Much scholarship in the humanities and social sciences can be defined as an attack on categories. Whenever categorization is viewed as a form of oppression, liberation becomes equated with rejecting categories. We often celebrate literary texts that defy genres because critics have come to value the transgression of boundaries. But despite our attempts to reject categories, they will not go away. Indeed, categories persist because we continually categorize. As John Taylor states in Linguistic Categorization, "Human beings are categorizing creatures par excellence" [xi]. If you doubt that fact, simply watch toddlers at play categorize their toys. As to the question of why do we categorize, according to Taylor, "Categorization makes it possible for an organism to reduce the limitless variation in the world to manageable proportions" , and "the function of categorization is to reduce the complexity of the environment" . In other words, to categorize is to survive.
Categories pervade both everyday life and linguistics, and for Taylor linguistic categorization refers to how verbal language reflects cognitive categories and how concepts used by linguists arise from categories. As an Oxford Textbook in Linguistics, Linguistic Categorization is ostensibly aimed at students. Indeed, each chapter ends with a few study questions. But like many so-called textbooks, teachers may also learn something useful here. This is because Taylor's prose is clear, his argument coherent, and his bibliography rather thorough. Now in its third edition, Linguistic Categorization answers a criticism made by the linguist P.H. Matthews. In his 1992 review of Taylor's first edition, Matthews thought Taylor fell short when what was needed was something "which will allow us to treat 'cognitive linguistics' with more respect." This is now that book.
In the first four chapters, Taylor covers the pioneering findings Berlin, Kay, Rosch, Mervin, and Labov, to name just a few of the researchers involved with categorization theory. In contrast to Bloomfield's belief that categories were arbitrary , Taylor points to a significant amount of irrefutable empirical evidence to show there is a remarkable cross-cultural consistency, for example, in the perception of colors even though color names vary. As Taylor reminds us, "For structural linguistics, language was a self-contained system of signs, independent of the cognition and experience of its users. In contrast, cognitive linguistics strongly emphasizes the non-arbitrary, motivated nature of language structure" . If cognition and conceptualization motivate language, then categorization offers clear evidence of this. As a cognitive linguist Taylor is able to critique the classical theory of categorization and promote the prototype theory of categorization. The so-called classical theory, which goes back to ancient Greece, posited clear limits and boundaries. Something either belongs to a category or it does not since boundaries are not fuzzy .
But the prototype theory of categorization maintains that boundaries are indeed fuzzy. Subjects in Labov's experiments, for instance, revealed "there was no clear dividing line" between cups, bowls, or vases since "one category merged gradually into another" . If Rosch is the mother of prototype theory, then Taylor is right to refer to her research countless times. For Rosch, categories are comprised of three levels. The superordinate level names the category as a whole. The basic level is the level at which we organize the category in terms of an exemplar or prototype while the subordinate level includes marked instantiations of items related to the prototype. For example, when asked to think about the category of "furniture," subjects in Rosch's American psychological experiments often picked "chair" as the basic level object (i.e. the prototype) and situated a marked object like "easy chair" at the subordinate level . The chair—the basic level object—is the prototypical exemplar around which subjects cognitively construct the furniture category as a whole. Simply put, we construct categories around prototypes and their perceived family resemblance with more peripheral members of the category, a concept Taylor refers to as "degree of membership" . On the classical view, something either is a member of a category or it is not. On the prototypical view, however, some objects are better examples of a category than others.
The better examples are prototypes although cross-cultural variations with prototypes are common. For example, in Dirven's möbel experiment, the furniture category's prototype for German subjects was "bed" . And while Taylor places cats and dogs in different categories  when both could belong to the same category (i.e. pets), there may be prototype divergences within the same culture as well—especially if some people prefer dogs to cats as their prototype for the household pet category. Such variety may be hard to explain, but the important finding has been that categories are universally constructed around prototypes that exemplify the category as a whole. This is obvious in literary studies. For example, in his third edition of the Penguin Dictionary of Literary Terms and Literary Theory, Cuddon reprints John Clare's "The Secret" as a way to define the lyric poem genre. "The Secret" thus becomes Cuddon's prototype for the "lyric poem" category as a whole, and it becomes a way of defining what a lyric poem is. Definition by prototype is probably a tactic used in other literary term handbooks too whenever genres are defined. Simply put, prototypes are ubiquitous.
Taylor ends Chapter 4 with an informative analysis of hedges. These are expressions like "par excellence" and "strictly speaking" that "require us to distinguish between central and peripheral members of a category [ ] as well as between different degrees of non-membership in a category" . Phrases like "A robin is a bird par excellence" or "Strictly speaking, a bat is not a bird" make explicit our categorizations. Related to hedges are reduplications, as in "Is he French-Canadian or French-French?" . Reduplications also establish clarifications over prototypes. Someone in France once told me she had studied at the Sorbonne. When I naively asked if she meant the Sorbonne or the Sorbonne Nouvelle, her answer (the "Sorbonne-Sorbonne") reinforced a prototype.
Ideas like these lead Taylor to cover vital concepts in cognitive linguistics in chapter 5. Such concepts include domains, schemas, frames, scripts, profiling, and perspectivization. The discussion of semantics Taylor starts here continues in various forms through Chapter 8. Despite the fact that Taylor's account of the creation of nominal compounds  overlooks Fauconnier and Turner's theory of conceptual integration and how it accounts for compounds like "land yacht," Taylor does treat the problem of polysemy at length. Of course, polysemy is not a problem—if it were, it would not exist. The problem, rather, is how to account for it. Polysemy is important for it offers "linguistic evidence [ ] that prototype categories can be linked to family resemblance structures" . After listing seventeen sentences in Chapter 6 that reveal the polysemy of the preposition "over" [113-114], Taylor points out that the senses of "over" can be reduced to "four major clusters" , which may be seen in phrases like "The lamp hangs over the table," "He put his hands over his face," "The plane flew over the city," and "The lesson is over." These ideas are no doubt relevant to teachers of English to speakers of other languages. But they also hint at an important question: are meanings themselves categories around which some singular, core, prototypical sense is located?
Although there is a "veritable cottage industry of over studies"  since many linguists do not seem satisfied with all the accounts published on the topic thus far, answering these questions is not easy. However, Taylor comes back to similar questions in Chapter 8 with the recognition that "there is virtually no affinity at all" , for example, between the sense of "round" in a phrase like "running round the park" and in a phrase like "driving round the city center." Proponents of "the two-level approach"  such as Bierwisch reject a prototype account of polysemy, but Taylor gives their argument with respect. Even Langacker's network model, which Taylor finds promising, seems to entail this consequence: "the question of whether a word is polysemous or not turns out to be incapable of receiving a definite answer" . That may not be satisfactory but it is nevertheless honest. That said, if "Metaphor has always been something of an embarrassment to generative linguistics" , it has been a gravy train for cognitive linguistics. This is why Chapter 7 covers metaphor and metonymy. Although Taylor covers ground many readers may be familiar with already, his overview is accurate. Yet those who expect to see references to Turner's work on XYZ metaphors  or Grady's work on similarity and correlation  might be disappointed. This is not to say Taylor's argument here is weak; rather, some readers may have expectations of what references a cognitive linguistic discussion of metaphor will include, especially if they have a prototype of such a discussion in mind!
Categorization in linguistics is addressed in chapters 9 through 12. While Taylor's discussion of Russian cases  overlooks Janda's research on this topic, his presentation of the Italian diminutive case is coherent, especially his explanation of the experiential cognitive motivation for its use . But if French speakers thought Taylor's discussion of the metaphorical sense of venir was interesting , they may be confused by this: "In some languages (French is a well-known example) the distinction between the deictic and narrative senses (of the past tense) is grammaticalized; the 'passé composé' (il a vu "he saw") is employed for deictic reference, while the 'passé historique' (il vit) is obligatory in historical and fictional narratives" . By "passé historique" Taylor means the passé simple. He is correct to say it is used specifically in writing, but when we find the imparfait and présent used on nearly every page of Michelet's Histoire de la Revolution Française, is the "passé historique" really "obligatory" for all "historical" writers? To his credit, in Chapter 10 Taylor highlights "the failure of successive generations of linguists to come up with a satisfactory set of core meanings for the elements of intonation" . Confusion over terms like tone, pitch, and key seems rooted in divergent prototypes among linguists of what the best examples of each are. Taylor shows the same is true in chapter 13 when it comes to phonology and the confusion over terms like phoneme, vowel, consonant, syllable, rhyme, onset, nucleus, coda, and sonority.
Grammar teachers will find most useful the discussion of grammatical categories in Chapter 11. Taylor follows Ross and Lakoff's footsteps to remind us of the "fuzziness of grammatical categories" . This is worth remembering when students ask for a precise definition of "noun," "verb," or "word." If "time-stability" for verbs and "enduring" spatial existence for nouns identify the traits of noun and verb prototypes , then some nouns will seem better than others as examples of what a noun is while some verbs will seem better than others as examples of what a verb is (just as prototype theory would predict). "Grammatical categories have a prototype structure," according to Taylor, and "semantic criteria are relevant to grammatical categorization" . This is part of Taylor's case against the semantic autonomy hypothesis, an argument that continues, in Chapter 12, with a discussion of constructions, perhaps the best proof we have that semantics and syntax are inseparable. As usual, Taylor begins the chapter by explaining what is wrong with the generative account. Then he defines constructions as form-meaning pairings whose semantics may reveal prototypical properties. For example, with the prenominal possessive (e.g. "John's car"), "the relation of possession"  is primary or prototypical to the construction's meaning (i.e. that John possesses the car in question). Expressions like "John gave Mary the Book" are what Taylor calls transitive constructions and what Goldberg calls ditransitive constructions. They encode the semantics of a transitive event whereby "the transitive construal of the state of affairs"  involving an agent, an action, and a recipient of an object is shown to be the prototype for the construction's semantics.
Linguistic Categorization concludes, in Chapter 14, on the acquisition of categories. In my opinion this is where Taylor most fully answers critics like Matthews. Generative linguists are allegedly concerned with language acquisition, but as Taylor boldly states: "Anyone studying the development of language in the young child is hardly likely to be able to find much inspiration in a theoretical model which makes a clean division between syntax and semantics, and between semantic structure and conceptualization, and which, moreover, assumes that the basic architecture of a language is already present in the child at birth. It is in child language, in fact, that we are particularly likely to find confirmation for the cognitive linguistic hypothesis of the grounding of language structure in non-linguistic cognition" . What follows is an analysis of the research findings of Piaget, Tomasello, and many others. Taylor is in eminent company here. If the category is the enemy of the social science scholar, the arbitrary is the enemy of the cognitive linguist. As Taylor confidently concludes, "categories are not arbitrary, but are structured along natural principles of category formation" . After reading Linguistic Categorization, the onus is on those who wish to argue otherwise.