Book announced at https://linguistlist.org/issues/34.2398

EDITOR: Manuel Diaz-Campos
EDITOR: Sonia Balasch
TITLE: The Handbook of Usage-Based Linguistics
SERIES TITLE: Blackwell Handbooks in Linguistics
PUBLISHER: Wiley
YEAR: 2023

REVIEWER: Geoffrey Sampson

SUMMARY

The title of this book might puzzle a newcomer to linguistics: what else could language description be based on, if not usage by speakers and writers? But any approach to describing a language involves a tension between the linguist’s desire to find neat patterns in the material, and the messy lack of neatness in the raw data; and until very recently, much of the data was sufficiently inaccessible that the urge to impose neatness arguably had far too free rein. Speech is all around us, but it vanishes as fast as it is uttered. Written language lasts, but unless it is available in electronic form it cannot easily be searched for items of interest. For most of the history of linguistics, computers were not available; sound-recording was, but recordings of speech are not much use to the linguist before they have been time-consumingly and expensively transcribed, and then the same problem about search arises. On the whole, linguists proceeded as best they could with little help from technology.

Usage-based linguistics is a newish approach to the subject, which holds that the neat arrays of phonemes and grammatical rules which emerged from that style of working were misleading caricatures of language. Nothing we know about the workings of the human mind requires us to believe that language acquisition involves abstracting structural generalizations from the utterances we encounter and then discarding all of the other detailed properties of the utterances: perhaps our minds retain individual utterances in their full phonetic and lexical detail – as Joan Bybee puts it on p. 13 here, “tokens of linguistic experience are retained in memory along with their context”. Bybee adds “and [are] grouped together with similar experiences to form patterns or generalizations”, but Dagmar Divjak and Petar Milin, on p. 317, urge that even when some abstract generalization about the structure of a language is in fact correct, we are not entitled to infer that the generalization is part of speakers’ mental model of the language. (Divjak and Milin do not address the question of how an abstract structural feature of a language could maintain itself if it were not a reality for speakers, but the question is perhaps not unanswerable.) Even if not every detail of every utterance is remembered, heard utterances must build up a constantly-updated store of frequency information about the forms they contain – usage-based theorists cite abundant evidence that frequencies affect speakers’ output, for instance unstressed syllables are more likely to be phonetically reduced when the words they occur in are common words. Indeed, Clay Beckner (p. 339) says that our minds must be tracking different statistical properties of the same forms simultaneously. Quoting Stefan Gries’s chapter, “frequency [is] not just an arbitrary performance phenomenon but a crucial component of the inner workings of mental grammar”. Now that computers have begun to make detailed language data accessible on an unprecedented scale, it is becoming easy to find evidence against the neat linguistics of the twentieth century.

Work reflecting ideas like these has been cropping up here and there for several decades now. The term “usage-based linguistics” pulls together what have been separate individual objections to the reigning paradigm, and efforts to replace that paradigm in particular areas of language structure (e.g. “Construction Grammar” with respect to syntax), into a fairly consistent and comprehensive attempt to overturn the entire paradigm. (The term “usage-based” itself is over twenty years old. The coinage is attributed here to Ronald Langacker 1987: 494, and several contributors cite Langacker 2000 and Tomasello 2003.)

The book comprises 31 numbered chapters by a total of 47 contributors (there is also an unnumbered editorial introduction, which as usual in this kind of book does little more than summarize the structure of the book and list acknowledgements). It would take too much space to list all the chapter titles, but the chapters are grouped into six Parts:
Part I: Overview
Part II: Phonology and the Usage-Based Approach
Part III: Morphosyntax and the Usage-Based Approach
Part IV: Psycholinguistics, Language Development, and the Usage-Based Approach
Part V: Variation, Change, and the Usage-Based Approach
Part VI: The Future of the Usage-Based Approach

The real “introduction” to the book is Chapter 1, “What is usage-based linguistics?”, by Joan Bybee of the University of New Mexico (a well-regarded linguist who is cited by some other contributors in terms suggesting that she has perhaps been a mentor figure). It offers a clear summary of the new paradigm and the nature of its contrasts with the established linguistic tradition, and the chapters that follow enlarge on particular aspects of the usage-based paradigm. Thirty-three of their contributors are based in the USA (quite a number of them, like Bybee, in its Mountain timezone); eleven work in Europe, and two in Brazil.

It seems fair to say that the book as a whole, while it gives the reader an explicit and rather complete understanding of what usage-based linguists see as wrong with the established linguistic tradition, does no more than adumbrate the new kind of linguistics which they hope to see replace it. We are not given anything like a skeleton description in the usage-based style of some individual language. But this is not a criticism, it is a reflection of the current state of the field. I believe it is usual that, when one Kuhnian paradigm replaces another (assuming that Kuhn’s 1962 picture of “scientific revolutions” has some validity), the new paradigm will at first be fairly vague and will only gradually be fleshed out into a clearcut theory.

It is not too early, though, to foresee some rather drastic implications for linguistics as an institution, if the usage-based paradigm does win.

For one thing, until now individuals who advance the discipline of linguistics have typically begun as students educated mainly in the humanities, who have encountered linguistics in connexion with studying a language or languages and found it interesting enough to want to pursue it as a specialization. But the centrality of frequency to the usage-based approach (Mark Hoff remarks on p. 200 that “While usage-based linguistics can be subdivided many ways … , frequency is key to them all”) means that contributing to this approach must involve heavy-duty statistical reasoning and calculation, as exemplified for instance in Baayen (2001) – a number of contributors here quote Harald Baayen’s work. In my experience, humanities students tend not only to lack familiarity with statistics (or indeed with any branch of maths), but to be very clear that they have no wish to become better acquainted. (And Divjak and Milin point out, p. 307, that the skills needed for the computational modeling which is becoming an important part of usage-based research may pose an even greater barrier.) So there will be an issue about where recruits to the new linguistics will come from.

Also, it is apparent from the contents of this book that there will be far more to the description of a language under the usage-based paradigm than was envisaged by traditional linguistics. It will be as if a schematic road map of a territory were replaced by documentation specifying exact width and camber of each stretch of road, the nature of their surfaces and state of repair, and so on. Consequently, becoming a useful member of the linguistics profession will be more like becoming a practicing chemist or engineer, where a student may have pursued the subject throughout his undergraduate career and as a postgraduate, but even at the postdoc stage is still an apprentice. Our societies accept this long learning period in the case of disciplines like chemistry and engineering, because voters can see the material benefits which flow from the existence of bodies of people competent in those fields. It is not so easy to identify what benefits to society at large will accrue if we have thoroughly adequate descriptions of our languages. The most obvious possibility would perhaps be to eliminate some of the remaining performance limitations, relative to human language performance, of software of the ChatGPT type – but in view of the consternation already aroused by ChatGPT, that might not be a strong selling point for the new linguistics.

All this does not, obviously, imply that usage-based theory is mistaken. But, if it is essentially correct, that might mean that linguistics has a limited future. Divjak and Milin comment on p. 309 that “Some in the cognitive sciences would claim that linguistics is becoming obsolete”.

EVALUATION

For anyone wishing to get to grips with this new “usage-based” paradigm, Díaz-Campos and Balasch’s book makes an excellent starting-point. It covers just about every aspect of language (not only different levels of structure, such as phonology, morphology, and so on, but topics such as language acquisition, aphasia, and others – there is not much about semantics, but then semantics is an area where the traditional, “neat” paradigm is weak, and so offers little to overthrow). And the book includes abundant references to other publications, making it easy to follow up any issue of particular interest.

On the other hand, not all the contributors are of Joan Bybee’s calibre. An impression given by some chapters is that their authors have learned and reproduced the lessons taught by the leaders of the movement, but lack a deep enough background knowledge of languages and linguistics to build successfully on those lessons. The book is sometimes repetitive: I lost count of the number of contributors who made the (not particularly surprising) point that syllables are more commonly reduced in high-frequency words, often using the same examples as one another – ‘camera’ and ‘memory’ as high-frequency words, ‘mammary’ as a low-frequency word. Several chapters attempt to define the standard distinction between “type” and “token” (pp. 200, 291, 514) but do so in a fashion I find incomprehensible. I suspect that what has happened there is that some usage-based theorist has stretched the “type” concept to apply it to a special issue within the theory, and these authors have not understood that this is a specialized use of “type”, not the simple, Peircian distinction.

One way in which the shallowness of some contributors’ acquaintance with linguistics betrays itself is that they equate the “traditional” paradigm with Chomskyan Universal Grammar; for instance on p. 511 Díaz-Campos and Matthew Pollock write “A usage-based approach to the study of linguistic phenomena contrasts with the more traditional approach based on Universal Grammar”. (William Labov may be surprised to read on p. 510 that his concept of variation in language “is rooted in the linguistic models advanced by Chomsky”.) Several chapters contain suggestions that if Chomskyan generative linguistics can be shown to be wrong, traditional linguistics must be wrong and the usage-based paradigm will triumph as the last man standing.

This attitude feels particularly weird in connexion with the objections expressed in several chapters to the concept of the phoneme. Earl Kjar Brown (p. 132) quotes Bybee as saying “ ‘Phonemes’ do not exist as units”; Michael Gradoville (p. 533) refers to “the category traditionally thought of as the phoneme /s/”, to emphasize that he himself does not believe in phonemes. Jessie Nixon and Fabian Tomaschek, in a chapter entitled “Does speech comprehension require phonemes?”, cite definitions of the phoneme concept by various early-twentieth-century structural linguists and continue: “Jakobson, Fant and Halle (1952) described the phoneme within the structuralist perspective as being composed of a bundle of binary distinctive features”, adding “Chomsky and Halle … take the idea of binary distinctive features further in ‘The Sound Pattern of English’, arguing for a finite, ‘universal set of phonetic features’ ”. But, over the decades during which phonemic analysis was the central focus of linguistics (represented for instance by Martin Joos’s 1957 collection ‘Readings in Linguistics’), the point of it was that this (and other structural features of languages) were held _not_ to be based on universals. The significance of linguistics as a discipline was agreed to be that it demonstrated how diverse human cultural institutions can be. Morris Halle was the man who persuaded much of the discipline, on the basis of data from Russian (Halle 1959: 22–3), that the phoneme concept does not stand up. Arguing against Halle, as Nixon and Tomaschek do, is certainly not arguing against the phoneme concept.

To claim that there are no phoneme units is ambiguous. It might mean that we should not consider, say, the English phoneme /t/ to be an atomic entity, but as shorthand for a combination of a distinctive value on the scale of voice-onset timing, a distinctive area of tongue–upper jaw contact, and perhaps other feature(s). That is a very reasonable idea, though it does not imply that the various distinctive features used by the English language coincide with those used by other languages – traditional phoneme theory would predict few such coincidences. But Nixon and Tomaschek are clear that this is not what they mean: “Changing the type of unit from phoneme to something else did not solve the problem.” For them, it seems, there are only mental records of individual utterances, some chunks of which may cluster near one another in regions of perceptual space.

I do not see how this radical rejection of the phoneme idea can work. Suppose in reading I encounter a word I do not know, say the word ‘chitin’. I will not be sure how to pronounce it, perhaps /ʧɩtɩn/ or /ʧaitɩn/ or (the correct pronunciation) /kaitɩn/. But there will be only a few possibilities, and if I look the word up in a dictionary which shows pronunciations in some essentially phonemic notation, perhaps as ‘kītin’, I shall be fully competent to pronounce it. If I talk about chitin to hearers familiar with the concept, they may well detect that I don’t know much about chitin, but they will not be able to tell that I have never heard the word spoken. How would this be possible, if something like a finite set of target sounds were not a mental reality for speakers?

In one important respect, Joan Bybee is herself guilty of conflating Chomskyan linguistics with traditional structural linguistics. In the very first paragraph in which she discusses linguists of the past (p. 2), she states that Saussure’s distinction between ‘langue’ and ‘parole’ was “roughly” the same as Chomsky’s “competence v. performance” distinction (and other contributors make similar remarks). Saussure’s and Chomsky’s distinctions were not at all the same. Saussure was simply distinguishing examples of a language from the general linguistic system to which the examples conform. Saussure did acknowledge, via a musical analogy, that speakers sometimes make mistakes (Bally and Sechehaye 1959: 18), but this was by no means central to his discussion. For Chomsky, it is a crucial claim that speakers’ utterances do _not_ conform to their mental competence – their performance is far too “degenerate”, “fragment[ed] and deviant” (Chomsky 1965: 13, 201) for infants to use it as a basis for inferring their elders’ underlying linguistic system, and hence he believes that much knowledge of language must be innate and hence universal. Chomsky’s claim about speakers’ typical output is of course highly controversial. In the context of a paradigm to which “usage” is central, it is surely important to be clear about whether or not one is endorsing that claim.

Apart from these theoretical issues, contributors to the book not infrequently make factually misleading statements. Javier Rivas (p. 480) tells us that the concepts “subject, predicate, object” were introduced into grammatical description in the eighteenth century as a consequence of the loss of Latin case marking in the Romance languages, which forced grammarians to turn their attention from morphology to syntax. This is perplexing. I am not sure about “object”, but “subject” and “predicate” come straight from Aristotle, who in the Middle Ages was seen as having said the last word that needed to be spoken on issues of grammar and logic. (The word ‘sub‑jectum’ was coined as a rendering into Latin of Aristotle’s ‘to hypokeimenon’, “what is placed under [one’s eyes]”.) These ideas were part of the Trivium, the basic (“trivial”) topics which people learned before moving on to the hard stuff in the Quadrivium, and they learned them through the medium of Latin with its full panoply of case-endings. A few paragraphs later, Rivas mentions that Japanese lacks verb morphology. Japanese has plenty of verb morphology.

Clancy Clements and Jordan Garrett claim (p. 63) that “English does not allow the absence of an overt subject”. The particular example they quote, ‘Am singing’, is admittedly unlikely, because it saves no syllables relative to the version with subject, ‘I’m singing’. But as a generalization about the grammar of colloquial English it is quite wrong: subjectless clauses are often heard. (A famous series of examples were produced by Clement Attlee, British prime minister in the late 1940s, when John Strachey, as a matter of form in view of the convention that government ministers must clear publications with their leader, asked for permission to publish a book of poetry, and was disconcerted when Attlee responded “Can’t publish. Don’t rhyme, don’t scan.”)

The person who realized that “Zipf’s Law” needed the addition of extra variables to make it empirically adequate was not András Kornai in 2008, as Chad Howe suggests on p. 264, but Benoit Mandelbrot (see Mandelbrot 1965). (Incidentally, the version of the Zipfian formula shown by Howe is virtually unreadable because of non-use of algebraic italics and spacing, and hyphen in place of minus sign.) On p. 215, Florent Perek says that linguistic knowledge comprises a “vast network” of constructions, and reasserts the concept of a “network approach” on a later page; “network” is a numinous word, but when one seeks a concrete example, the only example given (Fig. 12.2, p. 222) is not a network but a tree, a quite different structure. (The characteristic of a network is that between any two nodes there are typically many alternative paths – hence the choice of network geometry for the Internet, to make it robust against enemy action – whereas a tree provides only a single path between any two nodes.) And it was not sociolinguists but Hermann Paul and other Neogrammarians in the late nineteenth century who held that sound changes apply regularly across the board (p. 456).

I would also query Howe’s remark (p. 249) that the original British National Corpus from 1995 serves as any kind of “gold standard”, so far as its “demographically-sampled speech” section is concerned. The speech is very inaccurately transcribed, and the demographic data on speakers are full of gaps and obvious errors. Its 2014 replacement, see <cass.lancs.ac.uk/cass-projects/spoken-bnc2014/>, is on an altogether different and higher plane in terms of scholarly standards.

For a group of academics who lay such emphasis on fine details of linguistic usage, it is surprising how clunky some of their own prose is. “These generative approaches generally ascribe to a modular approach to language” (p. 60) – is “ascribe” being used for “subscribe”? “… studies of grammaticalization have led researchers to consider Usage-Based Theory more than any other emergent phenomena” (p. 233) – it was quite a while before I grasped that the author meant “… studies of grammaticalization, more than any other emergent phenomena, have led researchers …”. Pp. 269–70 refer to “the Poverty of the Stimulus, or the limited amount of input that would fail to account for users’ language abilities” – this seems to be a truly tortuous way of saying something like “the fact that the amount of input is too limited to account for …”, but if I had not already known what generative linguists mean by the phrase “poverty of the stimulus”, I could never have unravelled this passage. (In any case, surely the “poverty of the stimulus” idea is not about _quantity_ of input, but about – alleged – total absence of certain kinds of input?)

The book is well produced, with few misprints – though p. 43 includes a long quotation which is not clearly distinguished from the surrounding prose; on p. 263 “headed” must I think be an error for “heeded”; the topmost entry in the “4-grams” column of Table 18.1, p. 329, is not a 4-gram; and there is a curious passage on pp. 514–16 where there are many instances of an arrow character used for “becomes”, and each such character is immediately followed by “x02794;” (the hexadecimal Unicode codepoint for the arrow).

With a book of 600 pages, one relies on the index to find one’s way back to passages which one has read without making notes of page-references. I may have been unlucky with the two items I tried to check back on: I was defeated in the case of Pawley and Syder (1983) (cited in at least two chapters here, and a rare case in my experience of a publication emanating from the “Applied Linguistics” industry which is well worth reading) by neither author’s name appearing in the index, while with “construction grammar” I hit an endless loop – “construction grammar, see grammar”; “grammar, construction, see construction”.

This is a valuable book, which unveils (I believe for the first time) the full scope of a movement whose adherents see it as destined to sweep aside much or all of what we think we know about language. But some may feel that we need the advocates of the new linguistics to temper their early enthusiasm with more in the way of mature and sober scholarship, before we will be able to judge how seriously we must take the “usage-based” paradigm.

REFERENCES

Baayen, R.H. 2001. Word Frequency Distributions. Kluwer.
Bally, C. and A. Sechehaye, eds. 1959. Ferdinand de Saussure: Course in General Linguistics, tr. W. Baskin from 1916 original. Philosophical Library (New York).
Chomsky, A.N. 1965. Aspects of the Theory of Syntax. MIT Press (New York).
Chomsky, A.N. and M. Halle. 1968. The Sound Pattern of English. Harper & Row.
Halle, M. 1959. The Sound Pattern of Russian: a linguistic and acoustical investigation. Mouton (the Hague).
Jakobson, R., C.G.M. Fant, and M. Halle. 1952. Preliminaries to Speech Analysis: the distinctive features and their correlates. Acoustics Laboratory, Massachusetts Institute of Technology.
Joos, M., ed. 1957. Readings in Linguistics. American Council of Learned Societies (New York).
Kuhn, T.S. 1962. The Structure of Scientific Revolutions. University of Chicago Press.
Langacker, R.W. 1987. Foundations of Cognitive Grammar, vol. 1. Stanford University Press.
Langacker, R.W. 2000. “A dynamic usage-based model”. Pp. 1–63 in M. Barlow and S. Kemmer, eds, Usage-Based Models of Language. CSLI Publications (Stanford, Calif.).
Mandelbrot, B.B. 1965. “Information theory and psycholinguistics”. Pp. 550–62 in B.B. Wolman and E. Nagel, eds, Scientific Psychology: principles and approaches. Basic Books.
Pawley, A. and F.H. Syder. 1983. “Two puzzles for linguistic theory: nativelike selection and nativelike fluency”. Pp. 191–226 in J.C. Richards and R.W. Schmidt, eds, Language and Communication. Longmans.
Tomasello, M. 2003. Constructing a Language: a usage-based theory of language acquisition. Harvard University Press.

ABOUT THE REVIEWER

Geoffrey Sampson graduated in Chinese Studies from Cambridge University, and his academic career was spent partly in Linguistics and partly in Informatics, with intervals in industrial research. After retiring as professor emeritus from Sussex University in 2009, he spent several years as Research Fellow at the University of South Africa. He has published contributions to most areas of Linguistics, as well as to other subjects. His latest book is "The Delusion of Linguistics" (2017).