Categories of phlosophy

Categories

First published Thu Jun 3, 2004; substantive revision Wed Jan 23, 2013

A system of categories is a complete list of highest kinds or genera. Traditionally, following Aristotle, these have been thought of as highest genera of entities (in the widest sense of the term), so that a system of categories undertaken in this realist spirit would ideally provide an inventory of everything there is, thus answering the most basic of metaphysical questions: “What is there?”  Skepticism about the possibilities for discerning the different categories of ‘reality itself’ has led others to approach category systems not with the aim of cataloging the highest kinds in the world itself, but rather with the aim of elucidating the categories of our conceptual system. Thus Kant makes the shift to a conceptualist approach by drawing out the categories that are a priori necessary for any possible cognition of objects. Since such categories are guaranteed to apply to any possible object of cognition, they retain a certain sort of ontological import, although this application is limited to phenomena, not the thing in itself. After Kant, it has been common to approach the project of categories in a neutral spirit that Brian Carr (1987, 7) calls “categorial descriptivism”, as describing the categorial structure that the world would have according to our thought, experience, or language, while refraining from making commitments about whether or not these categories are occupied. Edmund Husserl approaches categories in something like this way, since he begins by laying out categories of meanings, which may then be used to draw out ontological categories (categories of possible objects meant) as the correlates of the meaning categories, without concern for any empirical matter about whether or not there really are objects of the various ontological categories discerned.

A system of ontological categories drawn out in any of these modes has the potential for a great many uses in philosophy, but those who would offer such systems of categories also face a variety of difficulties. They must address the issue of what the proper methods are by means of which categories are to be distinguished, how many categories there are and what they are, whether or not there is a single summum genus subsuming all other categories, and whether we should distinguish a single system of categories or multiple dimensions of categories—issues on which there has been little agreement.

Over the past hundred years, skepticism about the possibility of offering a uniquely true and complete system of ontological categories has led discussion of categories to shift from attempts to offer complete systems of categories to attempts merely to draw particular distinctions, especially among our conceptual or linguistic categories. Work on category differences, unlike that on category systems, does not generally purport to answer deep metaphysical questions about what things or kinds of things exist; instead, category differences are articulated as a way of diagnosing and avoiding various philosophical problems and confusions. Nonetheless, even those who merely argue for category differences owe an account of the conditions under which two concepts, terms, or objects belong to different categories.

• 1. Category Systems
• 1.1 Aristotelian Realism
• 1.2 Kantian Conceptualism
• 1.3 Husserlian Descriptivism
• 1.4 Contemporary Category Systems
• 1.5 Skepticism about Category Systems
• 1.6 Categories in Other Disciplines
• 2. Category Differences
• 2.1 The Uses of Category Distinctions
• 2.2 The Ryle/Husserl Method of Distinguishing Categories
• 2.3 Fregean Approaches to Distinguishing Categories
• Bibliography
• Academic Tools
• Other Internet Resources
• Related Entries

1. Category Systems

1.1 Aristotelian Realism

Philosophical interest in categories may be traced back to Aristotle who, in his treatise Categories, attempts to enumerate the most general kinds into which entities in the world divide. He does not begin from a single highest kind, but rather lists the following as the ten highest categories of things “said without any combination” (Categories 1b25):

• Substance (e.g., man, horse)
• Quantity (e.g., four-foot, five-foot)
• Quality (e.g., white, grammatical)
• Relation (e.g., double, half)
• Place (e.g., in the Lyceum, in the market-place)
• Date (e.g., yesterday, last year)
• Posture (e.g., is lying, is sitting)
• State (e.g., has shoes on, has armor on)
• Action (e.g., cutting, burning)
• Passion (e.g., being cut, being burned)

There are two sorts of substance: a primary substance is, e.g., an individual man or horse; the species (and genera) of these individuals (e.g., man, animal) are secondary substances. While the ten categories are all equally highest kinds, primary substances nonetheless have a certain sort of priority, since “all the other things are either said of the primary substances as subjects or in them as subjects. So if the primary substances did not exist it would be impossible for any of the other things to exist” (Categories 2b4).

Elsewhere, in Metaphysics (998b22), Aristotle argues explicitly that there cannot be a highest genus (e.g., of being or unity) shared by entities of different categories (cf. Ackrill 1963, 81). For a species is defined in terms of its subsuming genus and differentia (e.g., man is definable as an animal that is rational), and while the genus (animal) may be predicated of the species (man), it may not be predicated of the differentia (rational). As a result, if being (or unity) were a genus, no differentiae could be said to have being (or to be one); but “the differentiae of any genus must each of them both have being and be one” (Metaphysics 998b22–3).

The ancient Greek term ‘kategoria’ described what could be said against someone in a court of law, and indeed Aristotle uses what can be said of or in a subject as a route to distinguishing categories. There is controversy in the literature, however, about precisely how he arrived at his categories (Studtmann 2007). On one prominent interpretation, put forward by J. L. Ackrill, Aristotle arrived at his list of categories by way of distinguishing “different questions which may be asked about something” and noting “that only a limited range of answers can be appropriately given to any particular question” (Ackrill 1963, 78–9), e.g., the question ‘what is it’ can only be asked of a substance, and only answers describing substances are appropriate. The question ‘how much’, by contrast, requires a quantity for an answer, and so on.

But although on this interpretation Aristotle seems to have arrived at his categories by considering different sorts of question and answer, the categories he was offering were supposed to be categories of entities, not of language; language was just a clue to truths about the world. As J. L. Ackrill writes, Aristotle's Categories “is not primarily or explicitly about names, but about the things that names signify…Aristotle relies greatly on linguistic facts and tests, but his aim is to discover truths about non-linguistic items” (1963, 71).

Other interpretations have also been suggested about how Aristotle's categories were derived. Some hold that Aristotle's list was arrived at by reflecting on grammatical categories, and assuming a parallelism between structures of language and structures of the world (Baumer 1993). But others have developed interpretations that do not consider Aristotle to have arrived at his categories by considering linguistic matters such as grammatical structure or the questions we may ask. Instead, they take them to arise from more worldly considerations such as which types of entity any sensible particular must be related to (Moravscik 1967). For an overview of the interpretive options, see Studtmann (2007).

In any case, regardless of how the categories were derived, Aristotle's approach to categories is generally taken to be in the spirit of what Brian Carr calls “categorial realism”—an approach conceiving of a system of categories as a list of the highest genera of beings (not merely of language or thought--even if those may be used in deriving the metaphysical categories). As Studtmann (2007) puts it, Aristotle “ assumes rather than defends a posture of realism with respect to the metaphysical structures of the world”. Given this approach, a complete system of categories would offer a systematic inventory of what there is, considered at the most abstract level (although it is not clear whether Aristotle intended his categories to be exhaustive). Thus on a categorial realist approach, providing a system of categories can be seen as one, or even the central task of metaphysics (cf. Grossman 1983, 3). Such a system of categories may also play a central role in answering individual questions of nature, providing the most general sort of answer to questions of the form “What is this?”, and providing the basis for definitions of narrower sorts of things by specifying the most general category (genus) under which things of this sort fall, and the differentia that distinguishes them from other things of the same category. This has endured as the paradigmatic approach to categories, and several recent authors have offered new theories of categories in this spirit of Aristotelian realism (see §1.4 below).

1.2 Kantian Conceptualism

Others, however, have shied away from this robustly realist approach to categories, generally on grounds of skepticism about our ability to discern intrinsic divisions in ‘reality itself’, and have instead treated the project of categories as a matter of laying out the highest categories governing our conceptual scheme. This shift in approach to what Carr (1987, 6) calls “categorial conceptualism”  was made famous by Immanuel Kant. While Kant famously denied that we have access to intrinsic divisions (if any) of the thing in itself that lies behind appearances or phenomena, he held that we can discover the essential categories that govern human understanding, which are the basis for any possible cognition of phenomena. Thus, as H. J. Paton puts it, for Kant “We can have a priori knowledge by means of the categories, only if the categories are due to the nature of the mind and are imposed by the mind on the objects which it knows” (1936, 258).

In his Critique of Pure Reason, Kant arrives at his list of categories by first enumerating the forms of possible judgment (A70/B95-A93/B109). On this view, objective empirical judgments (i.e., empirical judgments which purport to refer to objects rather than merely subjective seemings or connections of sense impressions, and which purport to be universally valid for all judging subjects) are endowed with their objectivity and generality in virtue of the a priori concepts embodied in the relevant forms of judgment. If we can identify all of the possible forms of objective empirical judgment, we can then hope to use them as the basis to discover all of the most general concepts or categories that are employed in making such judgments, and thus that are employed in any cognition of objects (Körner 1955, 48–49).

Thus in distinguishing his categories, Kant begins from Aristotelian logic in outlining four respects in which one can classify any judgment: according to its quantity, quality, relation, or modality. In each of these respects or ‘moments’ of judgment, there are three alternative classifications; thus, e.g., in respect of quantity, a judgment may be universal, particular, or singular; in respect of its relation, a judgment may be categorical, hypothetical, or disjunctive, and so on. These Aristotelian ways of classifying judgments are the clue to discerning the twelve correlated concepts of the understanding. So, e.g., from noting that all judgments are either universal (e.g., All swans are white), particular (e.g., Some swans are white) or singular (e.g., Cygmund is white), we can arrive at the three corresponding categories of quantity: unity, plurality, and totality. Via this route, Kant ultimately distinguishes twelve pure concepts of the understanding (A80/B106), divided into four classes of three:

• Quantity
• Unity
• Plurality
• Totality
• Quality
• Reality
• Negation
• Limitation
• Relation
• Inherence and Subsistence (substance and accident)
• Causality and Dependence (cause and effect)
• Community (reciprocity)
• Modality
• Possibility
• Existence
• Necessity

The categories are presented as forming a single exhaustive list, with the four classes of categories imposing four different forms of unity on the object known (Paton 1936, 295–9). Thus, one may separately inquire after an object's quantity, quality, relation, and modality, receiving one of the three sub-answers in each case on the way to a more complete characterization of the object.

Although these are categories of the understanding, they nonetheless retain a certain sort of ontological import, as it is a priori that they apply universally to all objects of possible cognition (A79/B105). In this way, by delineating the concepts that are a priori necessary for the cognition of objects, we can acquire knowledge of categories governing any possible object of cognition, and so acquire a sort of descriptive set of ontological categories, though these must be understood explicitly as categories of objects of possible cognition, not of the thing in itself. Thus Kant was able to treat his system of concepts as a system of categories in something like the Aristotelian sense, “for our primary purpose is the same as his [Aristotle's], although widely diverging from it in manner of execution” (A80/B105). Nonetheless, it is clear that for Kant the categories find their original source in principles of human understanding, not in intrinsic divisions in mind-independent reality, and are discoverable by paying attention to possible forms of human judgment, not by study of the world itself, nor by study of our contingent manners of speaking.

An approach like Kant's has been defended more recently by P. F. Strawson and others following him, who undertake the project of “descriptive metaphysics”, which is concerned with describing “the most general features of our conceptual structure” (1959/1963, xiii), thus providing more general and durable results than mere analyses of language can be.

1.3 Husserlian Descriptivism

Edmund Husserl introduced two sorts of innovation to the study of categories. First, while Aristotle used language as a clue to ontological categories, and Kant treated concepts as the route to categories of objects of possible cognition, Husserl explicitly distinguished categories of meanings from categories of objects, and attempted to draw out the law-like correlations between categories of each sort (Smith 2007, 139ff.). Secondly, whereas Aristotle and Kant each lay out a single system of categories, Husserl distinguishes two ways of arriving at top-level ontological classifications: by formalization and by generalization, yielding two separate, orthogonal, systems of categories, in two different dimensions (cf. Smith 2004, chapter 8).

Husserl is careful to distinguish categories of meanings (by way of which we can think about the highest kinds or ‘essences’ of objects) from the categories meant—the latter are the categories of objects, or ontological categories, considered as the highest essences that entities might have: “by ‘categories’ we can understand, on the one hand, concepts in the sense of meanings, but on the other also, and to better effect, the formal essences themselves which find their expression in these meanings” (1913/1962, 61–2). But although the two sorts of categories must be distinguished, according to Husserl categories of the two sorts are essentially correlated (see below), so we can learn about one by way of the other.

Regardless of whether we are studying categories of meanings or of objects, Husserl is quite clear that the study of categories, for him, is an entirely a priori matter; the categories of meanings and objects alike “arise … solely in relation to our varying thought-functions: their concrete basis is solely to be found in possible acts of thought, as such, or in the correlates which can be grasped in these” (1913/2000, 237). As he puts it later, in the Ideas, the study of categories is a study of essences, based in essential insights about the types of meanings and correlative types of things. Such studies of essence may be conducted by way of imaginative variation of cases, independently of any matter of fact, including whether or not there actually is anything of the ontological kinds distinguished (1913/1962, 51). Thus Husserl’s ontological categories, in this sense, are descriptive categories of highest essences of possible things (that might fall under those essences), and do not purport to provide an inventory of what things actually exist (as a matter of empirical fact).

Husserl provides an extensive discussion of categories of meaning in the Logical Investigations, arguing that differences in categories of meaning (which seem to be more like syntactic than semantic categories) can be distinguished by noting where nonsense results from substituting one term for another. E.g., in the sentence “This tree is green” we may substitute “chair”—but not “careless”—for “tree” without turning sense into nonsense, marking the difference between the meaning categories of nominative material and adjectival material (1913/2000, 511–512). Husserl’s understanding of ‘nonsense’ is rather strict: he counts only those strings of words that are syntactically incorrect (so that they form a mere ‘heap of words’ and cannot be combined into any unified meaning (Husserl 1913/2000, 522)) as strictly nonsensical, and thus as signs of differences in categories of meaning. (Husserl repeatedly distinguishes the nonsense of verbal formations like “a round or” (in which no unified meaning emerges) from cases of mere absurdity such as “a round square”, in which the expression has a unified meaning, although it is a priori that no object can correspond to the expression (1913/2000, 516–17)).

Correlated with the categories of meanings are ontological categories; e.g., object, state of affairs, unit, plurality, number, and relation are (formal) categories that categorize objects, not meanings (Husserl 1913/2000, 237). Categories of the two sorts are, according to Husserl, connected by ‘ideal laws’. Thus, for example, presumably objects are the ontological correlates of the meaning category of nominative expressions, properties are the ontological correlates of adjectival expressions, and states of affairs are the ontological correlates of propositions. So while Husserl does not (to my knowledge) explicitly lay out a method of discerning ontological categories, it may be that we can derive them by beginning from the above nonsense test for distinguishing meaning categories, and then shifting attention to the correlative ontological categories, since “pure truths concerning meaning can be transformed into pure truths concerning the object” (1913/1962, 61).

As well as explicitly distinguishing categories of meanings from categories of the correlated objects that could possibly be ‘meant’, Husserl introduced a second innovation to the study of categories by distinguishing highest formal essences (which Husserl calls ‘categories’) from highest material essences (which he calls ‘regions’) (1913/1962, §10; cf. Smith 1995, 329–330 and Smith 2007, 142–148). Thus far I have been describing the formal ontological categories, the correlates of the different meaning categories distinguishable by the nonsense test. In fact, Husserl reserves the term ‘category’ for the highest formal genera, which are distinguished by a process of formalization—a removal of content. These ‘categorial essences’ begin with ‘object in general’ at the top of the tree, which is then divided at the next level into categories including (as examples) object, state of affairs, property, relation, number, etc. (compare lists 1913/2000, 237 and 1913/1962, 61). Much as Aristotle distinguished (independent) primary substances from (dependent) things of other kinds, within his formal categories Husserl distinguishes the ‘substrative’ category of individuals (or, more properly, the mere this-there) from the dependent ‘syntactic objectivities’—the correlates of nominative terms that are derived from ways in which we speak about the primary substances (1913/1962, 62–3 and 67) (as, e.g., the nominative term ‘this relation of brightness’ may be derived from claims that ‘A is brighter than B’ (1913/2000, 797–8)).

Husserl's material categories, by contrast, classify entities according to their nature or essence, with the highest material genera to be arrived at by a process of generalization to the most general kind of content involved, rather than by the formalization that involves an emptying of all content (1913/1962, 65). The highest material categories are the three regions: nature (including physical objects and events), culture (including artifacts, social entities, and values), and consciousness (cf. Smith 2004). While formal and material category systems each form a hierarchy (1913/1962, 64), considered jointly their categories are not mutually exclusive, since one and the same entity may be categorized either in terms of its material nature or its form. For further discussion of Husserl's categories, see Smith (2007, 135–161).

Husserl is nowhere explicit about the proper method for distinguishing material ontological categories, but he does distinguish material absurdity from formal absurdity, and from the formal nonsense that marks the difference in meaning categories (1913/2000, 523). Expressions are formally absurd if it is a priori that no object could correspond to them, based purely on formal, logical laws, without regard to which particular material concepts are employed, e.g., “a round not-round thing” is formally absurd; its absurdity would remain regardless of which adjective we substituted for ‘round’ or which noun for ‘thing’. On the other hand, expressions are materially absurd if the impossibility of there being any corresponding object is based in the particular material concepts employed, e.g., ‘a round square’ is a materially absurd expression based in the particular meanings of ‘round’ and ‘square’. Thus presumably one could attempt to distinguish material ontological categories by the material absurdity that results from substituting expressions for objects of different material kinds; ‘a round table’, for example, makes perfect sense, but if we substitute for ‘table’ a term for a geometric figure such as ‘square’ or for a day of the week such as ‘Thursday’, we get a materially absurd statement (to which it is a priori that nothing corresponds). As we will see in §2.2 below, Gilbert Ryle developed Husserl's nonsense test for category differences in something like this way.

Roman Ingarden (1960/1964, 22ff) took Husserl's multi-dimensional ontology one step further. Like Husserl, he distinguished formal categories from material categories, but he also distinguished categories in a third dimension: existential categories (those describing an entity’s mode of being). The highest existential categories on Ingarden’s list are the real (spatio-temporal being), the ideal (abstract), the absolute (completely independent, atemporal), and purely intentional (consciousness-dependent). While any conceivable entity should be uniquely locatable in a single category of each dimension, the three sorts of ontology are mutually orthogonal, providing different most abstract ways of considering the putative entity in question. Thus, e.g., a sculpture might be categorized formally as an object, materially as a work of art, and existentially as purely intentional.

1.4 Contemporary Category Systems

By the mid twentieth century, systems of ontological categories had fallen somewhat out of fashion (for reasons I will discuss in §1.5 below), with most discussion of categories shifting to merely articulating category differences rather than outlining a comprehensive system of categories. Nonetheless, in recent years there have been several notable attempts to offer new systems of categories in either the realist or descriptivist spirit, although little agreement exists about what the categories are or how one could decide among competing systems.

Ingvar Johansson (1989) and Roderick Chisholm (1996) both take a neo-Aristotelian realist approach to categories, attempting to lay out a complete system of the categories, where this is understood as providing a list of categories of real entities in the world. Ingvar Johansson explicitly insists that his interest is in the world: “This book is a book about the world. I am concerned with ontology, not merely with language” (1989, 1), and attempts to offer “a realist theory of categories regarded as real aspects of being” (1989, 2). His list (1989, 20) includes nine main categories (some of which subdivide further):

• Space-time
• State of affairs
• Quality
• Substance
• Property
• External Relation
• Grounded Relation
• Inertia
• Spontaneity
• Tendency
• Intentionality
• Real
• Presentational
• Representational
• Fictional

Unlike Aristotle, Johansson makes no explicit use of language in discerning ontological categories, instead appealing to the method of successive abstraction (Johansson 1989, 1–2). Thus, e.g., we arrive at the category ‘quality’ by moving up in abstraction from a particular shade of dark red, to red, color, and finally quality. Similarly (to use an example of Sellars') one might try to arrive at the category of ‘substance’ by considering an individual entity, say, Fido, and moving by successive abstraction from “Fido is a dachshund”  through “Fido is a dog” and “Fido is an animal”, ultimately reaching “Fido is a substance” (1970/1974, 321). Like Aristotle's categories, Johansson's categories top out with a number of distinctions without a highest single category subsuming them all.

Like Aristotle and Johansson, Roderick Chisholm presents his work on categories as being “about the ultimate categories of reality” (1996, 3). Unlike them, however, Chisholm (1996, 3) lays out categories in the form of a porphyrian tree starting from a single most general category comprising everything, but divided into successively narrower genera at lower levels of branching. (For an interesting discussion of whether such general terms as ‘entity’ or ‘thing’  could be seen as naming a highest category, see Thompson 1957, cf. §2.3 below). Chisholm's system of categories thus reads:

• Entia
• Contingent
• States
• Events
• Individuals
• Boundaries
• Substances
• Necessary
• States
• Nonstates
• Attributes
• Substance

Other contemporary authors have approached the issue of categories in a purely descriptive spirit. Reinhardt Grossman, for example, distinguishes eight highest categories (1983, xvi):

• Individuals
• Properties
• Relations
• Classes
• Structures
• Quantifiers
• Facts
• Negation

But although Grossman characterizes his book as an attempt to “bring Aristotle's Categories up-to-date” (1983, xv), he is explicit in denying that he is making any claims about whether or not there are things belonging to any of the eight categories he distinguishes, taking this as beyond the scope of ontology (1983, 10–12).

Joshua Hoffman and Gary Rosenkrantz (1994) lay out a tree-form system of categories, with entity the summum genus, subdivided into abstract and concrete (rather than Chisholm's contingent and necessary), each of which is further subdivided:

• Entity
• Abstract
• Property
• Relation
• Proposition
• Concrete
• Event
• Time
• Place
• Substance
• Material Object
• Spirit
• Limit
• Collection
• Privation
• Trope

They, too, explicitly offer their system of categories in the spirit of categorial description, as offering an analysis of the various possible categories of being, rather than making any claims about which of these categories is non-empty (1994, 7–8).

E. J. Lowe takes categories to be categories of “what kinds of things can exist and coexist” (2006, 5). Such categories, he argues, are to be individuated according to the existence and/or identity conditions of their members; fundamental categories are those regarding which the existence and identity conditions for their members cannot be exhaustively specified in terms of ontological dependence relations they bear to entities of other categories (2006, 8). Accordingly, he argues that there are four fundamental ontological categories: objects (individual substances such as Fido), modes (property or relation instances such as Fido's four-leggedness), kinds (substantial universals such as the kind dog), and attributes (property or relation universals, such as being four-legged). But although he argues that there are exactly four fundamental categories, Lowe nonetheless takes a hierarchical approach to arranging categories. The four fundamental categories appear at the third level of his hierarchical chart; the categories that appear at the higher levels (particulars and universals at the second level; entity at the top) are “mere abstractions and do no serious ontological work on their own account” (2006, 39). His fuller chart of categories appears as follows:

• Entities
• Particulars
• Objects
• Substances
• Non-substances
• Modes (monadic and relational)
• Universals
• Kinds
• Attributes (properties and relations)

1.5 Skepticism about Category Systems

Both realist and descriptivist category systems, at least as traditionally presented, seem to presuppose that there is a unique true answer to the question of what categories of entity there are—indeed the discovery of this answer is the goal of most such inquiries into ontological categories. Grossman, for example, argues that a list of categories must be complete, contain everything, with everything in its proper place (1983, 4). Johansson similarly takes his project as to “develop a coherent system of all the most abstract categories needed to give a true description of the world” (1989, 1). Arguments about which of the many systems of categories offered is correct likewise seem to presuppose that there is a uniquely correct list of categories.

But actual category systems offered vary so much that even a short survey of past category systems like that above can undermine the belief that such a unique, true and complete system of categories may be found. Given such a diversity of answers to the question of what the ontological categories are, by what criteria could we possibly choose among them to determine which is uniquely correct?

Some minimal standards of adequacy immediately suggest themselves (Butchvarov 1995, 75). Whether one takes a realist or descriptivist approach to providing a system of categories, if that system is supposed to be comprehensive, it clearly must meet at least the standard of being exhaustive—providing a category for everything there is (on the realist approach) or might be (on the descriptive approach). Nonetheless, one may, as Hoffman and Rosenkrantz (1994, 140) do, present a system of some fundamental categories without taking it to be exhaustive.

Another minimal criterion of adequacy is generally taken to be that the highest categories (or, for tree systems, the categories at each level of branching) be mutually exclusive, ensuring that whatever there is (or might be) finds its place in exactly one highest category, or one category at each level (Chisholm 1989, 162). (This still allows for nested categories, so that something may belong to both a more specific category like substance and a more general category like individual.)

But these criteria are not enough to provide the needed reassurance. First, we lack assurance that most proposed category systems meet even these minimal conditions. As mentioned above, Aristotle drew out his categories largely by considering the types of question that could be asked (and the types of answer appropriate to them). It is difficult to know, however, how one can be assured that all kinds of questions have been surveyed, and so difficult to know that an exhaustive list of categories has been offered—a point Aristotle does not attempt to demonstrate (Ackrill 1963, 80–81). Indeed, the fact that Aristotle provides different lists of categories in different places suggests that he did not consider his list final and exhaustive. Similarly, Kant’s system of categories can be thought to be exhaustive only as long as the list of forms of judgment from which he derives them exhausts the possible forms of judgment—but we have reason to think this is not so (Körner 1955, 50). Johansson, as we have seen, instead uses the method of successive abstraction. But it is not clear how following such a method could ensure either that the categories thereby distinguished are exhaustive (how do we know we have considered something of each highest kind if we do not yet know what the highest kinds are?) or even mutually exclusive.

Secondly, even if we can verify that the standards of mutual exclusivity and exhaustiveness are met, these conditions alone are far too weak to uniquely pick out a system of categories. Provided one accepts the law of the excluded middle, an endless supply of mutually exclusive and exhaustive classifications can be generated at will: we can divide things into the spatio-temporally located and the not-spatio-temporally-located, the intentional and the non-intentional, the extended and the non-extended, to name but a few of the more relevant ways in which things could be divided. Indeed one of the sources of puzzlement about categories comes from the fact that philosophers have selected so many different sorts of divide as the fundamental category difference—for Descartes, the extended and the thinking (unextended), for Chisholm the contingent and the necessary, for Hoffman and Rosenkrantz the concrete and the abstract, and so on. Thus another reason for skepticism about the existence of a unique set of categories comes from the fact that categories are supposed to be the most abstract genera under which things (may) fall. But from any given entity, abstraction may apparently be done in a variety of ways—even if we are careful to do so in ways that ensure mutual exclusivity and exhaustiveness.

Doubts about possibilities for discovering the one true category system have led many to eschew talk of category systems altogether, and others to adopt some kind of relativism about category systems that ceases to take systems of categories seriously as candidate lists of the single set of highest genera under which anything falls (or could fall). Jan Westerhoff (2005), for example, argues that there is no unique, absolute set of ontological categories. On his view categories in metaphysics turn out to be analogous to axioms in mathematical theories; in each case, there may be more than one way to systematize our knowledge from a relatively simple basis. The result is a kind of relativity about systems of categories: “which set of ontological categories we choose is primarily a matter of convenience, in the same way as specific axiomatizations of propositional logic or Newtonian mechanics are more convenient to use than others” (2005, 218). As a result, Westerhoff argues, we must reassess the importance of ontological categories in metaphysics--these should not be considered “the most fundamental parts of the world, but... the most fundamental parts of our systematization of the world” (2005, 135).

Others have taken the variety of category systems explicitly offered or presupposed by philosophers as mere evidence of the particular presuppositions of their thought, or prejudices of their age—not as evidence about anything to do with the world and its divisions. Thus, e.g., Stephan Körner's discussion of categorial frameworks is designed to make explicit how a thinker’s framework categorizes objects, making use of certain individuative principles, and making clear his reasons for holding that framework (1970, 10). R. G. Collingwood, in similar vein, treats the task of metaphysics generally as merely uncovering the “presuppositions underlying ordinary science” (1940/1998).

The specific worries about (1) guaranteeing the mutual exclusiveness and joint exhaustiveness of the categories, and (2) whether or not any single system of categories could purport to be uniquely correct, can, however, be met by certain ways of formulating ontological categories. The first sort of worry can be met by ensuring that categories (of the same level) are defined in ways that guarantee mutual exclusiveness and exhaustiveness. Thus, e.g., Thomasson (1999, chapter 8) distinguishes categories in terms of what relations of dependence a purported entity has or lacks on mental states (and a second dimension distinguished in terms of what relations of dependence a purported entity has or lacks on spatio-temporally located objects), so that the law of the excluded middle alone ensures mutual exclusiveness and exhaustiveness of the categories distinguished. (Dummett’s method of distinguishing categories provides another route for guaranteeing mutual exclusivity—see §2.3 below).

Multi-dimensional systems (Husserl 1913/1962, §10; Ingarden 1960/1964, Chapter 2; Thomasson 1999, Chapter 8; Smith forthcoming, Chapter 8) address the second worry to some extent by acknowledging that the different dimensions of categorization are possible, and that no one-dimensional list can purport to completeness. In principle, multi-dimensionalists may even accept that there is no fixed number or limit on how many one-dimensional lists of categories there may be, though each such list may purport to provide a unique, correct, exhaustive categorization of entities considered in the chosen respect.

In any case, given the great potential uses of a system of categories (many of which do not depend on claims that that category system is uniquely acceptable), we should not abandon attempts at systems of categorization prematurely. Even if we do not think of a category system as providing a realist inventory of all that exists, a system of categories laid out in the descriptivist spirit provides a framework within which existence questions can be answered in a systematic and wholesale way, by enumerating categories so that we can then undertake further investigations into whether or not there really is anything of each kind. Working from within a categorial framework can help ensure that whatever ontology we provide is principled and unified, avoiding ad hoc and piecemeal decisions. The descriptivist's categories also provide a tool that may be used elsewhere in ontology, e.g., in helping to ensure that comparisons of parsimony are legitimately made (by examining which categories of entity are accepted and which denied by various theories), and in checking that potential solutions to metaphysical problems are not overlooked by tacit use of a category system that is not exhaustive (Thomasson 1999, Chapters 8 and 9). Indeed assumptions about categorizations play such a strong role in philosophical discussions (e.g., discussions of the Cartesian theory of mind, Platonist theories of mathematics, etc.), that doing the work on categories necessary to make these categorial assumptions explicit and open them for examination must remain a highly useful exercise regardless of doubts about the prospects for discovering a uniquely correct system of categories.

1.6 Categories in Other Disciplines

Recently, work on ontological categories has attracted interest not only among philosophers, but also in information science and the biomedical sciences, where ontologies are used to organize the knowledge represented in information systems (Smith 2003). In some cases, the ontologies developed are domain-specific (e.g. specific to medical information, geographic information, etc.), but there has also been a great deal of interest in developing a ‘top-level’ ontology of maximally general categories applicable to all specific domains; it is such top-level ontologies that draw upon philosophical work on ontological categories most directly, although categorial distinctions also play a crucial role in domain-specific ontologies.

Discussion of categories also plays an important role in cognitive science, where the goal is not to discover the fundamental categories of being, but rather the means by which experiencers come to categorize their world. Here, debates have centered on how humans in fact come to group things into categories--whether this involves lists of definitional (observable or hidden) features, resemblance to prototypes, prominent features weighted probabalistically, etc. Debates also concern the relation between conceptual and linguistic categories, which levels of category are more basic, and whether or to what extent categorizations are consistent across cultural groups. For further discussion of the debates about categorization in cognitive science see Lakoff (1987) and Rakison and Oakes (2003). Whether information from cognitive science about our formation of conceptual and linguistic categories has any import for our understanding of what ontological categories there are is itself, of course, a contentious issue (to which I will return briefly in section 2.3 below).

2. Category Differences

Recent work on categories has been strongly directed by reaction to skepticism about the possibility of offering a system of ontological categories. On the one hand, difficulties like those mentioned above have undermined the idea that a uniquely true and comprehensive system of categories could be developed. As a result, while categories have continued to play a central role in analytic philosophy in the past century, the focus has shifted to articulating particular category differences, without attempting to provide an exhaustive inventory of metaphysical categories or even presupposing that such a list is possible.

On the other hand, an even more influential source of skepticism about the possibility of laying out a system of ontological categories in the realist spirit derives from the general positivist rejection of metaphysics: if all metaphysical talk is nonsense, then talk about what sorts of thing there really are in the world is just part of the general rubbish constituting metaphysics, and debates about, e.g., whether or not substance is a category of being, whether there are ten or twelve or twenty-seven top level categories, whether or not there is a single highest category, etc. are all meaningless. Moreover, even for philosophers who regained comfort with certain forms of metaphysical talk, naturalistic concerns about ontological commitment to such abstract entities as qualities, relations, etc., have made many shy away from providing a realist list of categories that would include any such things (Sellars 1970/1974, 322–3).

In response to such positivist and naturalistic qualms, talk about categories over the past century has also tended to shift to identifying differences in semantic categories rather than drawing out systems of ontological categories. Thus when Gilbert Ryle (1949, 1938/1971) talks of categories, he does not speak directly of categories of entities, but rather of differing logical types of concepts, where such type differences are detectable by the absurdities that result from substituting in terms of one sort for terms of another in sentences of certain kinds (see §2.2 below). Wilfrid Sellars, developing a strategy of Ockham’s, argues explicitly that we may construe category statements as disguised metalinguistic statements about the role of certain expressions (and their functional counterparts in other languages). According to Sellars, “Socrates is a substance”, for example, has the sense of “The ·Socrates· is a basic mental singular term”, and “Yellow is a quality” has the sense of “The ·yellow· is a (one-place) predicate (in mentalese)” (1970/1974, 328) (where the “·___·” notation has the function of enabling us to speak about linguistic roles without being tied to a particular natural language). As a result we can replicate the work done by traditional category distinctions between, e.g., substance and quality, without committing ourselves ontologically to the existence of qualities or other abstracta (1970/1974, 329).