The distinction between atype and itstokens is auseful metaphysical distinction. In §1 it is explained what itis, and what it is not. Its importance and wide applicability inlinguistics, philosophy, science and everyday life are brieflysurveyed in §2. Whether types are universals is discussed in§3. §4 discusses some other suggestions for what types are,both generally and specifically. Is a type the sets of its tokens?What exactly is a word, a symphony, a species? §5 asks what atoken is. §6 considers the relation between types and theirtokens. Do the type and all its tokens share the same properties? Mustall the tokens be alike in some or all respects? §7 explains someproblems for the view that types exist, and some problems for the viewthat they don't. §8 elucidates a distinction often confused withthe type-token distinction, that between a type (or token) and anoccurrence of it. It also discusses some problems thatoccurrences might be thought to give rise to, and one way to resolvethem.

• 1. The Distinction Between Types and Tokens
  • 1.1 What the Distinction Is
  • 1.2 What It Is Not
• 2. Importance and Applicability of the Distinction
  • 2.1 Linguistics
  • 2.2 Philosophy
  • 2.3 Science and Everyday Discourse
• 3. Types and Universals
• 4. What is a Type?
  • 4.1 Some General Answers
    • 4.1.1 A Set
    • 4.1.2 A Kind
    • 4.1.3 A Law
  • 4.2 What is a Word?
    • 4.2.1 Identity Conditions for Words
    • 4.2.2 What a Theory of Words Might Tell Us
    • 4.2.3 Orthography
    • 4.2.4 Phonology
    • 4.2.5 Conclusion
• 5. The Relation between Types and their Tokens
• 6. What is a Token?
• 7. Do Types Exist?
  • 7.1 Universals
  • 7.2 Realism
  • 7.3 Nominalism
    • 7.3.1 Goodman's Nominalism
    • 7.3.2 Sellars's nominalism
• 8. Occurrences
• Bibliography
• Academic Tools
• Other Internet Resources
• Related Entries

1. The Distinction Between Types and Tokens

1.1 What the Distinction Is

The distinction between atype and itstokens isan ontological one between a general sort of thing and its particularconcrete instances (to put it in an intuitive and preliminary way). Sofor example consider the number of words in the Gertrude Stein linefrom her poemSacred Emily on the page in front of thereader's eyes:

Rose is a rose is a rose is a rose.

In one sense of ‘word’ we may count three differentwords; in another sense we may count ten different words. C. S. Peirce(1931-58, sec. 4.537) called words in the first sense“types” and words in the second sense “tokens”.Types are generally said to be abstract and unique; tokens are concreteparticulars, composed of ink, pixels of light (or the suitablycircumscribed lack thereof) on a computer screen, electronic strings ofdots and dashes, smoke signals, hand signals, sound waves, etc. A studyof the ratio of written types to spoken types found that there aretwice as many word types in written Swedish as in spoken Swedish(Allwood, 1998). If a pediatrician asks how many words the toddler hasuttered and is told “three hundred”, she might well enquire“word types or word tokens?” because the former answerindicates a prodigy. A headline that reads “From the Andes toEpcot, the Adventures of an 8,000 year old Bean” might elicit“Is that a bean type or a bean token?”.

1.2 What It Is Not

Although the matter is discussed more fully in §8 below, it shouldbe mentioned here at the outset that the type-token distinction is notthe same distinction as that between a type and (what logicians call)itsoccurrences. Unfortunately, tokens are often explained asthe “occurrences” of types, but not all occurrences oftypes are tokens. To see why, consider this time how many words thereare in the Gertrude Stein line itself,the line type, not atoken copy of it. Again, the correct answer is either three or ten, butthis time it cannot be ten wordtokens. The line is anabstract type with no unique spatio-temporal location and thereforecannot consist of particulars, of tokens. But as there are only threeword types of which it might consist, what then are we counting ten of?The most apt answer is that (following logicians' usage) it iscomposed of tenoccurrences of word types. See §8 below,Occurrences, for more details.

2. Importance and Applicability of the Distinction

2.1 Linguistics

It is generally accepted that linguists are interested in types.Some, e.g. Lyons (1977, p. 28), claim the linguist is interested onlyin types. Whether this is so or not, linguists certainly appear to beheavily committed to types; they “talk as though” thereare types. That is, they often quantify over types in their theoriesand refer to them by means of singular terms. As Quine has emphasized,a theory is committed to the existence of entities of a given sort ifand only if they must be counted among the values of the variables inorder for the statements in the theory to be true. Linguistics is rifewith such quantifications. For example, we are told that an educatedperson's vocabulary is about 15,000words, but Shakespeare'swas more nearly 30,000. These are types, as are the twenty-sixletters of the English alphabet, and its eighteen cardinalvowels. (Obviously the numbers would be much larger if wewere counting tokens). Linguists also frequently refer to types bymeans of singular terms. According to theO.E.D., forexample,the noun ‘color’’ is from earlymodern English and in addition to its pronunciation [kɒ’lər] has two "modern current or most usual spellings" [colour,color], eighteen earlier spellings [collor, collour, coloure, colowr,colowre, colur, colure, cooler, couler, coullor, coullour, coolore,coulor, coulore, coulour, culler, cullor, cullour] and eighteen senses(vol. 2, p. 636). According toWebster's,the word‘schedule’ has four current pronunciations:['ske-(,)jü(ə)l], ['ske-jəl](US),['she-jəl] (Can) and ['she-(,)dyü(ə)l](Brit) (p. 1044). Thus, linguistics is apparently committedto the existence of these words, which are types.

References to types is not limited to letters, vowels and words, butoccur extensively in all branches of linguistics. Lexicographydiscusses, in terms that make it clear that it is types being referredto, nouns, verbs, words, their stems, definitions, forms,pronunciations and origins. Phonetics is committed to consonants,syllables, words and sound segments, the human vocal tract and itsparts (the tongue has five). Phonology also organizes sounds but interms of phonemes, allophones, alternations, utterances, phonologicalrepresentations, underlying forms, syllables, words, stress-groups,feet and tone groups. Morphology is apparently committed to morphemes,roots, affixes, and so forth, and syntax to sentences, semanticrepresentations, LF representations, among other things. Clearly, justas words and letters and vowels have tokens, so do all of the otheritems mentioned (nouns, pronunciations, syllables, tone groups and soforth). It is more controversial whether the items studied insemantics (the meanings of signs, their sense relations, etc.) alsocome in types and tokens, and similarly for pragmatics (includingspeaker meanings, sentence meanings, implicatures, presuppositions,etc.) It seems to hinge on whether a mental event (token) or part ofit could be a meaning, a matter that cannot be gone into here. SeeDavis (2003) for a view according to which concepts andthoughts—varieties of meaning—come in types andtokens.

It is notable that when one of the above types is defined, it isdefined in terms of other types. So for example, sentences might be(partly) defined in terms of words, and words in terms of phonemes.

The universal and largely unscrutinized reliance of linguistics on thetype-token relationship and related distinctions like that oflangue toparole, andcompetence toperformance, is the subject of Hutton's cautionary book(1990).

2.2 Philosophy

Obviously then, types play an important role in philosophy oflanguage, linguistics and, with its emphasis on expressions, logic.Especially noteworthy is the debate concerning the relation between themeaning of a sentence type and the speaker's meaning in using atoken (a relation that figures prominently in Grice 1969). But thetype-token distinction also functions significantly in other branchesof philosophy as well. Inphilosophy of mind, it yields twoversions of the identity theory of mind (each of which is criticized inKripke 1972). The type version of the identity theory (defended bySmart (1959) and Place (1956) among others) identifiestypesof mental events/states/processes withtypes of physicalevents/states/processes. It says that just as lightning turned out tobe electrical discharge, so pain might turn out to be c-fiberstimulation, and consciousness might turn out to be brain waves of 40cycles per second. On this type view, thinking and feeling are certaintypes of neurological processes, so absent those processes, there canbe no thinking. The token identity theory (defended by Kim (1966) andDavidson (1980) among others) maintains that every token mental eventis some token physical event or other, but it denies that a typematch-up must be expected. So for example, even if pain in humans turnsout to be c-fiber stimulation, there may be other life forms that lackc-fibers but have pains too. And even if consciousness in humans turnsout to be a brain waves that occur 40 times per second, perhapsandroids have consciousness even if they lack such brain waves.

Inaesthetics, it is generally necessary to distinguishworks of art themselves (types) from their physical incarnations(tokens). (See, for example, Wollheim 1968, Wolterstorff 1980 andDavies 2001.) This is not the case with respect to oil paintings liketheMona Lisa where there is and perhaps can be only onetoken, but seems to be the case for many other works of art. There canbe more than one token of a sculpture made from a mold, more than oneelegant building made from a blueprint, more than one copy of a film,and more than one performance of a musical work. Beethoven wrote ninesymphonies, but although he conducted the first performance ofSymphony No. 9, he never heard the Ninth, whereas the rest ofus have all heard it, that is, we have all heard tokens of it.

Inethics, actions are said to be right/wrong—but is itaction types or only action tokens? There is a dispute about this. Mostethicists from Mill (1979) to Ross (1988) hold that the hallmark ofethical conduct is universalizability, so that a particular action isright/wrong only if it is right/wrong for anyone else in similarcircumstances—in other words, only if it is the right/wrongtype of action. If certain types of actions are right andothers wrong, then there may be general indefeasible ethical principles(however complicated they may be to state, and whether they can bestated at all). But some ethicists hold that there are no generalethical principles that hold come what may—that there is always somecircumstance in which such principles would prescribe the wrongoutcome—and such ethicists may go on to add that only particular(token) actions are right/wrong, not types of actions. See, forexample, Murdoch 1970 and Dancy 2004.

2.3 Science and Everyday Discourse

Outside of philosophy and linguistics, scientists often quantifyover types in their theories and refer to them by means of singularterms. When, for example, we read that the “Spirit Bear” isa rare white bear that lives in rain forests along the British Columbiacoast, we know that no particular bear israre, but rather atype of bear. When we are told that these Kermode bears “have amutation in the gene for the melanocortin 1 receptor” (TheWashington Post 9/24/01 A16) we know that it is not a tokenmutation, token gene and token receptor being referred to, but a type.It is even more evident that a type is being referred to when it isclaimed that “all men carry the same Y chromosome…. Thisone and only Y has the same sequence of DNA units in every man aliveexcept for the occasional mutation that has cropped up every thousandyears” (The New York Times, Nicholas Wade 5/27/03).Similarly, to say the ivory-billed woodpecker is not extinct is to bereferring to a type, a species. (The status of species will bediscussed in greater detail in §4 below.)

The preceding paragraph contains singular terms that (apparently)refer to types. An even more telling commitment to types are thefrequent quantifications over them. Mayr (1970, p. 233), for example,tells us that “there are about 28,500 subspecies of birds in atotal of 8,600 species, an average of 3.3 subspecies perspecies…79 species of swallows have an average of 2.6subspecies, while 70 species of cuckoo shrikes average 4.6subspecies”. Although these examples come from biology, physics(or any other science) would provide many examples too. It was claimed(in the sixties), for example, that “there are thirty particles,yet all but the electron, neutrino, photon, graviton, and proton areunstable.” Artifactual types (the Volvo 850 GLT, the DellLatitude D610 laptop) easily lend themselves to reference also. Inchess we are told that accepting the Queen's Gambit with 2…dchas been known since 1512, but Black must be careful in thisopening—the pawn snatch is too risky. Type-talk is ubiquitous.

3. Types and Universals

Are types universals? They have usually been so conceived, and withgood reason. But the matter is controversial. It depends in part onwhat a universal is. (See the entry onproperties.) Universals, in contrast to particulars, have been characterized ashaving instances, being repeatable, being abstract, being acausal,lacking a spatio-temporal location andbeing predicable ofthings. Whether universals have all these characteristics cannotbe resolved here. The point is that types seem to have some, but notall, of these characteristics. As should be clear from the precedingdiscussion, types have or are capable of having instances, of beingexemplified; they are repeatable. To many, this is enough to count asuniversals. With respect to being abstract and lacking aspatio-temporal location, types are also akin to universals—thatis, they are if universals are. On certain views of types anduniversals, types, unlike their instances, are abstract and lack aspatio-temporal location. On other views, types and universals arein their instances and hence are neither abstract noracausal; far from lacking a spatio-temporal location, they usuallyhave many. (For more details, see §5 below,The Relationbetween Types and Tokens.) So far, then, types appear to be aspecies of universal, and most metaphysicians would so classifythem. (Although a few would not. Zemach (1992), for example, holdsthat there are no universals, but there are types, which arerepeatable particulars—the cat may be in many differentplaces at the same time.)

When it comes tobeing predicable, however, most typesdiverge from such classic examples of universals as the property ofbeing white or the relation ofbeing east of. Theyseem not to be predicable, or at least not as obviously so as theclassic examples of universals. That is, if the hallmark of auniversal is to answer to a predicate or open sentence such asbeing white answers to ‘is white’, then mosttypes do not resemble universals, as they more readily answer tosingular terms. This is amply illustrated by the type talk exhibitedin §2 above. It is also underscored by the observation that it ismore natural to say of a token of a word—‘infinity’,say—that it is a token of the word ‘infinity’ thanthat it is an ‘infinity’. That is to say, types seem tobeobjects, like numbers and sets, rather than properties orrelations; it's just that they are not concrete particulars butare general objects—abstract objects, on some views. If, then,we follow Gottlob Frege (1977) in classifying all objects as being thesort of thing referred to by singular terms, and all properties as thesort of thing referred to by predicates, then types would beobjects. Hence they would not fall into the same category as theclassic examples of universals such asbeing white andbeing east of, and thus perhaps should not be considereduniversals at all. (Although perhaps all this shows is that they arenot akin to properties, but are their own kind of universal.) Ageneral exception to what has just been claimed about how we refer totypes (with singular terms) might be inferred from the fact that we domore often say of an animal that it is a tiger, rather than that it isa member of the speciesFelis Tigris. This raises thequestion as to whether the speciesFelis Tigris is just theproperty ofbeing a tiger, and if it isn't, then whatthe relation between these two items is.

Wollheim (1968, p. 76) insightfully puts the point that types seemto be objects as that the relationship between a type and its tokens is“more intimate” than that between (a classic example of) aproperty and its instances because “for much of the time we thinkand talk of the type as though it were itself a kind of token, though apeculiarly important or pre-eminent one”. He (1968, p. 77)mentions two other differences worth noting between types and theclassic examples of universals. One is that although types and theclassic examples of properties often satisfy the same predicates, thereare many more predicates shared between a type and its tokens thanbetween a classic example of a property and its instances.(Beethoven'sSymphony No. 9 is in the same key, has thesame number of measures, same number of notes, etc. as a great many ofits tokens.) Second, he argues that predicates true of tokens in virtueof being tokens of the type are therefore true of the type (OldGlory is rectangular) but this is never the case with classicproperties (being white is not white.)

These considerations may not suffice to show that types aren'tuniversals, but they do point to a difference between types and theclassic examples of properties.

4. What is a Type?

The question permits answers at varying levels of generality. At itsmost general, it is a request for a theory of types, the way settheory answers the question “what is a set?” A generalanswer would tell us what sort of thing a type—anytype—is. For example, is itsui generis, or auniversal, or perhaps the set of its tokens, or a function from worldsto sets, or a kind, or, as Peirce maintained, a law? These options arediscussed in §4.1. At a more specific level, “what is atype?” is a request for a theory that would shed some light onthe identity conditions for specific types of types, not necessarilyall of them. It would yield an account of what a word (or a symphony,a species, a disease, etc.) is. This is in many ways a more difficultthing to do. To see just how difficult it is to give the identityconditions for an individual type, §4.2 considers what a word is,both becausewords are our paradigm of types, since thetype-token distinction is generally illustrated by means of words, andbecause doing so will show that some of the most common assumptionsmade about types and their tokens are not correct. It will alsoilluminate some of the things we want from a theory of types.

4.1 Some General Answers

As mentioned in the previous paragraph, one way a theory of typesmight answer the question “what is a type?” is the way settheory answers the question “what is a set?” If typesare universals, as most thinkers assume, then there are asmany theories of types as there are theories of universals. Someaxiomatic theories include Zalta 1983 and Jubien 1988. Since theoriesof universals are discussed at length in this Encyclopedia elsewhere,they will not be repeated here. (Seeproperties.) However it might be said that types arenot universals forthe reasons mentioned in §3 above, where it was urged that typesmight be neither properties nor relations but objects, and there is anabsolute difference between objects and properties. Identifying typesas universals would appear to fly in the face of thatconsideration.

4.1.1 A Set

It might appear that types are better construed as sets (assuming setsthemselves are not universals). The natural thought is that a type isthe set of its tokens, which is how Quine sometimes (1987, p. 218)construes a type. After all, a species is often said to be “theclass of its members”. There are two serious problems with thisconstrual. One is that many types have no tokens and yet they aredifferent types. For example, there are a lot of very long sentencesthat have no tokens. So if a type were just the set of its tokens,these distinct sentences would be wrongly classified as identical,because each would be identical to the null set. Another closelyrelated problem also stems from the fact that sets, or classes, aredefined extensionally, in terms of their members. The set of naturalnumbers without the number 17 is a distinct set from the set ofnatural numbers. One way to put this is that classes have theirmembersessentially. Not so the specieshomosapiens, the word ‘the’, nor Beethoven'sSymphonyNo. 9. The set of specimens ofhomo sapiens withoutGeorge W. Bush is a different set from the set of specimens ofhomo sapiens with him, but the species would be the same evenif George W. Bush did not exist. That is, it is false that had GeorgeW. Bush never existed, the specieshomo sapiens would nothave existed. The same species might have had different members; itdoes not depend for its existence on the existence of all its membersas sets do.

Better, then, but still in keeping with an extensional set-theoreticapproach, would be to identify a type as a function from worlds to setsof objects in that world. It is difficult to see any motivation forthis move that would not also motivate identifying properties as suchfunctions and then we are left with the question of whether types areuniversals, discussed in §3.

4.1.2 A Kind

The example ofhomo sapiens suggests that perhaps a type isakind, where a kind is not a set (for the reasons mentionedtwo paragraphs above). Of course, this raises the question of what akind is; Wolterstorff (1970) adopts thekind view of types andidentifies kinds as universals. In Wolterstorff 1980, he takesbeing an example of as undefined and uses it to definekinds—so that, for example, a possible kind is one such that it ispossible there is an example of it.Norm kinds he then definesas kinds “such that it is possible for them to have properlyformed and also possible for them to have improperly formedexamples” (p. 56). He identifies both species and artworks asnorm-kinds. Bromberger (1992a) also views the tokens of a type as aquasi-natural kind relative to appropriate projectible (“What isits freezing point?” e.g.) and individuating questions(“Where was it on June 13^th, 2005?”). However,he doesn't identify the type as the kind itself, since to do sodoes not do justice to the semantic facts mentioned in §2 above, thattypes are largely referred to by singular terms. Instead he views thetype as what he calls thearchetype of the kind, defined assomething that models all the tokens of a kind with respect toprojectible questions but not something that admits of answers toindividuating questions. Thus for Bromberger the type is not the kinditself, but models all the tokens of the kind. We shall see somedifficulties for this view in §5 below.

4.1.3 A Law

It wouldn't do to ignore what the coiner of the type-tokendistinction had to say about types. Unfortunately it cannot beadequately unpacked without an in-depth explication of Peirce'ssemiotics, which cannot be embarked upon here. (See the entries onCharles Sanders Peirce andPeirce's theory of signs.) Peirce said types “do not exist”, yet they are“definitely Significant Forms” that “determinethings that do exist” (4.423). A type, or “legisign”as he also calls it, “has a definite identity, though usuallyadmitting a great variety of appearances. Thus, &and andthe sound are all one word” (8.334). Elsewhere he tells us thata type is “a general law that is a sign. This law is usuallyestablished by men. Every conventional sign is a legisign. It is not asingle object, but a general type which…shall besignificant. …[E]very Legisign requires Sinsigns”(2.246). Sinsigns are tokens. (It should be mentioned that for Peircethere is actually a trichotomy among types, tokens andtones,or qualisigns, which are “the mere quality of appearance”(8.334).) Thus types have a definite identity as signs, are generallaws established by men, but they do not exist. Perhaps all Peircemeant by saying they do not exist was that they are “notindividual things” (6.334), that they are, instead what he calls“generals”—not to be confused with universals. Whathe might have meant by a “general law” isuncertain. Stebbing (1935, p.15) suggests “a rule in accordancewith which tokens … could be so used as to have meaningattached to them”. Greenlee (1973, p. 137) suggests that forPeirce a type is "a habit controlling a specific way of respondinginterpretatively." Perhaps, then, types are of a psychological nature.Obviously two people can have the same habit, sohabits alsocome in types and tokens. Presumably, types are then habit types. Thisaccount may be plausible for words, but it is not plausible forsentences, because there are sentences that have no tokens because ifΦ,Ψ are sentences, then so is (Φ & Ψ) and it isclear that for Peirce "every [type] requires [tokens]" (2.246). And itis much less plausible for non-linguistic types, like types ofbeetles, some of which have yet to be discovered.

4.2 What is a Word?

No general theory of types can tell us what we often want to knowwhen we ask: what is a species, a symphony, a word, a poem, or adisease? Such questions are just as difficult to answer as what a typeis in general. Even if types were sets, for example, set theory byitself will not answer the burning question of which set a species is.One would then have to go to biology and philosophy of biology to findout whether a species is (i) “a set of individuals closelyresembling each other” as Darwin (1859, p. 52) would have it, (ii)a set of “individuals sharing the same genetic code” asPutnam (1975, p. 239) would have it, (iii) a set of “interbreedingnatural populations that are reproductively isolated from other suchgroups”, as Mayr (1970, p. 12) would have it, or (iv) a setcomprising “a lineage evolving separately from others and withits own unitary evolutionary role and tendencies”, as Simpson(1961, p. 153) would have it. Similarly, if there is a question ofcopyright infringement, one had best look to industry standards andaesthetics for what a film or a song is, and not set theory. Ingeneral, questions such as “what is a poem, a phoneme, a disease,a flag,….?” are to be pursued in conjunction with aspecific discipline, and not within the confines of a general theory oftypes. It is largely up to linguistics and the philosophy of it, e.g.,to determine the identity conditions for phonemes, allophones, cardinalvowels, LF representations, tone groups and all the other linguistictypes mentioned in §2 above.

4.2.1 Identity Conditions for Words

It's instructive to consider what our paradigm of a type is—aword. It will reveal how complicated the identity conditions are foreven one specific type, and help to dispel the idea that tokens are totypes as cookies are to cookie cutters. It will also show what wedesire from a theory of types, by exhibiting the facts that a theoryof types has to accommodate. We illustrated the type-token distinctionby appealing towords, so presumably we think we know atleast roughly what a word type is. Unfortunately, everyone seems tothink they know, but there is massive disagreement on the matter inphilosophy. However, as urged in the preceding paragraph, it iscrucial to rely on linguistics when we consider what a word is. Whenwe do we find that there are different linguistic criteria for what aword is, and a good deal of the disagreement can be chalked up to thisfact. McArthur 1992'sThe Oxford Companion to the EnglishLanguage (pp. 1120-1) lists eight: orthographic, phonological,morphological, lexical, grammatical, onomastic, lexicographical andstatistical— but adds that more can be demarcated.There aredifferent types of types of words. However deserving of attentioneach of these is, it will be useful to focus on just one, and I willdo so in what follows. There is an important and very common use ofthe word ‘word’ that lexicographers and the rest of us usefrequently. It is, roughly,the sort of thing that merits adictionary entry. (Roughly, because some entries in thedictionary, e.g.,‘il-,’ ‘-ile,’ and‘metric system,’ are not words, and some words, e.g. manyproper names, do not get a dictionary entry.) This notion was at playin our opening remarks in §2 about Shakespeare's vocabularycontaining 30,000 words, and the twenty spellings and eighteen sensesof the noun ‘color’/’colour’, the verb‘color’/’colour’, and four currentpronunciations of the noun ‘schedule’. These examples show(in this ordinary sense of ‘word’) that the same word canbe written or spoken, can have more than one correct spelling, canhave more than one correct spelling at the same time, can have morethan one sense at the same time and can have more than one correctpronunciation at the same time. It also shows that different words canhave the same correct spelling and pronunciation; further obviousexamples would show that different words can have the samesense—e.g. English ‘red’ and French‘rouge’. A theory of types, or of word types, that cannotaccommodate this notion of a word is worthless. In what follows, Ishall use ‘word’ in this sense.

4.2.2 What a Theory of Words Might Tell Us

Ideally, a theory of words and their tokens should tell us not only(i) what a word is (in the sense indicated), but (ii) how a word is tobe individuated, (iii) whether there is anything all and only tokens of aparticular word have in common (other than being tokens of that word);(iv) how we know about words; (v) what the relation is between wordsand their tokens; (vi) what makes a token a token of one word ratherthan another; (vii) how word tokens are to be individuated; and (viii)what makes us think a particular is a token of one word rather thananother. These questions are distinct, although they are apt to runtogether because the answer to one may give rise to answers to others.For example, if we say in answer to (iii), that all tokens of acertain word (say, ‘cat’) have something in common besidesbeing tokens of that word—they are all spelled‘c’-‘a’-‘t’, forexample—then we may be inclined to say to (vi) thatspelling makes a word token a token of ‘cat’rather than some other type; and to (vii) that word tokens of‘cat’ are to be individuated on the basis of their beingspelled ‘c’-‘a’-‘t’; and to (viii)that we think something is a token of ‘cat’ when we seethat it is spelled ‘c’-‘a’-‘t’;and to (ii) that the word ‘cat’ itself is to beindividuated by its spelling; and to (i) that a word type is asequence of letters—e.g. the word ‘cat’ justis the sequence of letters <‘c’,‘a’, ‘t’>; and to (iv) that we know about aparticular word, about what properties it has, by perceiving itstokens: it has all the properties that every one of its tokens has(except for properties types cannot have, e.g.,beingconcrete).

4.2.3 Orthography

The advantage of starting with (iii) is that if there is somenontrivial property that all tokens of a word (in the sense indicated)have in common, then perhaps we can use it to individuate the tokens,and also to get a handle on what the type is like and on how we knowwhat the type is like. Unfortunately, it is not spelling, contrary towhat many philosophers seem to think. Stebbing, e.g., considers theinscribed word ashape. But not even the linguist's muchnarrower notion of anorthographic word (‘a visual signwith a space around it’) requires a canonical spelling. We haveseen that not all tokens of ‘color’ have the samespelling, even when they are spelled correctly, which they sometimesare not. Not all tokens are spelled at all—spoken tokens arenot. Moreover, two words can have the same spelling, as the noun‘color’ and verb ‘color’ prove, or to take adifferent example, the noun ‘down’ from German meaning“the fine soft covering of fowls” and the different noun‘down’ from Celtic meaning "open expanse of elevatedland". (They are not the same word with two senses, but differentwords with different etymologies.) Notice that even if, contrary tofact, all tokens of a word had the same spelling and we concluded thatthe word type itself just is the sequence of letter types that composeit, we would have analyzed word types in terms of letter types, butsince we are wondering what types are in the first place, we wouldstill need an account of what letters are since they are typestoo. Providing one is surprisingly difficult. Letters of the alphabetsuch as the letter ‘A’ are not justshapes, forexample, contrary to what is implicit in Stebbing 1935 and moreexplicit in Goodman and Quine 1947, because Braille and Morse codetokens of the letter ‘A’ cannot be said to have “thesame shape”, and even standard inscriptions of the letter‘A’ do not have the same shape—in either a Euclideanor a topological sense—as these examples obtained from a fewminutes in the library illustrate:

Moreover, the letter ‘A’ has a long history and many ofits earlier “forms” would not be recognizable to the modernreader.

4.2.4 Phonology

If we switch instead to aphonemic analysis of words, asbeing more fundamental, similar problems arise. Not all tokens of aword are composed of the samephonemes, because some tokensare inscriptions. But even ignoring inscriptions, the example of thetwo ‘down’’s shows that neither can we identify aword with a sequence of phonemes. This particular difficulty might beavoided if we identify a word with a phonemic analysis paired with asense. But this is too strong; we saw earlier that the noun‘color’ has eighteen senses. Moreover,‘schedule’ has more than one phonemic analysis. A phonemeitself is a type with tokens, and so we'd also need an account of whata phoneme is, and what its tokens have in common (if anything). Sayingwhat a phoneme is promises to be at least as hard as saying what aletter is. Phonology, the study of phonemes, is distinct fromphonetics, the scientific study of speech production. Phonetics isconcerned with the physical properties of sounds produced and is notlanguage relative. Phonemes, on the other hand, are language relative:two phonetically distinct speech tokens may be classified as tokens ofthe same phoneme relative to one language, or as tokens of differentphonemes relative to another language. Phonemes are theoreticalentities, and abstract ones at that: they are sometimes said to besets of features. The bottom line is that thephonological word is not the lexicographical one either.

It might be thought that we started at too abstract a level—thatif we think there is a hierarchy of types of words, we started "toohigh" on the hierarchy and we should start lower on thehierarchy. That is, that we should first gather together those tokensthatare phonetically (and perhaps semantically) identical onthe grounds that this is a perfectly good notion of aword. But notice: this would mean that different dialects ofthe same language would have far fewer "words" in common than onewould have supposed, and it would misclassify many words because, forexample, according to Fudge (1990, p. 39) a Cockney ‘know’is like the Queen's ‘now’; the Queen's ‘know’is like Scottish ‘now’; and a Yorkshire ‘know’is like the Queen's ‘gnaw’. Worse, even within the verysame idiolect it would distinguish as different "words" what one wouldhave thought were the same word. For example, the word‘extraordinary’ is variously pronounced with six, five,four, three or even two syllables by speakers of British English.According to Fudge (1990, p. 40) it ranges "for most British Englishspeakers from the hyper-careful['ekstrə'ʔɔ:dɪnərɪ] through the fairlycareful [ɪk'strɔ:dnrɪ] to the very colloquial['strɔ:nrɪ]." That is, the very same person may use any offive pronunciations for what should be considered the same word. Onlyan absolute diehard of this "bottom-up" approach would insist ondistinguishing as different words representations for the sameidiolectal word. Not only would a phonologist take this as excessivelycomplicated, but the representation types themselves can receiverealizations that are acoustically very different (for the small childand the man may speak the same idiolect). Fudge (1990, p. 31) assuresus that "It is very rare for two repetitions of an utterance to beexactly identical, even when spoken by the same person." Pretty soon,each word token would have to count as tokens of different"words".

The example of ['strɔ:nrι] demonstrates that there may beno phonetic signal in a token for every phoneme that issupposed to compose the word: it is "missing" several syllables! Thisis also demonstrated by reflection on casual speech: [jeet?] for‘did you eat?’. No wonder, then, that many phoneticianshave given up on the attempt to reduce phonological types toacoustic/articulatory types. (See Bromberger and Halle 1986). Even thephysicalist Bjorn Lindblom (1986, p. 495) concedes that "for a givenlanguage there seems to be no unique set of acoustic properties thatwill always be present in the production of a given unit (feature,phoneme, syllable) and that will reliably be found in all conceivablecontexts."

However, the final nail in the coffin for the suggestion according towhich all tokens of the same word have the "same sound" is that wordscan be mispronounced. As Kaplan (1990) has argued, a word can sufferextreme mispronunciation and still be (a token of) the same word. Heasks us to imagine a test subject, who faithfully repeats each wordshe is told. After a time, we put filters on her that radically alterthe results of her efforts. Nonetheless, we would say she is sayingthe word she hears. Kaplan concludes that differences in sound betweentokens of the same word can be just about as great as we wouldlike. Notice that in such circumstancesintention—whatword the test subject intended to produce—is key. This suggeststhat perhaps what all tokens of a word, say, ‘color’, havein common is that they were produced as the result of an intention toproduce a token of the word ‘color’. Unfortunately,counterexamples are not hard to manufacture. (A clear phonemic exampleof ‘supercalifragilisticexpealadocious’ in English wouldprobably not count as a token of ‘color’, for example.)Counterexamples aside, it would be putting the cart before the horseto try to explain what the word ‘color’ is by appealing tothe intention to produce a token of the word ‘color’. Itwould be like trying to explain what a fork is by appealing to theintention to produce a fork. Intentions are important in helping toidentify which type a token is a token of—question(viii)—but will not help us with what the type is—question(i)—and so I shall ignore them in what follows.

4.2.5 Conclusion

The upshot of all this is that there is no nontrivial, interesting,"natural", projectible property that all tokens of a word have incommon, other than being tokens of that word (in the sense of‘word’ indicated). Tokens are all the same word, but theyare not all the same. That is, the answer to (iii) is no. What then,of the other questions, (i)-(viii)? They become more difficult toanswer. Wetzel (2002) attempts to answer them. The primary conclusionof Wetzel 2002 is that words aretheoretical entities,postulated by and justified by linguistic theory. Words, in the senseindicated, are individuated by a number of variables, includingorthography, phonology, etymology, grammatical function andsense(s). As for their tokens, they are apt to have some but not allof the properties of the type. And, as the story from Kaplan shows,tokens may even be quite deformed. These considerations impactsignificantly on the relation between types and their tokens,discussed in the next section, §5.

5. The Relation between Types and their Tokens

The relation between types and their tokens is obviously dependent onwhat a type is. If it is a set, as Quine (1987, p. 217) maintains, therelation isclass membership. If it is a kind, asWolterstorff maintains, the relation iskind-membership. Ifit is a law, as Peirce maintains, it is something else again, perhapsbeing used in accordance with. (Although Peirce also says atoken is an “instance” of a type and that the tokensignifies the type.) Nonetheless, it has often been taken to be therelation ofinstantiation, or exemplification; a token is aninstance of a type; it exemplifies the type. (Not that every instanceof a type is a token—e.g. capital ‘A’, small‘A’, and all the other types of ‘A’s tokenedin the display in §4.2 above may be said to be instances of theletter ‘A’.) As with other universals, there are twoversions of this relation, Platonic and Aristotelian. Although the twoversions of property realism are discussed at length under thisencyclopedia's entry for properties, a few remarks about the typeversions are in order.

According to Platonic versions of type realism, e.g., Bromberger 1989,Hale 1990, Katz 1981 and Wetzel 2002, the type is an abstract object,not located anywhere in space-time, although its tokens are. Thisversion appears to give rise to serious epistemologicalproblems—we don't see or hear the type, it isn't locatedanywhere in space-time, so how do spatiotemporal creatures such asourselves know it exists, or what properties it has? Admittedly, wesee and hear tokens, but how are they a guide to what the type islike? One answer, given for example, by Wolterstorff, for whom we sawin §4 that a type is a norm-kind, is that ordinary induction fromtokens would give us knowledge of types, at least in the case ofinstantiated types. Bromberger (1992a, p.176) claims that linguists“often impute properties to types after observing and judgingsome of their tokens and seem to do this in a principled way”and calls the principle that licenses this inference thePlatonicRelationship Principle. More specifically, he proposes (1992a)that the type, as the archetype of the quasi-natural kind whichcomprises the tokens, has just those projectible properties that allthe tokens have. He has in mind properties such as the same underlyingphonological structure, for words, and the same boiling point, forelements.

However, as we saw in §4, generally there are no such propertieshad by all and only tokens of a type, at least in the case ofwords—not same phonological structure, nor same sense nor samespelling. Not all tokens have any such natural projectible property(except for the property of being tokens of the sametype). It should be clear from §4 that the cookie cuttermodel—the idea of the type as just a perceptible pattern forwhat all the tokens look like—does not work. Goodman (1972,pp. 437-8) follows Peirce in using the word ‘replica’ toapply to all tokens of the same type, (although Peirce seemed to thinkthey were replicas of the type, whereas Goodman, being a nominalist,cannot think this) but not all tokens resemble each other in anyordinary sense beyond being tokens of the same type (although ofcourse some do). Goodman himself is clear about this, for he notesthere that “Similarity, ever ready to solve philosophicalproblems and overcome obstacles, is a pretender, an impostor, aquack…. Similarity does not pick out inscriptions that are‘tokens of a common type’, or replicas of each other. Onlyour addiction to similarity deludes us into accepting similarity asthe basis for grouping inscriptions into the several letters, words,and so forth.” But others, e.g. Stebbing (1935, p. 6) and Hardie(1936), claim that all spoken tokens are more or less similar to eachother. Because they are not, Wetzel (2002) and (2008) proposes thatsince the only property all the tokens of a type generally share isbeing tokens of the type, one of the primary justifications forpositing word types is that being a token of the word‘color,’ say, is the glue that binds the considerablevariety of space-time particulars together. The type is thus a veryimportant theoretical object, whose function is to unify all thetokens as being “of the same type”; in accordance with thePlatonic Relationship Principle, the type has properties based on theproperties ofsome of its tokens, but in a complexway—in addition, the tokens have some oftheirproperties in virtue of what properties the type has.

In Aristotelian versions of exemplification, such as Wollheim 1968 andArmstrong 1978, the type has no independent existence apart from itstokens. It is “in” each and every one of its tokens, andso can be seen or heard just as the tokens can be. This avoids theepistemological problem mentioned in the preceding paragraph, butmakes it hard to explain how some types—such as very longsentences—can have no tokens.

As against instantiation of any sort, Stebbing (1935, p. 9) arguesthat a token is not an instance of a type, because “the type isa logical construction out of tokens having similarity or conventionalassociation [as the inscribed word with the spoken]. It follows thatthe type is logically dependent upon the tokens, in the sense that itis logically impossible to mention the type-word without using atoken, and further, the meaning of the type has to be defined byreference to the tokens.” These claims are quitecontroversial. For example, it is clear one can refer to a type wordwithout using a tokenof it—one can say “the worda token of which is the first word on the page”.

Another alternative toexemplification isrepresentation. According to Szabo (1999, p. 150), types areabstract particulars, as with Platonic realism, but tokensrepresent their types, just as “paintings, photographs,maps, numerals, hand gestures, traffic signs and horn signals”represent, or “stand for” their representata. A word tokenof ‘horse’ represents the word ‘horse’, whichin turn represents horses. Just as a correct map of the planet canprovide us with knowledge of the planet, so too a token can provide uswith knowledge of properties of the type, thus addressing theepistemological problem. The representation view gives rise to aproblem however, for it turns out that what we have been calling wordtokens aren't words at all on this view, anymore than a map of aplanet is a planet, and this runs contrary to our usual thinking.

6. What is a Token?

It might seem that tokens are less problematic than types, beingspatio-temporal particulars. But certain complications should benoted. (We continue to use linguistic examples, but the remarks holdtrue for tokens generally).

• As Kaplan (1990, p. 97) pointed out, a token of a word could besuitably circumscribed empty space (e.g., in a piece of cardboardafter letters have been cut out of it).
• It would seem that a token might also be a mentalparticular—because a poem might be composed prior to its beingread or written down—although there is disagreement over whethersuch a mental particular is to count as a token.
• It will generally be the case that some tokens of the typeresemble in their outward physical appearance some other tokens of thetype, but as we saw in §4-5, it is generally not the case thatall the tokens of the same type resemble one another inappearance. (Recall the example of the letter ‘A’ in§4.)
• Even being similar in appearance (say by sound or spelling) to acanonical exemplar token of the type is not enough to make a physicalobject/event a token of that type. The phonetic sequence [Ah‘key ess ‘oon ah ‘may sah] is the same phonetic(type) spoken in Spanish or Yiddish. Yet if a Spanish speaker uses itto saya table goes here, she has not, in addition, saida cow eats without a knife in Yiddish. She has not saidanything in Yiddish, however phonetically similar what she said mightbe to a sentence token of Yiddish. So her token is a token in Spanish,not Yiddish. Meaningful tokens are tokens in a language.
• Along the same lines, being a physical object/event that issimilar in appearance (say by sound or spelling) to a canonicalexemplar token of the type is not even enough to make the physicalobject/eventmeaningful, because, as Putnam (1981, pp. 1-2)among others has argued, the token must have been appropriatelyintentionally produced.
• If it is not already clear from the foregoing, it should be notedthat a physical object that is a token of a type is not oneintrinsically—merely by being a certain sequence ofshaped ink marks, say. Just as being a brother is not an intrinsicproperty of brothers because one is a brother only relative to anotherperson, so a token is a token only relative to a type, and to alanguage, perhaps to an orientation, and perhaps, as Hugly and Sayward(1981 p. 186) have argued, to a tokening system (e.g., Morse Code),which they define as “a set of instructions that tell how, forany given expression, a speaker of the language could construct aperceptible particular that tokens that expression given that thespeaker had enough physical and mental resources…”.

7. Do Types Exist?

7.1 Universals

Since types are usually thought to be universals, the debate overwhether they exist is as longstanding as the debate over universals,and debaters fall into the same camps. Realists say they do, as we sawin §5 which surveyed several varieties of realism. Realism'straditional opponents were nominalists andconceptualists. Nominalists, who renounce universals and abstractobjects, say they don't. (See, e.g., Goodman and Quine 1947, Quine1953, Goodman 1977 and Bromberger 1992b). Conceptualists argued thatthere are no general things such as the specieshomo sapiens;there are only generalideas—that is, ideas that applyto more than one thing. Applied to words, the thesis would be thatwords are not abstract objects “out there”, but objects inthe mind. Their existence then, would be contingent on having beenthought of. While this contingency may have a good deal ofplausibility in the case of linguistic items, by itself conceptualismis just a stopgap measure on the issue of types and tokensgenerally. For ideas themselves are also either types or tokens (asevidenced by the fact that two people sometimes have the sameidea). So either the conceptualist is proposing that types are ideatypes—which would be a species of realism—or she isproposing that there are no types, only mental idea particulars inparticular persons, which is a version of nominalism. Conceptualism,therefore, will be ignored in what follows.

7.2 Realism

Realism is the most natural view in the debate with nominalism,because as we saw in some detail in §2 type talk isubiquitous. That is, we talk as though there are types in philosophy,science and everyday life. To say that we talk as though there aretypes is not to invoke the traditional argument for universals, whichis that a sunset and a rose are both red, so they have something incommon; and this something can only be the property ofbeingred; so properties exist. Quine (1953) convinced many atmid-century that this traditional argument for universals, whichrelies on predicates referring to something, fails. He objected that“the rose is red because the rose partakes of redness” isuninformative—we are no better off in terms of explanatory powerwith such extra objects asredness than we are without them;perhaps a rose's being red and a sunset's being red are just brutefacts. Rather, to say we talk as though there are types is, as we sawin §2, to appeal to the fact that in our theories we frequentlyuse singular terms for types, and we quantify over them. As we saw,Frege emphasized that singular term usage is in indicator ofobjecthood, and Quine stressed that we are ontologicallycommitted to that over which we quantify. Such considerations ledQuine himself (1987, p. 217) to hold that expression types such as‘red’ exist, even while he denied thatrednessdoes. Since at least on the face of it we are committed to types inmany fields of inquiry, therefore, it is incumbent upon the nominalistto “analyze them away”. (Or to maintain that all theorieswhich appear to refer to types are false—but this is a prettyradical approach, which will be ignored below.)

7.3 Nominalism

Realism is not without its problems, as was noted in §5above. Also favoring nominalism is Occam's principle which would haveus prefer theories with fewer sorts of entities, other things beingequal. The main problem for nominalists is to account for ourapparent theoretical commitment to types, which, whatever types are,are not spatio-temporal particulars (according to nearlyeveryone). Traditional nominalists argued (as their name implies) thatthere are no general things, there are only generalwords,and such words simply apply to more than one thing. But this is not asolution to the current problem, presupposing as it does that thereare word types—types are the problem. (Attempts to avoid thisare given by Goodman and by Sellars; see below.) So-called classnominalists hold that a word type is just the class, or set, of itstokens. But this is unsatisfactory because, first, as we saw in§5, classes are ill-suited for the job, since classes have theirmembership and their cardinality necessarily, but how many tokens atype has is usually a contingent matter. (For the same reason,mereological sums of tokens are unsuited for the job of types, as theyhave theirparts essentially.) And second, classes areabstract objects too, so it is hard to see how this is really a formof nominalism about abstract objects at all.

Initially more promising is the nominalistic claim that the surfacegrammar of type talk is misleading, that talk of types is justshorthand for talk of tokens and is thus harmless. To say ‘Thehorse is a mammal’ is just to say ‘All horses aremammals’; to say ‘The horse is a four-legged animal’is to say, as Frege himself (1977) suggested, ‘All properlyconstituted horses are four-legged animals’. The idea is to“analyze away” apparent references to types by offeringtranslations that are apparently type-free, and otherwisenominalistically acceptable. The problem is how to do this for eachand every reference to or quantification over a type/types. Given theubiquity of such references/quantifications, only the procurement ofsome sort of systematic procedure could assure us it can be done.However, chances of formulating a systematic procedure appear slight,in view of the following obstacles that would have to be overcome.

• If the “translation” is not to turn out to betrivially true because there are no tokens, it must be the case thatall types have tokens. We saw in §5 there is reason to deny this,because it would falsify the law of syntax that says that ifΦ,Ψ are sentences, then so is (Φ & Ψ), since thereare only finitely many sentence tokens.
• The examples of ‘The horse is a mammal’ and ‘Thehorse is four-legged’ suggest that ‘The Type is P’is to be analyzed as ‘All tokens, or all normal tokens, areP’. But this won't work. The noun ‘colour’ isspelled ‘color’ (among other ways) , but not all properlyformed tokens of it are. ‘Colour’ also meansof theface or skin (among other things), but not only do not all tokensof it meanof the face or skin, probably most do not, and theaverage one does not.
• These semantic facts show that ‘The Type is P’ cannotbe analyzed as ‘Either all tokens are P, all properly formedtokens are P, most tokens are P or the average token is P’. Noris there any point in searching for some other magic statisticalformula thatwould express exactly how many tokens have to beP for the Type to be P. This is because some of the properties of theType are collective properties—e.g., when it is said that“The grizzly bear,Ursus Horribilis, had at one time aU.S. range of most of the West, and numbered 10,000 is Californiaalone, but today its range is Montana, Wyoming and Idaho and itnumbers less than 1000.” Having a range of most of the West istrue of no individual bear.
• It is very difficult even to find a nominalistic paraphrase for“Old Glory had twenty-eight stars in 1846 but now hasfifty”, as Wetzel (2008) argues.
• So far we have only been considering analyzing away singularterms. Quantifications are indeed a nightmare. Here is Quine andGoodman's (1947 p. 180) nominalized version of “There are morecats than dogs”: “Every individual that contains a bit ofeach cat is bigger than some individual that contains a bit of eachdog” where a bit “is an object that is just as big as thesmallest animal among all cats and dogs”. Ingenious though theirstrategy is, it offers no help for nominalizing “Of 20,481species examined, two-thirds were secure, seven percent werecritically imperiled, and fifteen percent were vulnerable”.

7.3.1 Goodman's Nominalism

I have been writing as though it were easy to pick out and quantifyover tokens of a type without referring to the type. Sometimes it is:‘The horse…’— ‘Allhorses…’. But this will not generally be thecase. Consider the noun ‘color’. The natural way to pickout its tokens is “all tokens of the noun‘color’”, but obviously this will not do as anominalistic paraphrase, for it contains a reference to atype. Goodman (1977) suggests a systematic way of paraphrasing and ageneral procedure for substituting nominalistic paraphrases forlinguistic type-ridden sentences. In 1977 (p. 262) he claims that“any “ ‘Paris’ consists of five letters”is short for “Every ‘Paris’-inscription consists offive letter-inscriptions” ”. The idea seems to be toreplace singular terms by predicates, which nominalists such asGoodman think carry no ontological commitment. So instead of “aninscription of ‘color’” write “a‘color’-inscription.” It seems pretty clear,however, that, based on the rules of quote-names for words, thispredicate still retains a singular term for a type, the word‘color’ just as clearly as “a George W. Bushappearance” still contains a reference to Bush. That this is sois even more obvious in the case of “an ‘a cat is on themat’-inscription”. So the question is how we mightidentify these tokens grammatically but without referring to the noun‘color’ itself and still say something true and (in someappropriate sense) equivalent. Is there perhaps some other way ofquantifying over tokens of a word, without referring to the word or toany other type? The fact that tokens are all “the sametype” suggests they are all “the same” somehow,which begets the idea that the type must embody certainsimilar features that all and only its tokens have, such asspelling, or sense, or phonological structure or some combination ofthem. This is a beguiling idea, until one tries to find such afeature, or features, amid the large variety of its tokens—eventhe well-formed tokens. As we saw in §4, there is no such feature(consider again ‘color’ and ‘schedule’). Theydemonstrate that e.g. neither same spelling, same sense, nor samepronunciation prevail. In any event, such possible defining featuresinvolve reference to types: letter types in the spellings, phonemetypes in the pronunciation. These too would have to be “analyzedaway” in terms of quantifications over particulars. It might bethought that Sellars solved this problem by appealing to the notion ofalinguistic role, which Loux (1998, p. 79) defines as:two word (tokens) have the same linguistic role if they“function in the same way as responses to perceptual input; theyenter into the same inference patterns; and they play the same role inguiding behavior”. It is dubious whether this notion can beunpacked without referring to abstract objects (same inferencepatterns?), but in any event it cannot be used to pick out all tokensof a word, as we have been using the word ‘word’. Thereason is that ‘red’ and ‘rouge’ are differentwords in our sense, but their tokens play the same linguistic role forSellars.

But even if expressions like “a‘color’-inscription” did not contain singular terms,Goodman's suggestion that whatever is true of the type is true of allthe tokens has two fatal defects. First, it would turn truths intofalsehoods. ‘Paris’ consists of five letters, as does‘color’, but not every ‘color’-inscriptionconsists of just five letter-inscriptions since some are spelled‘c’-‘o’-‘l’-‘o’-‘u’-‘r’.(As for denying that these are inscriptions of the word‘color’ see “Orthography” in §4 above.)It doesn't even work for the word ‘Paris’, as it has theforms ‘Pareiss’ and ‘Parrys’ also. Second,Goodman's technique of replacing singular terms by predicates onlyworks if we are employing a realist semantics. That is, the key to hisprogram is that for every word in the dictionary and every sentence inthe language there corresponds a unique predicate that is true of justthe tokens of that word/that sentence. (Without such predicates, truestatements apparently referring to types would be short for nothing atall.) But this is only plausible on a realist semantics. If‘predicate’ meant ‘predicate token’ therewould not even be enough predicates for every word in the dictionary,much less every sentence in the language. That is, it is certainlyfalse that for every word α in the dictionary, there is anactual predicate token of the form ‘is anα-inscription’. For more details on the problems withGoodman's strategy, see Wetzel (2000).

7.3.2 Sellars's nominalism

Sellars's suggestion is rather similar to Goodman's, and faces similardifficulties, but is sufficiently different to be worthmentioning. Loux (1998, pp. 75-83), for one, argues that Sellars(1963) achieves the best nominalist account available by overcomingcritical objections to Carnap's “metalinguistic” accountas follows. Carnap (1959, pp. 284-314) had suggested that all claimsinvolving apparent reference to abstract objects, such as “Redis a color”, are systematically to be understood asmetalinguistic claims about the word involved (“The word‘red’ is a color predicate”). There are two obviousobjections to Carnap's suggestion. The first is that we still haveword types being referred to; the second is that translation of“Red is a color” and “The word ‘red’ isa color predicate” into French, say, will not result insentences that are equivalent, since the latter (“Le mot‘red’ est un prédicat de coleur”) will stillrefer to an English word, but the former will not. To counter thefirst objection, Sellars (1963, pp. 632-33) claims “the word‘red’” is adistributive singular term,which typically consists of ‘the’ followed by a commonnoun that purports to name a kind—e.g., ‘the lion’in “The lion is tawny”—and that use of it results ina generic claim that is about token lions, not thetype—“All lions are tawny”—since “The Kis f” is logically equivalent to “All Ks aref”. Note that this is similar to Goodman's suggestion and assuch is subject to the same criticism as is Goodman's: the twosentence forms said to be logically equivalent are not. See the secondbullet point three paragraphs above. Nor is there some other simpleand straightforward logical equivalence that does the job; see thethird and fourth bullet points above. At best this is a suggestionthat some of the sentences in question—the ones that do notcontain “collective” predicates like‘extinct’—are logically equivalent to genericclaims—e.g. “Lions are tawny”—which arecapable of being true while admitting of exceptions. However, thatthere is a uniform semantics (even a very complex one) for genericclaims, one moreover that “analyzes away” kind talk, is astrong empirical claim about language, one that does not receivestrong support from current efforts to analyze generic claims. (SeeCarlson and Pelletier 1995.)

To counter the second objection (about non-equivalent translations)Sellars suggests we introduce a new sort of quotation device, dotquotation, that would permit translation of the quoted word. Then,e.g. “The word •red• is a color predicate”translates as “Le mot •rouge• est un prédicatde coleur”. The justification is that ‘red’ and‘rouge’ are functionally equivalent—they play thesame linguistic role. As mentioned above, this means that they havethe same causes (of perceptual stimuli, e.g.) and effects (conduct,e.g.) and function similarly in inferential transitions. Again,whether Sellars can unpack the notion of “same linguisticrole” without appealing to any universals remains an openquestion. Less open is the question of how successful such an approachis to dealing with talk of words, e.g., “The word‘red’ consists of three letters”. If we applySellars' analysis systematically, this becomes “”The word‘red’” consists of a three-lettered predicate”which is just false; if instead we reckon that “the word‘red’” is already a distributive singular term andapply his analysis to it, getting “The word •red• is athree-lettered predicate”, this then translates incorrectly as“Le mot •rouge• est un prédicat trois-enlettres”. If we stick instead with regular quotation, whichresists translation, then we cannot appeal to “same linguisticrole” to say what class of things “the word‘red’” distributes over. In any event, Sellars'saccount also faces the difficulty noted above with Goodman's account:it is only plausible on a realist semantics, because if‘predicate’ meant ‘predicate token’ therewould not even be enough predicates for every word in the dictionary,much less every sentence in the language. (For other problems withSellars's account, see Loux 1978: 81-85.)

8. Occurrences

As mentioned in §1 above, there is a related distinction thatneeds to be mentioned in connection with the type-tokendistinction. It is that between a thing, or type of thing, and (whatis best called) anoccurrence of it—where an occurrenceis not necessarily a token. The reason the reader was asked above tocount the words in Stein's linein front of the reader'seyes, was to ensure that tokens would be counted. Tokens areconcrete particulars; whether objects or events they have a uniquespatio-temporal location. Had the reader been asked to count thewords in Stein's line itself, the reader might still have correctlyanswered either ‘three’ or ‘ten’. There areexactly three word types, but although there are ten word tokens in atoken copy of the line, there aren't any tokens at all in the lineitself. The line itself is an abstract type, as is the poem in whichit first appeared. Nor are there ten word types in the line, becauseas we just said it contains only the three word types,‘a,’ ‘is’ and ‘rose,’ each ofwhich is unique. So what are there ten of?Occurrences ofwords, as logicians say: three occurrences of the word (type)‘a,’ three of ‘is’ and four of‘rose’. Or, to put it in a more ontologically neutralfashion: the word ‘a’ occurs three times in the line,‘is’ three times and ‘rose’ four times.Similarly, the variable ‘x’ occurs three times in theformula ‘∃x (Ax & Bx)’.

Now this may seem impossible; how can one and the same thing occurmore than once without there being two tokens of it? Simons (1982)concludes that it can't. Wetzel (1993) argues that it is useful todistinguish objects from occurrences of them. For example, in thesequence of numbers <0,1,0,1> the very same number, the numberone, occurs twice, yet its first occurrence is distinct from itssecond. The notion ofan occurrence of x in y involves notonly x and y, but also how x is situated in y. Even aconcrete object can occur more than once in asequence—the same person occurs twice in the sequence of NewJersey million dollar lottery winners, remarkably enough. If we thinkof anexpression as a sequence, then the air of mystery overhow the same identical thing can occur twice vanishes. Does this meanthat, in addition to word types and word tokens, word occurrences mustalso be recognized? Not necessarily; we can unpack the notion of anoccurrence using “occurs in” if we have the notion of asequence available; see Wetzel 1993 for details.

The need to distinguish tokens of types from occurrences of typesarises, not just in linguistics, but whenever types of things haveother types of things occurring in them. There are 10,000 (or so)notes in Beethoven'sSonate Pathétique, but there areonly 88 notes (types) the piano can produce. There are supposed to befifty stars (types) in the currentOld Glory (type), but thefive-pointed star (type) is unique. And what could it mean to say thatthe very same atom (type), hydrogen, “occurs four times”in the methane molecule? Again, the perplexing thing is how the verysame thing can “occur” more than once, without there beingmore than one token of it. Armstrong (1986), Lewis (1986a,b) andForrest (1986) called such types “structural universals”,which were the subject of a debate among the three. Armstrong andForrest defended Armstrong's (1978) view of universals against Lewis,who delineated seven different views of structural universalscompatible with Armstrong's theory, and found all of them wanting.Basically, Lewis (1986a) assumed that there are two sorts of waysstructural universals might be constructed of simpler universals:set-theoretically and mereologically. He argued that set-theoreticalconstructions resulted in ersatz universals, not universals worthy ofthe name, and that the various mereological constructions justresulted in a heap of simpler universals, where there could not betwo of the simpler universals. Wetzel (2008) argues thatthere is a conception of a structural universal, the “occurrenceconception”—which is an extension of the occurrenceconception of expressions mentioned above—that escapes Lewis'sobjections.

Bibliography

• Allwood, Jens (1998), “Some Frequency based Differencesbetween Spoken and Written Swedish,”Proceedings from theXVIth Scandinavian Conference of Linguistics, Department ofLinguistics, University of Turku.
• Armstrong, David (1986), “In Defense of StructuralUniversals,”Australasian Journal of Philosophy, 64:85-88.
• ––– (1978),Universals and ScientificRealism, Vol II. A Theory of Universals. Cambridge: CambridgeUniversity Press.
• Block, Ned (ed.) (1980),Readings in Philosophy of Psychology,vol. 1. Cambridge, MA: Harvard University Press.
• Bromberger, Sylvain (1992),On What We Know We Don't Know.Chicago: University of Chicago Press.
• ––– (1992a), “Types and Tokens inLinguistics,” in Bromberger 1992, pp. 170-208.
• ––– and Halle, Morris (1992b), “TheOntology of Phonology” in Bromberger 1992, pp. 209-228.
• ––– and Halle, Morris (1986), “On theRelationship of Phonology and Phonetics,” in Perkell and Klatt1986.
• Carlson, Gregory and Pelletier, Francis Jeffry (eds.) (1995),The Generic Book. Chicago: University of Chicago Press.
• Carnap, Rudolf (1959),The Logical Syntax of Language.Patterson, NJ: Littlefield, Adams and Co.
• Chomsky, Noam (1957),Syntactic Structures. The Hague:Mouton & Co.
• Dancy, Jonathan (2004),Ethics WithoutPrinciples. Oxford; New York: Clarendon Press.
• Davidson, Donald (1980), “Mental Events,” inEssayson Actions & Events. Oxford: Clarendon Press.
• Davies, Stephen (2001),Musical Works & Performances: aPhilosophical Exploration. Oxford: Clarendon Press.
• Collinge, N. E. (ed.) (1990),An Encyclopedia of Language.London: Routledge.
• Darwin, Charles (1859),On the Origin of Species. London:John Murray.
• Davis, Wayne (2003),Meaning, Expression and Thought.Cambridge, UK: Cambridge University Press.
• Forrest P. (1986), “Neither Magic Nor Mereology: A Reply toLewis,”Australasian Journal of Philosophy, 64:89-91.
• Frege, Gottlob (1977), “On Concept and Object,” inGeach, P. & Black, M.(eds.)Translations from the PhilosophicalWritings of Gottlob Frege. Oxford: Basil Blackwell.
• Fudge, Eric (1990), “Language as Organized Sound:Phonology,” in Collinge 1990.
• Goodman Nelson (1977),Structure of Appearance,3^rd edition. Dordrecht, Holland: Reidel.
• ––– (1972a),Problems and Projects,Indianapolis: Bobbs-Merrill Co.
• ––– (1972b), “Seven Strictures onSimilarity,” in Goodman 1972a.
• ––– and Quine, W.V. (1947), “Steps Toward aConstructive Nominalism,”Journal of Symbolic Logic, 12:105-22. Reprinted in Goodman 1972a..
• Greenlee, D. (1973),Peirce's Concept of Sign. TheHague: Mouton.
• Grice, Paul (1969), “Utterer's Meaning andIntentions,”The Philosophical Review, 78: 147-177.
• Hale, Bob (1987),Abstract objects. Oxford and New York:Basil Blackwell.
• Hardie, C.D. (1936), “The Formal Mode of Speech,”Analysis, 4: 46-48.
• Hugly, Philip, and Sayward, Charles (1981), “Expressions andTokens,”Analysis, 41: 181-87.
• Hutton, Christopher (1990),Abstraction & Instance: TheType-Token Relation in Linguistic Theory. Oxford: PergamonPress.
• Jubien, Michael (1988), "On Properties and Property Theory" inChierchia, Gennaro (ed.),Properties, Types and Meaning, Vol. I:Foundational Issues, Dordrecht, Holland:Kluwer-Academic-Publishers.
• Kaplan, David (1990), “Words”,Proceedings of theAristotelian Society, (Supp. Vol.) 64: 93-120.
• Katz, Jerrold J. (1981),Languages and Other AbstractObjects. Totawa, N.J.: Rowman and Littlefield.
• Kim, Jaegwon (1966), “On the Psycho-Physical IdentityTheory,”American Philosophical Quarterly, 3:227-235.
• Kripke, Saul (1972), “Naming and Necessity,” inDavidson, D. and Harman, G. (eds.),Semantics of NaturalLanguage. Dordrecht, Holland: D. Reidel.
• Lewis, David (1986a), “Against Structural Universals,”Australasian Journal of Philosophy, 64: 25-46.
• ––– (1986b), “Comment on Armstrong andForrest,”Australasian Journal of Philosophy, 64:92-93.
• Lindblom, Bjorn (1986), “On the Origin and Purpose ofDiscreteness and Invariance in Sound Patterns,” in Perkell andKlatt 1986.
• Loux, Michael (1998),Metaphysics: A ContemporaryIntroduction, 2^nd edition. London: Routledge.
• ––– (1978),Substance and Attribute.Dordrecht, Holland: D. Reidel Publishing Co.
• Lyons, John (1977),Semantics, Vol. 1. Cambridge:Cambridge University Press.
• Mayr, Ernst (1970),Populations, Species, and Evolution.Cambridge, MA: Belknap Press of Harvard University Press.
• McArthur, Tom (1992),The Oxford Companion to the EnglishLanguage. Oxford: Oxford University Press.
• Mill, John Stuart (1979),Utilitarianism, edited by GeorgeSher. Indianapolis: Hackett Pub. Co.
• Murdoch, Iris (1970),The Sovereignty of Good. New York:Schocken Books.
• Peirce, Charles S. (1931-58),Collected Papers of CharlesSanders Peirce, Hartshorne and Weiss (eds.), Cambridge, MA: HarvardUniversity Press.
• Perkell, Joseph S. and Klatt, Dennis H., eds. (1986),Invariance and Variability in Speech Processes, Hillsdale,NJ: Lawrence Erlbaum Associates.
• Place, U.T. (1956), “Is Consciousness a Brain Process?”British Journal of Psychology, 47: 44-50.
• Putnam, Hilary (1981),Reason, Truth and History.Cambridge: Cambridge University Press.
• ––– (1975), “The Meaning of Meaning,”Mind, Language and Reality. Cambridge: Cambridge UniversityPress.
• Quine, Willard.Van (1987),Quiddities: An IntermittentlyPhilosophical Dictionary. Cambridge, MA: Harvard UniversityPress, pp.216-219.
• ––– (1953), “On What There Is,”From A Logical Point of View. Cambridge, MA: HarvardUniverity Press.
• Ross, W.D. (1988),The Right and the Good. Indianapolis:Hackett Pub. Co.
• Sellars, Wilfrid (1963), “Abstract Entities,”Review of Metaphysics, 16: 627-671.
• Simons, Peter (1982), “Token Resistance,”Analysis, 42/4: 195-203.
• Simpson, G.G. (1961),Principles of Animal Taxonomy (NewYork: Columbia University Press).
• Smart, J.J.C. (1959), “Sensations and Brain Processes,”Philosophical Review, 68: 141-56.
• Stebbing, Susan (1935), “Sounds, Shapes and Words,”Proceedings of the Aristotelian Society, (Supp. Vol.) 14:1-21.
• Szabó, Zoltán (1999), “Expressions and TheirRepresentations,”The Philosophical Quarterly, 49:145-163.
• Wetzel, Linda (1993), “What Are Occurrences ofExpressions?”Journal of Philosophical Logic, 22:215-220.
• ––– (2000), “The Trouble WithNominalism,”Philosophical Studies, 98: 361-370.
• ––– (2002), “On Types and Words,”Journal of Philosophical Research, 27: 237-263.
• ––– (2008),Types and Tokens: An Essay onUniversals. Cambridge, MA: MIT Press.
• Wollheim, Richard (1968), Art and Its Objects. New York: Harper andRow.
• Wolterstorff, Nicholas (1980),Works and Worlds of Art.Oxford: Clarendon Press.
• ––– (1975), “Toward An Ontology of ArtWorks,”Noûs, 9: 115-142.
• ––– (1970),On Universals: An Essay inOntology. Chicago: University of Chicago Press.
• Yablo, Stephen (2002) “Abstract Objects: A CaseStudy,”Nous, 36 (Supp. Vol. 1): 220-240.
• Zalta, Edward (1983),Abstract Objects: An Introduction toAxiomatic Metaphysics. Dordrecht: D. Reidel.
• Zemach, Eddy (1992),Types: Essays in Metaphysics. Leiden,the Netherlands: E.J. Brill.

Academic Tools

	How to cite this entry.
	Preview the PDF version of this entry at theFriends of the SEP Society.
	Look up topics and thinkers related to this entry at the Internet Philosophy Ontology Project (InPhO).
	Enhanced bibliography for this entry atPhilPapers, with links to its database.

Other Internet Resources

[Please contact the author with suggestions.]

Related Entries

mind/brain identity theory |nominalism: in metaphysics |Peirce, Charles Sanders |Peirce, Charles Sanders: theory of signs |properties |realism