You are on page 1of 21

Kinds of Things: A Study of Natural and Conventional Kind Classifications

[Summary and Introduction] By David Lockwood (http://www.david-lockwood.co.uk) SUMMARY How do we classify and refer to the various kinds of things encountered in the world? Since traditional, or Descriptivist, theories of reference fail to explain our use of natural kind terms (NKTs), I defend the Causal Theory of Reference (CTR) associated with Hilary Putnam and Saul Kripke. However, contradictions arise within that theory. On the one hand, the problem of indeterminacy of the same-kind relation suggests that members of natural kinds possess no mind-independent Real Essences. On the other hand, Kripke makes a convincing case for regarding such statements as "Water is H2O" as necessary a posteriori truths. The notion of regrounding, much used by defenders of the CTR, does not account for "Water is H2O" being a truth both necessary and discoverable empirically. The contradiction is plausibly resolved by Alan Sidelle, who provides a Conventionalist interpretation of Real Essence and of necessity a posteriori. My defence of the CTR entails a clear distinction between natural and conventional kind terms (CKTs), since applications of that theory to CKTs cannot be justified. I argue that no one theory explains our use of CKTs. Strict definitions may be given in some cases, and both Wittgensteinian family resemblances and Bede Rundle's notion of the 'leading idea' help to give the meaning of artefact terms. I also critically discuss and compare T.E. Wilkerson‟s and John Dupré‟s accounts of natural kinds, and argue for an emphatic distinction between natural and conventional kinds. Most of the ideas discussed here are borrowed, but the connections made between them are largely my own. In particular no other philosopher has (as far as I am aware) shown how we might defend both a Conventionalist interpretation of the CTR (which is the only defensible interpretation of the only defensible theory of reference, given problems with the Descriptivist approach, the plausibility of Kripke‟s advocacy of necessity a posteriori and the problem of Indeterminate Reference) and the notion that causally effective essences determine conformity to laws of nature. Uncritical supporters of the CTR assume a connection between these ideas which involves the false notions of Real Necessity and of (metaphysically) Real Essence. I propose a different type of connection, one suggested by Colin McGinn‟s argument that some apparently dispositional properties (such as malleability) are identical to the microstructural essence (atomic number) which is held to determine such properties. I also offer a solution to the qua- problem by distinguishing between the grounding of natural kind terms and the process of referring to natural kinds (Chapter 3). I suggest how regrounding and a Conventionalist interpretation of the CTR may be harmonised (Chapter 5).
1

I propose that there exist not only natural and conventional kinds, but also an intermediate category, scientific kinds (Chapter 1). Terms denoting such kinds behave like CKTs, but the kinds themselves are amenable to scientific investigation. Against Dupré, I argue that dispositional properties could not form the basis of laws (Chapter 1). I also suggest that the difference between those dispositional properties which are and those which are not reducible to invariant microstructures is scientifically testable (Chapter 5). In summary, the chief purpose of this book is to show how our classifications of natural kinds can be both conventional (in that they derive from human decisions) and nonetheless objective (in that they are based on real features of natural kinds which make them subject to laws of nature.) A subsidiary, and largely implicit, objective is to defend the Realist claim that the world is as it is independently of our perspective upon it. TABLE OF CONTENTS Introduction A Defence of (Weak) Realism and of Externalism The Analytic-Synthetic Distinction Thought Experiments The Contribution of Psychology Chapter 1: Kinds of Things T.E. Wilkerson's Categorisation of Kinds Scientific Kinds Characteristics Of Natural Kinds Conventional Kinds De Sousa on Laws of Nature Implications of a Conventionalist Account of Natural Kinds Dupre on Natural Kinds Chapter 2: Theories of Reference The Description Theory of Reference The Causal Theory of Reference Putnam's Account of Meaning Kripke on Necessity A Posteriori Chapter 3: The Grounding of Natural Kind Terms Grounding and the qua- Problem Referring to Natural Kinds Regrounding Theoretical Kind Terms Chapter 4: The Problem of Indeterminate Reference Competing Essences The Stipulation of Essence Scientific and Vernacular Kind Terms Fail to Coextend

2

sounds. unclassified experience is both an amorphous mass of colours. The ability to divide the world into kinds is an evolutionary trait essential to survival. is one of the most fundamental human attributes. smells. tastes and textures and a seemingly infinite number of particular things. Without the ability to classify we would experience the world as does a new-born child. speech and thoughts.” Raw. The child gradually learns to sort its experiences into groups and as it learns to speak. shapes. 3 . it acquires general terms which allow it to order and classify its perceptions verbally.from which follow the faculties of memory. Without this generalising ability . as (in the words of William James) “a blooming buzzing confusion.Chapter 5: A Conventionalist Defence of the Causal Theory Sidelle and Conventionalism McGinn on Essence Chapter 6: Conventional Kind Terms The Causal Theory: Artefact Terms The Causal Theory: Attributive Terms Hybrid Terms Wittgenstein on Family Resemblances Strict Definitions Rundle on the Leading Idea Conclusion Appendix I: The Notion of Similarity Appendix II: Universals Appendix III: Species Neo-Aristotelianism Wilkerson and Narrow Kinds The Species-as-Individuals Thesis Pluralism: Ruse and Kitcher Appendix IV: Classification and Cultural Relativism Bibliography INTRODUCTION Introduction The ability to classify. speak or think we are implicitly classifying the objects of our perceptions.human life would not be possible. Whenever we perceive. reasoning and indeed perception itself (we always perceive an object as something) . to sort things into useful kinds.

empiricists. do our classification schemes dissect the world as it is „in itself‟. a 4 . The second approach is associated with a heterogeneous group of philosophers who lean towards Idealism or Conventionalism. since most philosophers incline to one position or the other. However. Many commentators infer that the ordering of the world would be largely independent of linguistic categories. From the Pre-Socratic philosophers onwards. lawyers. we must initially examine the question in dialectical terms. farmers and builders. cricket bats. However. holding that things possess their properties essentially only as a function of the way they are described. gold ingots. and so on. for example. scientists have sought to classify the world in ways which yield maximum understanding and predictability of phenomena. Generally speaking. irrespective of human concerns. and followers of Quine and of Wittgenstein. members of natural kinds possess essences which determine the extension of the corresponding term. the question arises: Does that perception result from the world being ordered independently of our experience of it. The first approach is associated primarily with Aristotle and with medieval scholasticism.We perceive the world. say) into useful groups according to their function. The purely instrumentalist aims of such systems began to be supplemented by a desire for knowledge for its own sake. in three stages. or does the human mind impose order? In other words. or that our conceptual schemes map precisely onto external reality and necessarily 'carve Nature at the joints'. including Kantians. they acknowledge only de dicto necessity. it holds that any property possessed essentially by some object is possessed essentially in all possible worlds by everything that possesses it. who needed each to sort things (animals and stones. Scope and Strategy I attempt here to adjudicate between these viewpoints by defending. Given that we perceive the world as an orderly place in which objects follow regular patterns of behaviour. According to the CTR. universities. then. Early practical systems of classification would have originated in the needs of. many more formal systems of classification coexist alongside our (largely unreflective) everyday categories. as being divided up into indefinitely many kinds of things: parrots. or along lines corresponding to our needs and interests? Our prima facie response might be that we have been presented with a false dichotomy. The Causal Theory of Reference (CTR) associated with Hilary Putnam and Saul Kripke is (perhaps inappropriately) viewed as a modern continuation of this Essentialist tradition. Broadly. Purely scientific classifications have in turn immeasurably furthered the development of the applied sciences of agriculture and architecture.

resulting in indeterminate reference of the corresponding NKT (Chapter Four). T. In particular. Many separate questions are interwoven in a discussion of natural and conventional kinds. such as Stephen Schwartz (1977) and Michael Devitt (1990). the existence of which has been convincingly demonstrated by Kripke but which is not established by regrounding? In Chapter Five I interpret Essentialism and necessity a posteriori from a Conventionalist viewpoint. Five approaches are mentioned in the course of this thesis. How. Conversely. Locke. then. Kant and Mill to these questions. I do not investigate several related philosophical issues. However. they seem to differ from conventional kinds insofar as they sustain exceptionless laws of nature. The problem of universals is tangential to my enquiry. In Chapter Two I demonstrate the inadequacies of the rival Description Theory of Reference (DTR). mind and science. 5 . discussed chiefly in Chapters One and Six. but I demonstrate in Appendix II how one plausible solution to the problem (in terms of abstract particulars) may be harmonised with my conclusions with respect to natural kinds. but not of NKTs. might we retain a belief in the necessary a posteriori. Wilkerson presents a cogent theory of natural kinds. such as the status of scientific laws. John Dupré (1993) rejects Essentialism. is the question of whether there is a clear-cut distinction between natural and conventional kinds. The topic embraces metaphysics and epistemology. including D.H. the sameness relation between members of one natural kind is sometimes ill-defined. Others. Such indeterminacy indicates the problematic nature of essence. Mellor (1977). A secondary theme of this thesis. The notion of similarity is discussed in Appendix I. but offers little discussion of natural kinds. account of promiscuous realism. and property to be unproblematic. Many philosophers. concluding that the CTR can offer a viable account of NKTs. I argue that it does not apply to conventional kind terms (CKTs). Alan Sidelle (1989) attempts to synthesize causal and Conventionalist approaches to reference. support the CTR with respect to NKTs largely without reservation. I offer the outline of such a approach here. Although natural kinds are individuated conventionally. the philosophies of language. offering the alternative. and philosophical logic. Having defended the CTR with respect to NKTs. some problems remain intractable.modified version of the CTR with respect to natural kind terms (NKTs). but not fully worked out. It follows that there must be a fundamental distinction. Philosophers vary greatly in their response to the CTR and to Essentialism. No one philosopher seems to have attempted an all-inclusive account of both kinds and terms which acknowledges both the strengths and weaknesses of the CTR. In Chapter Three I defend certain details of the CTR against its critics by employing the notion of regrounding.E. repudiate it completely. and take such concepts as substance. Substantive issues in the physical and biological sciences and in psychology are also relevant. There is little discussion from a historical perspective: a fuller account would examine the contribution of Aristotle.

I therefore expound and criticise Putnam's early theories without tracking subsequent vacillations and changes of mind. argues that knowing the meaning of NKTs involves the ability to recognise appropriate occasions for their use. from (2) Gold necessarily has atomic number 79 in all possible worlds. We cut up the world into objects when we introduce one or other scheme of description.independent insofar as (1) remains the case whether or not human observers are present. This essence. Putnam 1990) suggest that he still considers them broadly correct. designated 'Real Essence(A)'. Putnam emphasises our role in conceptualising experience. is mind. The views ascribed to Putnam are largely those summarised in "The Meaning of 'Meaning'" (1975b).) will necessarily (invariably) conform to law of nature L. in Reason.Some philosophers offer Wittgensteinian accounts of NKTs. Secondly. arguing that forms of life track essential properties. how could we know not only that 6 .E.though at the risk of over-simplification . (Putnam 1981. being capable of many different interpretations. T. discussed in Chapter One. To avoid ambiguity . Wilkerson provides a modern account. for example. 52) In rejecting the notion of a 'ready-made world'. In later writings and influenced by Goodman (1978). irrespective of our conventions we may infer the putative metaphysical Real Essence(M) associated with the concept of Real Necessity. Gregory McCulloch (1995) aims to fuse this insight with Putnamian externalism. we may infer a causally effective essence of the kind associated primarily with Aristotle. Wittgenstein can be considered an externalist in that he holds that word-meaning and understanding cannot be divorced from their contexts. but lack of space precludes discussion here. Oswald Hanfling (1984). Alan Sidelle (1989) offers several epistemological and metaphysical reasons for rejecting the notion of Real Necessity. However. Firstly. these remain influential and more recent papers (eg.I distinguish between three cases of real essence. For example. Necessity and Essence The notions of necessity and of real essence are notoriously elusive. Truth and History he writes that: ‟Objects‟ do not exist independently of conceptual schemes. For example. structured independently of our descriptions. from (1) Any substance with atomic number 79 and isotope number 197 (etc. Putnam apparently repudiates his earlier views.

118) In other words. Moreover. in other words. mind-independent necessities in the world and the belief that which feature of gold determines the extension of 'gold' is established conventionally: how. a real necessary truth would rule out some state of affairs. designated 'Real Essence(C)' and examined in Chapter Five.. 'gold' is used only of that substance with atomic number 79 we may infer Sidelle's notion of a conventionally derived essence. Real Essence(A) and (C) may be reconciled. our linguistic conventions prohibit Elizabeth II being described as the daughter of another monarch. Kripke's writings can be broadly harmonised with the notion of Real Essence(C). The problem of indeterminate reference also controverts the notion of Real Essence(M).such as Elizabeth II being the daughter of George VI . However. (Sidelle 1989. from (3) As a matter of linguistic policy.do not themselves rule out other states of affairs. I take de re Essentialism to follow from Real Essence(A). because this notion can be interpreted from a conventionalist viewpoint. I avoid his usage in order to prevent confusion with the Lockean notion of nominal essence as (superficial) defining features. although I attempt no detailed textual analysis.. what could be added to any state of affairs to make it necessary? Necessity is not an ingredient: the factual content of 'Hesperus is necessarily identical to Phosphorus' is nothing beyond 'Hesperus is identical to Phosphorus'. source. Chenyang Li (1993). if necessity were a real feature of the world. properties apparently possessed as a matter of real necessity . I ignore the many attacks (eg Tichy 1983) which have been made on Kripke's arguments for the necessity of identity and of origin. However. Similarly. I also argue that a paper by Colin McGinn (1974) suggests how we may retain both the idea of de re. assumes the repudiation of Real Essence(M) to involve the denial of the necessary a posteriori. Many philosophers take the notions of Real Essence(M). Finally.something is true. in Sidelle's words . Essentialism and the necessary a posteriori to be inseparable parts of one package. but that it is necessarily true? The empirical evidence for 'Gold has atomic number 79' could not establish that it is necessarily true. Sidelle frequently uses 'nominal essence' to denote what I call Real Essence(C). Accepting the validity of those arguments need not entail accepting Real Essences(M). However. The fact that we need a description to explain what is ruled out shows that the necessity has a linguistic.. considered nonverbally. 7 . I reject Real Essences(M) but defend Real Essence(A) and (C). for example. and not a metaphysical. Belief in Real Essences(M) is frequently attributed to Kripke.

or (ii) physical entities and mental entities. alongside Darwinian principles. after all. and (in some cases) to decide species membership on the basis of a frequency distribution of characteristics (x might be held to belong to species S if it possessed sufficiently many of a cluster of features associated with S. a classification of diseases is pointless unless directed towards explaining the causes of. Categories Kinds must be distinguished from categories. quantity and quality from different kinds of question that might be asked about something.Classification and Theory For much of their history. Linnaean principles still govern much of biological taxonomy. an established nomenclature tends to preserve established taxonomies and hence the theories upon which they are based. that a classification is not an ultimate objective. For example. Some philosophers have argued that classification per se is a lower order activity: that is. “What is x made of?” “How tall is y?” If a system of categories is complete then every entity in the world will belong to one and only one category. all things (here. However. Aristotle derived such categories as substance. Categories are genera which might contain species but are not themselves species of any higher genera. all samples of silver) that have the atomic number 47 also have the boiling point 21630C: everything that belongs to class x also belongs to class y. but the scheme is in turn influenced by the theory. Early botanists. However. it is arguable whether there is a real distinction between classification and theory/explanation in science.) Nonetheless. There is a reciprocal relationship between theory and classification: every theory presupposes a classification scheme. since 8 . For example. The formulation of universal (exceptionless) scientific laws presupposes classification. it is arguable whether it is possible to produce a set of categories containing more than two items. disease. according to Aristotle. which are. the ultimate or most fundamental divisions of things. such as (i) things and non-things. is less important than understanding the reasons for those differences. biological taxonomies were profoundly changed by Darwin‟s theory of evolution: they were forced to discard the notion of immutable species (and hence of homogeneous classes). For example. The mere observation of differences. zoologists and geologists were concerned chiefly with arranging their specimens in some kind of helpful order.with the exception of physics . to incorporate borderline cases. For example. Every universal law may be analysed as the generalisation to the effect that everything that possesses the set of properties P1 (properties essential to the kind) also possesses the set of properties P2.seem to have been primarily classificatory activities. such as. the theory underlying the law assumes the coherence of classes. but rather a preliminary to the more important business of explanation. In other words. the sciences . and thus preventing.

For stylistic reasons I use several expressions interchangeably to designate microstructural features of natural kinds. Although genetic essences are mutable. His system does not classify things existing in the world („noumena‟) because the true nature of those things is ultimately unknowable. for we presumably only perceive the former because the latter structure our perceptions The Problem of Species I discuss biological species chiefly in Appendix III. The categories are consequently a priori in nature: they are presupposed by and not derived from perceptual experience.'gold'. Aristotelian and Kantian categories are not mutually exclusive. They were originally selected because . 'tiger' etc.) Consequently the modified CTR presented here applies to all NKTs. the indeterminacy of the same-kind relation between different instances of many chemical kinds suggests that there is no fundamental difference between these and biological kinds. since the issues raised do not bear directly on the main lines of my argument. These include 'essence'.being stipulated the Real Essence(C) of a species. on the other hand. A basis for dividing the world into kinds is conventional if it has been established by human practices. They are best characterised as innate mental structures which we impose on the objects of perception. (Indeterminacy of the same kind relation does not apply to the narrowest chemical kinds such as diamond or D2O. 'attribute' and 'characteristic' are treated as synonyms. 'Property'.for the sake of historical continuity.they have belonged simultaneously to both vernacular and scientific English for many years. 'water'. 'plutonium' . From the arguments advanced in Chapters Four and Five it clearly follows that nothing prohibits one feature . I suggest that species might usefully be considered natural kinds. say. Other Terminology The expression 'possible world' is taken to mean 'possible alternative counterfactual situation'.unlike. For example. classifies the ways in which the human mind (necessarily) imposes order on the world. Against Dupré and other commentators. Otherwise we would perceive seemingly random events. the members of which are identical in all relevant respects. and 'internal/ hidden/underlying structure/constitution/traits. with no connection between movements of snookerplayer and ball.normally the genetic constitution . It is important to distinguish between the conventional and the merely arbitrary. but this being so is of course compatible with the existence of good grounds for the choice of principles of division. .' In this context 'essence' has no modal implications. 9 . No such grounds underlie an arbitrary way of grouping things. when we see a snooker-player hitting a ball we view that event under the categories of substance and cause. I normally use standard examples . Kant‟s system of categories. 'feature'.larger sets always contain some overlap.

36) argues that we make worlds by drawing one boundary rather than another. Discussions of Realism frequently conflate two logically separate theses: the weaker claim that reality exists independently of our representations. The doctrine of Conceptual Relativism states that because we can always devise alternative classifications of the world. I defend a (weak) version of Realism and the value of the analytic-synthetic distinction. 163) Moreover. the reality represented is interestrelative. 159) points out. Nothing in the sky dictates that the Pole Star. as John Searle (1996. and the stronger claim that reality 10 . the CTR is unaffected. the world must logically be at least partly independent of our experience and theories. However. Postmodernists such as Jacques Derrida apparently assume that because our representations are relative to our interests. however. I present reasons for treating thought-experiments with circumspection and mention the contribution of psychology to the topic. "make worlds" but rather describe the world with varying degrees of accuracy.Putative initial baptisms would have employed ancestors of English NKTs. lies in the constellation Ursa minor. such as Proto Indo-European *wedar. for example. all are conventional. as when astronomers determine arbitrarily that a star belongs to a certain constellation." (Searle 1996. it does not follow that reality is constructed by those mechanisms. A DEFENCE OF (WEAK) REALISM AND OF EXTERNALISM I assume a Realist account of the external world to be broadly correct. because our knowledge of reality is mediated through biological and social mechanisms. Many philosophers have challenged this commonsense assumption. Searle notes that without the presupposition of an external reality to provide standards against which to measure claims to knowledge. In the remainder of this Introduction I state the reasoning behind some of the presuppositions which underlie the main body of my discussion. Conceptual Relativists could not explain how the truth of opposing views might be established.the world of brute facts .is logically independent of human representations. Since this word is causally related to 'water'. Nelson Goodman (1984. by another it does not. We do not. arguing that we have no direct access to reality other than via social constructs. …the real world does not care how we describe it and it remains the same under the various different descriptions we give of it. Goodman's assertion that we make constellations by drawing boundaries already assumes something on which boundaries can be drawn. Realism can be defined as the metaphysical thesis that the way things are . Searle disagrees that such considerations refute Realism. They presuppose a languageindependent reality which can be divided in various ways. However. This is simply a non sequitur. By one criterion the Pole Star lies in Ursa minor.

In other words. In Wittgenstein's Tractatus Logico-Philosophicus (1922). This aim might seem incompatible with my conclusion that the connection between an NKT and its referent is established by convention. but assert that scientific interests should take precedence. Indeterminacy of the same kind relation consequently does not entail Relativism. because the features which comprise the Real Essence(A) (one of which is selected to constitute Real Essence(C)) jointly ensure compliance with invariable laws. The weak version can be further subdivided. one feature being chosen to determine the extension of the NKT. Internalism. presuppose an external object and are thus. for example.) plausibly argues. with minor reservations.itself determines how it must be described. I and my Twin-Earth Doppelgänger have different thoughts about what we both call 'water' because our thoughts are directed towards different substances.which entails the possibility of a uniquely correct classification scheme and is associated with Real Necessity . being self-contained with respect to the mind. which convincingly demonstrates that the content of my thoughts about water is determined by how the world is. One of the implicit themes of this thesis is to defend a broadly Realist and anti-Relativist approach to metaphysics. I broadly support this view. holds that all psychological states are ultimately 'narrow'. in his terminology.does not follow from the first. Others. the truth or falsity of propositions is given by corresponding states of affairs in the world. Putnam's externalism is manifested in his TwinEarth narrative. Since there may be indefinitely many true descriptions of the same reality given under competing conceptual schemes. The seeming contradiction is resolved by Colin McGinn‟s argument to the effect that all nomic features of a natural kind are linked. I further discuss (and reject) Relativism with respect to classification in Appendix IV. Such laws entail a reality wholly independent of how we describe it. The content of our mental states depends partly on the external environment and. not by my subjective experiences. such as being jealous. A defence of the CTR is implicitly a defence of 11 . Externalism Externalism may be characterised as the view that the mind is not self-contained with respect to the objects of its experience. 'wide' states. Putnam notes that many psychological states. 149ff. Some Realists acknowledge that there are many ways of classifying the world. in particular. truth and reality can never correspond precisely. One impetus behind much of Western philosophy is the belief that truth and reality must coincide. However. including Dupré (1993). things which are external to the body are not necessarily external to the mind. The second thesis . as Searle (1996. representations of the world are made and are true seen only under a particular aspect. on the other hand. claim that the scientific is just one among many objectively defensible classifications. our understanding of NKTs is bound up with our interactions with the kind designated.

and synthetic propositions such as (5) There is no life on Mars .V.which are true by virtue of the meaning of the words alone.such as (4) Vixen are female foxes . The analytic has been characterised as what is 'true come what may'. However.of the distinction. but Quine responds that no claims. in the 1850s Georg Riemann demonstrated that in non-Euclidean geometries an infinite number of parallel lines pass through a given point.Externalism.whose truth is discoverable only by empirical investigation. philosophers have distinguished between purportedly analytic propositions . Unrevisability Quine dismisses all proffered definitions of 'analytic' as either circular or nebulous. THE ANALYTIC-SYNTHETIC DISTINCTION W.. insofar as stable generalisations about natural kinds are synthetic while generalisations about the function of some conventional kinds are analytic.. which he describes as ".including the axiom that two straight parallel lines never meet could not be conceived false. Quine's belief that all propositions are potentially revisable has two sources. a massive alliance of beliefs which face the tribunal of sense experience not individually but only as a corporate 12 . For two thousand years mathematicians believed that Euclidean geometry expressed analytic truths because its principles .or at least the usefulness . He is a sceptic with respect to whether empirical methods guarantee the truth of our propositions. claiming that reference is unmediated through analytic definitions. However. However. and hence analytic. Quine (1953) famously contends that the traditional distinction between the analytic and the synthetic is untenable. A causal theory of meaning appears to make analyticity superfluous. Since Kant. I also claim that natural and conventional kinds can be sharply distinguished. Clearly a full scale critique of his position is impractical here. including the apparently timeless truths of mathematics. are immune to revision. I argue that a sound causal approach requires principles of individuation which are conventional in origin. and a holist with respect to our conceptual system. I am obliged to indicate how Quine may be answered because later discussion assumes the validity .

Circularity Quine also claims that we lack a non-circular criterion for analyticity. as Quine describes events running counter to previous experience. Evidently these principles are philosophically important. If peripheral beliefs. such as (5). 41) All our central beliefs are interdependent. is one of degree rather than kind. but being so connected renders them subject to revision. but. only in being more difficult to imagine false. Beliefs such as (4) appear unrevisable and have the semblance of analyticity. our central beliefs would remain broadly unchanged. such as "All members of natural kinds possess their chemical or biological microstructure essentially". immeasurably more probable that such experiences were freakish and unreliable than that a proposition to which all living scientists assent is false. Consequently. and if one is jettisoned.. differ from obviously revisable beliefs. permit synthetic and revisable statements to be predicated of NKTs. Does it make sense to claim that we cannot be certain of the truth of any contingent proposition. The difference. Quine does acknowledge a distinction between statements playing central roles in our theories . are neither coextensive with nor deeply linked to central beliefs. if there are analytic truths." (Quine 1953. Consequently any statement is revisable. that the truth of (6) might be disproved by recalcitrant experiences? It is. however. Sidelle suggests that it may be a brute fact that there exist a very few interesting analytic truths. they are of little significance. Their centrality does not therefore explain our belief in their unrevisability. The holistic argument is susceptible on the grounds that some beliefs most difficult to imagine false. such as (5). In response to Quine.such as (6) The Earth goes round the Sun . 139) argues that the denial that analytic definitions can be given for NKTs is compatible with accepting that analytic principles of individuation underpin the necessary a posteriori and our use of NKTs. Defenders of Quine might argue as follows: philosophically significant beliefs are clearly connected with others. Alan Sidelle (1989.and those on the periphery. we may query the cogency of the underlying sceptical thesis. but would readily disclaim (5) when appropriate. (6) could not be revised without indefinitely many other central beliefs being affected.. required revision. We would resist the denial of (4) in all currently conceivable circumstances. Holism blurs the distinction between the analytic and the synthetic: they are distinguished only by the extent to which we are willing to abandon beliefs in the face of 'recalcitrant experiences'. If we define analyticity 13 .body. These.. however. according to Quine. after all. we have no way of knowing in advance which others must be amended. such as (4).

It is. We need not define synonyms as terms which necessarily pick out the same objects. we need independent accounts of these notions. Two responses may be made. it does not follow that they are necessarily synonyms. as H. However. Consequently. nor that because counterexamples to some purportedly analytic statements have 14 . if we explain analyticity in terms of synonymy we are reasoning in a circle.F. However. Although 'vixen' and 'female fox' necessarily refer to the same object. 148ff. Many . Strawson (1956. In any case. For example. 'Vixen' and 'female fox' are not merely coextensive terms like 'renates' and 'cordates'. Quine's definition of synonymy as 'necessary identity of reference' is vulnerable to criticism. and do not explain. synthetic insofar as it expresses a claim which could be verified or falsified by empirical investigation. We might assert that (7) is indeed straightforwardly analytic: the meaning of 'green' entails that whatever it describes must be extended. Suppose that (4) is considered analytic because subject and predicate are synonyms. or of synonymy. arguably. (7) Everything green is extended is analytic in that. the relation of (say) synonymy. it follows neither that because there are difficulties in categorising some propositions.) point out. Grice and P. Borderline cases Quine notes that some borderline propositions cannot be easily assigned to either analytic or synthetic categories. dictionary definitions merely report. which differ in meaning despite denoting the same objects.perhaps all . or of necessity. Analyticity might be plausibly characterised in terms of necessary identity of reference. Analyticity depends on synonymy which depends on necessity which depends on analyticity. incoherent.P. arises from terms being necessarily equivalent to each other. and the very notion of precise synonymy is. arguably. any coloured thing must comprise more than a mathematical point. circularity need not be vicious. none can be categorised as either analytic or synthetic.in terms of that which cannot be denied without self-contradiction. since necessary identity of reference cannot be conflated with synonymy it seems that we might achieve a non-circular definition of 'analyticity'.apparent synonyms have different connotations. Quine separately argues that necessity is explicable only in terms of analyticity. according to Quine. necessarily. However. The usefulness of 'good' and 'duty' is unaffected by our inability to provide strict definitions using terms outside a cluster of related expressions. Claiming that 'vixen' and 'female fox' are intersubstitutable salva veritate in all contexts merely amounts to saying that "'All and only vixens are female foxes is analytic' is true". precise sameness of meaning. More importantly. even though 'green' does not mean 'extended'.

since 'analytic' and 'synthetic' can normally be applied to new cases without controversy. 143) The practical usefulness of the analytic-synthetic distinction is undeniable. What. In Chapter Four I argue that the same-kind relation is indeterminate because Real Essences(M) do not exist. not because (4) is less easily imaginable false. without seeking to 15 . to say that there is no such distinction. his holistic account of revisability is not unassailable. (Grice and Strawson 1956. A one year-old child cannot logically be an adult. Conclusion Quine demonstrates the weaknesses of naive accounts of analyticity. My justification for relying upon it in the following discussion therefore rests on both pragmatic grounds and on Quine's failure conclusively to prove his case. but only practical considerations prevent it from mastering the differential calculus. THOUGHT EXPERIMENTS Introduction Putnam and Kripke rely heavily upon thought-experiments in attempting to demonstrate the invariable sameness of kind which some of their supporters associate with Real Essence(M) . of the claim that a one day-old has solved Fermat's last theorem? The dividing line between logical and physical impossibility blurs. Presupposition of essentialism Nathan Salmon notes that adherents of the CTR assume the truth of. and this seems to suggest that it is absurd. however. some kind of distinction surely exists. Clearly there are indeterminate areas between the analytic and the synthetic. We distinguish between (4) and (5) on the grounds that the truth of each is established differently. Moreover. and hence unproblematically analytic.been produced. However. Physical/Logical Possibility The distinction between analytic and synthetic is allied to that between the logically and the physically (im)possible. I claim here that philosophers' beliefs in such essences originate in flawed or inconclusive thought-experiments . even senseless. How could there be a counterexample to (4)? Surely it is unrevisable in any circumstances. counterexamples will always be found. Grice and Strawson write: 'Analytic' and 'synthetic' have a more or less established philosophical use.

but in fact presuppose. (8) is just of this form. and nothing against. for example. Questions of the form "What is it to be cat?" asks what cats have essentially. He smuggles hidden stipulations into his arguments: namely. (8). in giving an essence for a kind of kind. Putnam's tale of Martians plotting to spy on Earthlings is fundamentally no more than colourful stage-setting. whether. This premise begs the question. Moreover.) argues that Putnam and Kripke employ hypothetical cases which purport to demonstrate the existence of. we would continue to describe cats. He then seeks to prove his argument by supporting it with intuitions which presuppose that stipulation. Putnamian and Kripkean scenarios omit detail which would draw attention to their own implausibility. newly revealed to be Martian robots. The response (9) To be a cat is to have the genetic essence of cats presupposes cats to be of a kind because they share genes. that (8) Being an instance of the same kind as x consists in having the same microstructure as x. as 'cats'. 16 . its hidden essence. Standard Wittgensteinian responses to scepticism also apply to thought-experiments describing bizarre cases. with the most cursory of detail. presupposes essentialism. the Anti-Realist philosopher Garth Hallett (1991. Cats have been dissected and been observed giving birth for thousands of years: the idea that in every case people were deceived by Martian technology is all but meaningless. Real Essences(M). that the only valid relation of sameness derives from the sharing of microstructural features. 95ff. Putnam already knows what determines whether x is a cat: namely. In similar vein. like (9). but at a one-step remove because it makes the more general claim that microstructure determines kind-membership.prove. the circumstances described are frequently so bizarre and fantastic that we remain unsure whether we are obliged to apply concepts in these new circumstances as in the past. However. cats always having been animals (cf. is couched in less precise terms than (9). We might counter Putnam's Martian example by asserting that everything counts towards. Their thought-experiments are designed to ascertain whether we would apply familiar concepts to new and uncertain cases. Wittgenstein 1969b #203). in using such phrases as "the cat as we know it" and "genuine cats" Putnam might presuppose what he intends to demonstrate. Moreover. We notice this less readily in the first case because (8).

the relationship between everyday terms and their scientific counterparts is asymmetrical. and concludes that this animal is both a scientific and a logical impossibility. with the most cursory of detail. 263ff. With each new supposition other laws of nature are affected.indicates that as presently constituted they could not exist as carnivores. but hardly provide conclusive support for Real Essences(M). . 109) Wittgenstein explains why we cannot pass judgement. For example. how can we be expected to pass judgement without knowing the missing details? (Hallett 1991. as Enc (1976. George Seddon (1972. Our intuitions give guidance as to our likely responses in counterfactual situations. Hallett notes If the details omitted [in Putnam's narrative] would be relevant in real-life situations. Theories are presented in terms of necessary and sufficient conditions and assessed by rules regarding consistency. we must assume a dramatic drop in the rabbits' birth-rate. there may be no means of deciding between them. since carnivores require more territory than vegetarians. Many other conjectures would be needed to accommodate this change of diet. Putnam is probably correct with respect to ordinary speakers. why are they irrelevant in imagined situations? If they are relevant in imagined situations. Putnam's case gives rise to similar difficulties. which weighs the claims of competing intuitions. Putnam (1962) assumes that our use of 'cat' would remain unchanged following the discovery of the real nature of cats. Everything we know about rabbits . He describes …a game of chess translated according to certain rules into a series of actions which we 17 . breeding habits.) examines the notion of a carnivorous rabbit. Putnam's tale is fundamentally no more than colourful stage-setting. However. Neither conclusion is forced on us. Pictures. digestive system. He thereby neutralises criticism insofar as counter-examples cannot refute pictures. etc.Reliance on intuition Kripke characterises his views as 'pictures' rather than as 'theories'.about their teeth. but zoologists might rather deny the existence of Felis domesticus than make wholesale revisions to the theoretical foundation of their taxonomies. every other part of the scenario is affected in unpredictable ways. 491ff. Suppose some rabbits to have overcome the lack of canine teeth by holding meat in their mouths until it dissolves. on the other hand. Lack of detail Thought-experiments specify attendant circumstances only in the vaguest terms. could intelligent life have arisen? Could Martians visit Earth unperceived? How are robot cats repaired when they misfunction? When we respond plausibly to one question. economy and coherence. Given what we know of Mars. and appear less persuasive when details are sketched in. When intuitions conflict.) points out. but we would be uncertain as to which and to what degree. are appraised by the faculty of judgement.

we might conclude that it constituted an unusual mourning ritual. 303). However. that cats are Martian robots] (since. say. But this predictability is exactly a criterion of constancy in meaning. taking P as "All cats are mammals". and Since there is no general way to determine how many of our beliefs may need to be altered as the result of such a discovery [for example.enable us to predict whether certain concepts are applicable in changed circumstances. Goosens writes: “That the properties associated with 'toad' change under the impact of the new knowledge is completely predictable.say into yells and stamping of feet‟ (Wittgenstein 1953. Goosens argues that we would still call T2s 'toads' because they share the underlying genetic traits of T1s. We can not. It is not part of the meaning of 'toad' that kind-members have lumpy blotched skin.do not ordinarily associate with a game . William Goosens (1977) offers a thought-experiment similar to Putnam's.that our correct prediction of 'toad' being used of T2s entails that it cannot have changed in meaning is less an argument than an essentialist presupposition.” (Goosens 1977. if we discovered that all cats were Martian robots we would discard the belief that all cats are mammals because it stands in contradiction to not-P. carrying disease-resistant genes. Hallett concedes that we might correctly predict linguistic usage in unusual situations. 148) Goosens. the ability to answer questions about what we would say in such circumstances is not implicit in our mastery of the rules of our language. A new generation of toads. based on merely a knowledge of current usage. are as sleek-skinned and agile as otters. there can be no general way to determine how much of our way of talking such a discovery may require us to revise. claims that our shared linguistic habits . with Putnam. since we lack certain knowledge of what we would say. (Fodor 1971. the set of beliefs that it would be reasonable to abandon should the belief that P prove false need not be coextensive with the set of beliefs formally incompatible with not P). Suppose that all past toads (T1s) suffered from a disease the symptoms of which included calloused skin. #200) Are the participants playing a game? We cannot answer unless informed about the context of the activity. If it were performed at funerals. 305) In other words.which incorporate deference to science . however. sameness of meaning cannot be guaranteed. T2s. To claim that predictability provides a sufficient criterion for sameness of meaning . whether 'toad' is appropriately used of otter-like creatures. that is. Jerry Fodor writes that …there is no reason to trust our intuitions about what we would say in situations in which some of our relatively secure beliefs have proved false. (Fodor 1971. ascertain in advance of the situation arising how 18 .

say. For example. THE CONTRIBUTION OF PSYCHOLOGY Although the following discussion is primarily philosophical. 9) have argued that general concepts are abstracted from particular objects. we are supposed to form the concept house by abstracting the essential properties of houses from particular instances. in forming the concept house we do not examine a range of house and then discard features not common to each: in which case we would be left with barely anything. since we may conceive of Twin-Earth as simply another planet. we cannot have a general idea of objects lacking particularised features. such as "All cats are carnivorous" (we can conceive of a robot which ingests meat. which is plainly impossible. The case of Twin Earth might become susceptible to the objections previously raised against reliance on bizarre counterfactual situations. xi. but by learning to recognise paradigm 19 . However.) Hallett summarise his argument thus: What we say can be observed and reported. George Berkeley (The Principles of Human Knowledge #6ff. including John Locke (Essay Concerning Human Understanding. When we form the idea of a triangle. How do we form the notion of a kind? Many philosophers. A house is not something which is made of no specific material or built in no specific style.many other beliefs which are not logically incompatible with not-P we would have to relinquish. can we be sure what we would say? (Hallett 1991. the empirical findings of psychology are clearly relevant. Similarly. one in which a substance could share all the superficial qualities of water and yet have a wholly different formula. 101) Twin-Earth Putnam's Twin-Earth thought-experiment apparently avoids problems arising from possible world scenarios. How. it might be argued that we are being asked to imagine another possible world. we always conceive a particular triangle. most psychologists believe that the actual formation of categories is governed not by acquiring understanding of necessary and sufficient conditions for x belonging to category X. Moreover.) argues that any putative general idea would have to be general enough to include all xs yet precise enough to include only xs. in imaginary and often fantastic circumstances. Following the pioneering work of Eleanor Rosch (1978). cannot. In response. II. overlooking differences in materials and styles of construction. What we would say. then.

the discoveries of scientists. to one category by comparing individuals with a prototype. Judgements about similarity and about category membership thus vary independently of each other. Subjects were then told that sorps have been exposed to pollutants which change their appearance and habits: they now have six legs and a brittle shell. In the experiment mentioned above. They moreover identified categorymembers without relying on perceptual similarities alone. For example. However. Gelman and Coley demonstrate that the categories formed by young children are not. 148) children were presented with the description of a "sorp". and clung to this belief when confronted with an anomalous case. and that these categories are the product of convention. but complex and theory-laden. and there is no means of weighting properties independently of a theory. our narrower categories of particular natural kinds. without either having acquired further theoretical understanding or having learned to make perceptual distinctions. In other words. They conclude both that we are predisposed to see the world in terms of natural kind categories. children acquire the concept of bird not by learning definitions but by assimilating robins. Children. Nelson Goodman (1977. Children frequently modify their concept of an x merely by hearing x designated a kind-member.cases of X. Susan Gelman and John Coley (1991) have also studied categorisation. and not a visual comparison of some substance with a known example of gold. Once formed.backed by no theoretical framework . according to the Causal Theory of Reference. Evidently such conclusions assist my line of argument. are ultimately established by science.provides no constraints on category membership. establish that „gold‟ is used correctly of certain things but not of others. which nests in trees. has feathered wings and two legs. as we might suppose. Both linguistic convention and features of the world led subjects to form appropriate categories. such as gold. Judgements of similarity depend on the relative importance of attributes. Language plays a crucial role in the construction of natural kind categories. In one experiment (cited in Gelman and Coley 1991. In other words. but produce normal offspring. sparrows etc. Subjects now described sorps as insect-like. Consequently similarity alone . They agreed that sorps were probably birds. They rather 20 . they considered category membership to be determined not by surface similarities but by underlying properties. naive and intuitive. overlook surface similarities in preference to non-obvious properties of inductive potential when forming categories. although our broad categories may be formed by reference to prototypes. A theory posits the relative importance of some features over others. a category serves to guide inferences beyond what is already known. In deciding whether penguins and sorps were of one kind they did not weigh similarities against differences. 107ff. but still categorised them as birds. like adults.) points out that any two things are similar in indefinitely many respects (see Appendix I). children assumed that membership of the kind bird captured certain deep properties of sorps.

children usually encounter stereotypical members of a category before peripheral ones. If a child's first introduction to the category bird was a collection of ostriches. we distinguish between lemons and the trees to which they are attached. Firstly. For example. 21 . In our capacity to discriminate and individuate we employ theoretical models of how the world is divided up. they assume that proper nouns are more likely than other parts of speech to refer to natural kinds. A cat is an animal. for example.decided whether both were birds by assessing the coherence of the kind bird. Thirdly. To what extent does linguistic usage overrule surface appearances in determining categorymembership in difficult cases? Research indicates that children detect various clues as to whether a word names a coherent kind. This finding supports Putnam's (1962) claim that discovering that cats are Martian robots would not affect everyday uses of 'cat'." rather than the converse. These ensure that. parents say "This is a cat. and the precondition for all. Such models correspond broadly to the a priori categories which Kant considered to be imposed by the mind upon. sparrows and penguins it might conclude that the category was incoherent. experience. Secondly. superordinate terms are presented in a manner which intimates that they matter less to the identity of the subject than more specific terms. Our apprehension of order in the world rests on an interlocking network of abilities. Children acquire the (partial) concept of 'cat' before that of 'animal': possessing one concept does not presuppose the other.