LINGUIST List 24.32

Tue Jan 08 2013

Review: Cognitive Science; Philosophy of Language: Taylor (2012)

Editor for this issue: Monica Macaulay <>

Date: 08-Jan-2013
From: Stefan Hartmann <>
Subject: The Mental Corpus: How Language is Represented in the Mind
E-mail this message to a friend

Discuss this message

Book announced at

AUTHOR: Taylor, John R.TITLE: The Mental CorpusSUBTITLE: How Language is Represented in the MindPUBLISHER: Oxford University PressYEAR: 2012

Stefan Hartmann, German Department (Deutsches Institut), University of Mainz


In this book, John R. Taylor defends the hypothesis that linguistic knowledgecan be conceived of as a repository of memories of exposure to actual usageevents. According to Taylor, language users keep track of the utterances theyencounter, thereby compiling a “mental corpus” of constructions at variouslevels of abstraction. This hypothesis is elaborated on and substantiated by agreat number of examples in the thirteen chapters of this book. In line withhis plea to focus linguistic attention on actual language use, most ofTaylor’s examples are derived from corpora (mainly from the British NationalCorpus and the Corpus of Contemporary American English).

The first two chapters of the book are concerned with the generative model oflinguistic knowledge, which Taylor rejects. Chapter 1, “Conceptualizinglanguage”, introduces Chomsky’s (e.g. 1986) distinction between E-language,i.e. ‘external’ language as encountered in the world, and I-language, i.e. the‘internal’ system of linguistic knowledge in the speaker’s mind. In contrastto the generative model, which treats E-language as an “epiphenomenon”(Chomsky 1986: 25), Taylor argues that I-language emerges from the exposure toE-language events. He then discusses to what extent E-language can be studiedwith the help of corpora, especially focusing on the pivotal question ofrepresentativeness. He also addresses the use of the World Wide Web as acorpus, reviewing research that shows data obtained from the Internet tocorrelate highly both with data from established corpora and with informants’acceptability judgments.

The title of Chapter 2, “The dictionary and the grammar book”, refers to thegenerative model of linguistic knowledge, which sees language as consisting ofthe lexicon on the one hand, i.e. “a finite list of structural elements”(Jackendoff 2002: 39), and the grammar on the other, i.e. “a finite set ofcombinatorial principles” (ibid.). While this model accounts neatly for thephenomenon of linguistic creativity, i.e. the ability of speakers to create apotentially infinite set of sentences from a finite set of elements, itentails a rule-based approach to linguistic knowledge that proves problematicin various respects. In evaluating the model against actual language use asattested by corpus data, Taylor argues that grammaticality intuitions do notin all cases reliably reflect the contents of the mental grammar. Heillustrates this with the use of “explain” in the double object construction,e.g. “explain me this”. While this construction strikes most native speakersof English as ungrammatical, Taylor’s corpus data suggest that in the mentalgrammar of a fair number of language users, “explain” can be used analogicallyto semantically related words such as “tell”. The critical assessment of thegenerative model leads Taylor to question both the classical definition of thelexicon as “a list of ‘exceptions’” (Chomsky 1995: 235) and the principle ofcompositionality, which “requires that each component of a complex expressioncontributes a fixed and stable chunk of semantic [...] material to theexpressions in which it occurs” (p. 41). In sum, he concludes that “[w]e willneed to abandon the model, its assumptions, and all that it entails” (p. 43).

The subsequent chapters deal with various aspects of actual language use.Chapter 3, “Words and their behaviour”, questions the generative model’sassumption that all items in the lexicon can be assigned to a small number ofcategories and that all members of a (lexical) category behave identicallywith respect to syntactic rules. He shows that a number of words -- e.g.“fun”, “much”, or so-called defective verbs such as “beware”, cf. “*I alwaysbeware of dogs” -- exhibit a unique distribution that cannot be predicted fromtheir membership in a syntactic category. These findings suggest “thatknowledge of a language can only be attained through exposure to actual usageevents, whose particularities are noted and laid down in memory” (p. 68).

Chapter 4, “Idioms”, and Chapter 5, “Speaking idiomatically”, discuss atlength the pervasiveness of idiomatic language use. While the generative modeltreats idioms as peripheral exceptions to the otherwise rule-governed use oflanguage, “[t]here comes a point [...] when the periphery looms so large thatit may no longer be an option to regard it as peripheral” (p. 43). In Chapter4, Taylor presents corpus findings on semantic (e.g. “kick the bucket”) aswell as syntactic (e.g. “by and large”), lexical (e.g. the usage range of“fun”), and phrasal idioms (e.g. “of course”, “in fact”, “under way”). InChapter 5, Taylor proposes to understand idiomaticity in terms of conformitywith usage norms. On the one hand, this concerns the appropriateness oflinguistic expressions to the respective context of use (e.g. the formality ofa situation as well as the medium of communication). On the other hand, itconcerns language-internal relations as manifested by preferred collocations(e.g. “merry Christmas”, but “*merry birthday”). His case study of “X-minded”then shows that such usage norms, which are governed by “language-internalstatistics” (p. 114), are subject to diachronic change. While this compoundpattern enjoyed a peak of popularity in the middle of the 20th century, itsuse declined afterwards. He concludes that the usage norms are notcategorical, but statistical in nature.

Chapter 6 is dedicated to the controversial notion of “constructions”. After abrief overview of Langacker’s (1987, 1991, 2008a) Cognitive Grammar, hecontrasts two approaches towards defining a construction and then proposes athird definition incorporating elements from both approaches. On the one hand,constructions can be conceived of as internally complex entities, i.e. “anylinguistic form which can be analysed into its parts” (p. 124). On the otherhand, constructions can be defined as pairings of form and meaning. Innarrower approaches (e.g. Goldberg 1995), however, only those form-meaningpairings that have unit status or even only those whose properties cannot bederived from the properties of any other construction are regarded asconstructions proper. Taylor’s own proposal conceives of constructions asbasically synonymous with the notion of “unit” in Cognitive Grammar, i.e. “anyelement of a language that has been learned and that forms part of a speaker’slinguistic knowledge” (p. 126). He then discusses the differences betweenrule-based and construction-based approaches. Crucially, a construction-basedapproach tolerates a certain amount of redundancy, predicting that evenregularly formed expressions may be stored in memory rather than beinggenerated in every single case (cf. Langacker’s discussion of the “rule/listfallacy”, e.g. Langacker 1987: 29f., 492). This view is substantiated by theresults of experimental studies such as lexical decision tasks. Furthermore,Taylor contrasts the autonomy of syntax hypothesis with constructional andcollostructional (e.g. Stefanowitsch and Gries 2003) analyses, concluding that“even the most general syntactic patterns of a language [...] need to beregarded as constructions” (p. 145).

Chapter 7 deals with the key notion of “Frequency”. Taylor reviews a broadvariety of research indicating that speakers know, at least implicitly, therelative frequency of all elements of their language. His discussion ofChomsky’s well-known “Dayton Ohio argument” reveals that the allegedcorrespondence between corpus attestations and real-world facts only pertainsto a relatively small number of cases. While the frequency of “I live in NewYork” is indeed not only higher than the frequency of “I live in Dayton,Ohio”, but even corresponds almost exactly to the number of inhabitants of NewYork City in relation to Dayton, Ohio (cf. Stefanowitsch 2005), the same isnot true for the relative frequencies of “He lives in New York” and “She livesin New York”, respectively. While it can reasonably be assumed that thedistribution of both genders should be roughly equal, the former sentence isabout twice as frequent as the latter. Further frequency-related issues thatTaylor discusses include collocation patterns (for example, the word“unmitigated” is highly probable to be followed by “disaster”, but not viceversa), prevalent phonological patterns, and ambiguity resolution. Hisdiscussion of these matters leads him to conceive of “Skewed frequencies as adesign feature of language”, which is the title of Chapter 8. Rather thanbeing a mere side-effect emerging in any language-like system, skewedfrequencies contribute to the learnability of language in that they facilitatelinguistic categorization at all levels of language structure and use.

The question of language acquisition is further elaborated on in Chapter 9,“Learning from input”. Focusing on phoneme acquisition, Taylor argues thatlanguage learners record the statistical properties of the input, which leadsto the emergence of phonetic categories. He cites the so-called “recencyeffect” as evidence that people are registering the properties of the languagethey encounter: As both experimental and corpus studies show, the choicebetween linguistic alternatives is often determined by the immediatelypreceding context. For example, if a speaker uses a comparative constructionwith “more”, her interlocutor is likely to do so as well, even if forming thecomparative with “-er” would be the more natural choice in other contexts.

Chapter 10 addresses one of the most widely-discussed topics of CognitiveLinguistics, namely, polysemy. Contrary to most accounts, Taylor suggests thateach word should not be associated with a fixed number of discrete meanings,but that word meaning should be thought of in terms of the ways in which alinguistic form can be used, i.e. seeing words in their respective “contextualprofile”. On this account, the variability and dynamicity of meaning thatconstitute polysemy can be seen as a consequence of its embeddedness inlanguage use. This hypothesis is substantiated by the finding that theacquisition of prepositions rarely follows the pattern predicted bytraditional network accounts of polysemy: Irrespective of their most centralsense, prepositions are mostly acquired as parts of fixed expressions, whichmay vary from individual to individual (cf. Rice 2003). This lends support toTaylor’s overall hypothesis in that it demonstrates that a word cannot beconceived of as a fixed entry in the mental lexicon but rather “providesaccess, not only to the conceptual domains against which it is understood, butalso to the linguistic contexts in which it has been used” (p. 244).

Chapter 11 deals with the notions of “Creativity and innovation”, which play acentral role in the classic generative model. While creativity, in thegenerative sense, refers to the application of grammatical rules to lexicalitems, the notion of innovation refers to the coinage of new words orexpressions. Taylor shows this distinction to be problematic in many respectsas it proves to be difficult to apply in many cases. In the case ofreanalysis, for example, language change comes about not through innovation byan individual speaker, but through the hearer’s interpretation of thespeaker’s utterance. Case studies of progressive “busy” as attested primarilyin South African English (“My essay is busy being typed”), ditransitive“explain” (“explain me it”), and constructions involving “all over” illustratethat “innovation is incremental” (p. 262). Patterns that emerge as minorvariants, deemed ungrammatical by most language users, gain acceptance whenthey become more frequent.

As a model to account for manifestations of linguistic creativity andinnovation, Taylor proposes “Blending” (cf. Fauconnier and Turner 2002), whichis the topic of Chapter 12. After a brief outline of blending theory, Tayloraddresses the phenomena of word blending and phrasal blending both in speecherrors (e.g. “that’s torrible”) and in creative language use (e.g.“glitterati”, a blend of “glitter” and “literati”, cf. Kemmer 2003). As theemergence of the plural form “process[i:]s” (in analogy to the plural forms ofsome words of Greek origin such as “thesis”) demonstrates, blending can alsoaffect inflectional morphology. In the domain of phrasal blending, idioms suchas “time and again” can be analyzed as blends of two (or more) distinctphrases (e.g. “time after time” + “again and again”). He concludes thatblending not only “is able to do the work traditionally assigned to generativerules” (p. 279) but also accounts for linguistic creativity and innovation.

The final chapter, “The mental corpus”, summarizes the main hypotheses of thebook: 1) Linguistic knowledge emerges from exposure to actual language use. 2)Knowing a language means much more than just knowing the words of a language.Instead, language is organized in terms of constructions: “Knowledge of wordsand knowledge of constructions cannot easily be teased apart” (p. 282). Alarge part of what constitutes a language has to be considered idiomatic. 3)Frequency of occurrence is highly important with regard to the emergence ofusage norms in a language. Speakers know, at least implicitly, the frequencyprofile of linguistic items; moreover, skewed frequencies can be considered adesign feature of language. 4) The mental corpus does not (only) consist of anumber of fixed words and expressions. Instead, speakers are capable ofgeneralizing over the data they encounter. This facilitates languagecomprehension in that it provides schemas against which utterances can beunderstood. Furthermore, it lays the ground for linguistic creativity andinnovation. 5) This has important bearings for a theory of languageacquisition, which has to be conceptualized as an input-driven, bottom-upprocess. Considering the dialectic relation between I-language and E-language,acquisition can also be seen as a life-long process. 6) Importantly, thesehypotheses are substantiated by both corpus studies and experimental findings.


In the tradition of approaches to language that highlight the importance ofusage frequency (e.g. Langacker 1988, 2000; Bybee 2007), Taylor makes aconvincing case for a radically usage-based account of linguistic knowledge.While the foundational works of Cognitive Grammar often relied on inventedexamples, almost all the examples Taylor presents are derived from corpora.Furthermore, Taylor takes into account a broad variety of experimentalstudies. This makes his book a major step towards an empirically groundedcognitive-linguistic theory of language.

Taylor maintains the highly readable style that already characterizes histextbooks on Cognitive Grammar (2002) and linguistic categorization ([1989]2003) without ever oversimplifying matters. Therefore his book is equallyrecommendable to students of linguistics, advanced scholars, and eveninterested laypersons.

What is more, his “mental corpus” hypothesis ties in neatly with currentapproaches in cognitive linguistics and also has important implications forsome issues that lie outside of the scope of Taylor’s monograph. For example,his hypothesis is highly compatible with approaches that view language as acomplex adaptive system (CAS, e.g. Beckner et al. 2009), a framework that hasalso been proposed for research in historical cognitive linguistics (Frank andGontier 2010). As both the “mental corpus” hypothesis and the CAS approachdraw attention to the “intrinsically diachronic aspect of language as asystem” (Frank and Gontier 2010: 48), his work also has implications forhistorical linguistics in that it highlights the role of usage and frequencyin language change. Moreover, Taylor’s observations, especially thoseconcerning the paramount importance of idiomatic knowledge for a speaker’slanguage proficiency, can prove fruitful for foreign language teaching (cf.also Langacker 2008b: 84).

All in all, Taylor’s book is highly recommended for anyone interested inusage-based theories of language and linguistic knowledge. It offers valuableinsights not only to cognitive linguists and corpus linguists, but also tohistorical linguists and second language teachers.


Beckner, Clay; Blythe, Richard; Bybee, Joan; Christiansen, Morten H.; Croft,William; Ellis, Nick C.; Holland, John; Ke, Jinyun; Larsen-Freeman, Diane;Schoenemann, Tom. 2009. Language is a Complex Adaptive System. Position Paper.Language Learning 59 Suppl. 1, 1–26.

Bybee, Joan. 2007. Frequency of Use and the Organization of Language. Oxford:Oxford University Press.

Chomsky, Noam. 1986. Knowledge of Language. Its Nature, Origin, and Use. NewYork: Praeger.

Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press.

Fauconnier, Gilles; Turner, Mark. 2002. The Way We Think. Conceptual Blendingand the Mind’s Hidden Complexities. New York: Basic Books.

Frank, Roslyn M.; Gontier, Nathalie. 2010. On Constructing a Research Modelfor Historical Cognitive Linguistics (HCL). Some Theoretical Considerations.In: Winters, Margaret E.; Tissari, Heli; Allan, Kathryn (eds.): HistoricalCognitive Linguistics. Berlin, New York: De Gruyter (Cognitive LinguisticsResearch, 47), 31–69.

Goldberg, Adele E. 1995. Constructions. A Construction Grammar Approach toArgument Structure. Chicago, London: The University of Chicago Press.

Jackendoff, Ray. 2002. Foundations of Language: Brain, Meaning, Grammar,Evolution. Oxford: Oxford University Press.

Kemmer, Suzanne. 2003. Schemas and Lexical Blends. In: Cuyckens, Hubert; Berg,Thomas; Dirven, René; Pather, Klaus-Uwe (eds.): Motivation in Language.Amsterdam: John Benjamins, 69-97.

Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol. 1:Theoretical Prerequisites. Stanford: Stanford University Press.

Langacker, Ronald W. 1988. A Usage-Based Model. In: Rudzka-Ostyn, Brygida(ed.): Topics in Cognitive Linguistics. Amsterdam; Philadelphia: JohnBenjamins (Amsterdam Studies in the Theory and History of Linguistic Science,50), 127–161.

Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Vol. 2:Descriptive Application. Stanford: Stanford University Press.

Langacker, Ronald W. 2000. A Dynamic Usage-Based Model. In: Barlow, Michael;Kemmer, Suzanne (eds.): Usage-based models of language. Stanford: CSLIPublications, 1-63.

Langacker, Ronald W. 2008a. Cognitive Grammar. A Basic Introduction. Oxford:Oxford University Press.

Langacker, Ronald W. 2008b. Cognitive Grammar as a Basis for LanguageInstruction. In: Robinson, Peter; Ellis, Nick C. (eds.): Handbook of CognitiveLinguistics and Second Language Acquisition. London: Routledge, 66-88.

Rice, Sally. 2003. Growth of a Lexical Network. Nine English Prepositions inAcquisition. In: Cuyckens, Hubert; Dirven, René; Taylor, John R. (eds.):Cognitive Approaches to Lexical Semantics. Berlin: De Gruyter, 243-280.

Stefanowitsch, Anatol. 2005. New York, Dayton (Ohio), and the raw frequencyfallacy. Corpus Linguistics and Linguistic Theory 1, 295-301.

Stefanowitsch, Anatol; Gries, Stephan Th.. 2003. Collostructions:Investigating the Interaction of Words and Constructions. InternationalJournal of Corpus Linguistics 8, 209-243.

Taylor, John R. 2002. Cognitive Grammar. Oxford: Oxford University Press.

Taylor, John R. [1989] 2003. Linguistic Categorization. 3rd ed. Oxford: OxfordUniversity Press.


Stefan Hartmann is a PhD student in historical linguistics at the Universityof Mainz, Germany. He is currently conducting a corpus-based study on thediachronic change of German nominalization patterns. Apart from historical andcorpus linguistics, his research interests include Cognitive Linguistics,sociolinguistics, and psycholinguistics.

Page Updated: 08-Jan-2013