, R. A. (2007). Social reality and institutional facts: Sociality within and without intentionality. In P. S. L. Tsohatzidis (Ed.), Intentional acts and institutional facts (pp. 139– 153). Dordrecht: Springer.