In general, we endorse what William A. Kretzschmar, Jr has set out to do in his book, Language and Complex Systems. We subscribe to a complex systems view of language. In fact, in the interest of full disclosure, we note that the title of his book addresses a topic that has been a focus of our ongoing research for many years. There is no doubt in our minds that an understanding of language as a complex system helps us think both differently and usefully about language on a number of levels, and that Kretzschmar is a fearless advocate of this approach (thus taking on some of the leading figures in linguistics today). Although his book appears to be more of a compilation of articles with different foci than a coherent narrative, Kretzschmar's straightforward style and broad scope make interesting reading, and we concur with some of his observations. At the same time, in this review we also identify instances where there seem to us to be conflation, contradiction and simplification.
In chapter 1 Kretzschmar provides principles and examples of complex systems, concluding the chapter with experimental evidence drawn from LAMSAS (Linguistic Atlas of Middle and South Atlantic States), which he puts forth as evidence of language as a complex system. Since this is the title of a book co-edited by one of us (Ellis & Larsen-Freeman Reference Ellis and Larsen-Freeman2009), we were keen to see why Kretzschmar had moved from arguing that speech is a complex system (in his 2009 book) to asserting the same for language in the present volume (p. 79). Does he mean to equate speech with language? If so, this should both be made explicit and justified. Along the same lines, he begins chapter 2 by claiming that empirical findings introduced in chapter 1 concerning language support the idea that speech is a complex system: which is it then to be? Speech or language? The two are not necessarily the same (cf. Saussure's distinction between langue and parole).
Also in chapter 2, Kretzschmar associates cause and effect reasoning with reductionism, writing that ‘[r]eductionism involves a commitment to cause and effect’ (p. 43). There may be a common association between reductionism and cause and effect argumentation, but the inverse is untrue. One can seek cause and effect relationships in complex systems without being reductionist. This might require a different view of causality (Larsen-Freeman & Cameron Reference Larsen-Freeman and Cameron2008), one that does not involve prediction and isolation of an independent variable, but rather traces changes that do result from earlier interactions, such as from the co-adaptation between interlocutors. We certainly support a non-reductionist approach; in fact, one of the core principles in complexity theory is the notion of emergentism, i.e. novelty arising from the interaction of the components of the system. However, there is no need to associate cause and effect reasoning solely with reductionism and to pit cause and effect against complex systems. We also question Kretzschmar's claim that ‘reductionism is a reversal of induction’ (p. 42). This is simply not true. Here Kretzschmar appears to conflate methodology with ontology. Induction is a method that can be, but need not be, employed in a reductionist way.
In chapter 2, Kretzschmar discusses the status of linguistics in modern times. He begins by contrasting the epistemology of biologists Edward O. Wilson (Reference Wilson1998), who identifies science with reductionism, and Stephen Jay Gould (Reference Gould2003), whom Kretzschmar points to as committed to emergence and contingency as addressed by complexity science. He goes on to trace the evolution of academia from antiquity to modern times, and he assails what he feels is a political position that the Linguistic Society of America (LSA) has taken in its alignment of linguistics with modern physical and natural science, as opposed to the humanities. He points out that the ‘humanities can be scientific just as well as the natural and physical sciences can’ (p. 52), and suggests that ‘the humanities is better characterized scientifically by complex systems’ (p. 53). We concur with this point and in fact would go a step further, maintaining that complexity theory as a metatheory is transdisciplinary and can be used to elucidate many phenomena across the disciplinary spectrum: humanities, social sciences, physical sciences, business, the health sciences, etc.
Larsen-Freeman's (Reference Larsen-Freeman2011) review of Kretzschmar's Reference Kretzschmar2009 book, The Linguistics of Speech, contended that Kretzschmar had not provided a proper overview of the relevant literature. He has, at least in part, redressed this in the present volume. Chapter 3 is devoted entirely to a review of more recent literature, most of which can be said to adhere to ‘usage-based’ linguistics. Kretzschmar begins his discussion with the work of Paul Hopper (Reference Hopper1987), and proceeds (with one exception) chronologically to discuss in turn publications of Hopper & Traugott, Bybee, Pierrehumbert, Goldberg, Tomasello, Ellis & Larsen-Freeman, and a later publication of Bybee. The thrust of the chapter is that Hopper's original position was the correct one and the rest have failed to get it right. His thesis is that Hopper characterizes grammar as a temporal phenomenon, with its structure being epiphenomenal, always contingent, always coming into being, always in process. He claims that, to their discredit, the others have moved away from a process view (however, cf. Larsen-Freeman's (Reference Larsen-Freeman2003) ‘grammaring’), by objectifying grammar and still striving to accommodate their ideas to those of formal linguistics, thus seeing grammar as a system instantiated in the mind as language data, instead of a dynamic process, or continual movement, ‘a process that never creates a product’ (p. 70).
Thus, among his criticisms of a usage-based approach is its alleged conformity to traditional linguistic categories of structure. However, we note that at the same time as he criticizes others for categorizing structures, he himself admits that ‘[w]e need to use categories so that we can count variant expressions and observe the nonlinear distributions’ (p. 78), an admission which seems to contradict his position and that which he has faulted others for taking. He also acknowledges the value of using the term ‘constructions’ employed by the usage-based linguists he has criticized ‘since the use of constructions does not entail a commitment to a fixed hierarchy of categories’ (p. 78). He further adds that ‘[t]here is no reason that we cannot use traditional terms like “noun” or “verb” or “ditransitive” to name them, as long as we do not invoke an entire hierarchical system when we do’ (p. 78). In a more recent publication even Hopper (Reference Hopper, MacWhinney and O'Grady2015) writes of ‘emergent noun phrases’, which exist as linguistic entities retrospectively.
Indeed, while we, too, reject the notion of a traditional ‘fixed’ grammar, we recognize the challenge in conceiving language as a process without alluding to some sort of pattern, which we have referred to as ‘patterns in the flux’ (Larsen-Freeman Reference Larsen-Freeman, Ortega and Han2017a) and even Kretzschmar himself invokes the term ‘stable patterns’ and claims ‘[f]or language, as for other domains in which complex systems operate, complex systems allow the stable patterns to emerge on which we rely for effective communications in our social interactions’. We agree, but then we wonder why he spends so much time criticizing others for discussing patterns. Indeed, it seems to us that you cannot talk about complex systems without patterns, ones that have to be recognizable retrospectively and corroborated emically.
In another apparent contradiction, Kretzschmar takes the Five Graces Group to task for referring to ‘the process of “selection” as the emergence of a preference for one state of the feature, one variant, over other possible variants’ (p. 74). However, he subsequently comments that ‘speakers as agents using speech are not dull bodies who have no choice in what features they want to use on any given occasion, including all of the different kinds of conversations and text types that people enact’ (p. 79). And, then, in chapter 6, he posits that speakers have ‘a pool of choices’ (p. 152). It is not clear to us that what Kretzschmar has written about speakers having choices differs at all from saying that speakers have the option to select one variant over another. Of course, they do not select the same variant for all time, and every variant is relevant for inclusion in the system, but on any given occasion, we would maintain that speakers do indeed select the variant that best suits their purposes.
Chapter 4 offers a way to assess different approaches to grammar, specifically through complex systems, and ‘to write better grammars when we apply the principle of scale-free networks to what we have learned from the 80/20 Rule’ (p. 100). As he did in his 2009 book, Kretzschmar promotes use of an A-curve (after the asymptotic shape of the curve) to plot nonlinear frequency distributions of linguistic features (although it should be noted that the A-curve is produced by linear recursion). According to him, a linguistic system as a fixed objective language hierarchy ‘originates as an observational artifact, something that we perceive to be there just because we usually do one of just a few things for any construction. Variants of features that are massively more common have a strong perceptual advantage because of the ever-present nonlinear distributional pattern, and they come to be perceived as “grammatical” or “acceptable”’ (p. 95). Interestingly, despite arguing vociferously for a different approach to grammar, he concludes chapter 4 by claiming ‘[t]he 80/20 Rule does not replace traditional grammars’ (p. 104). One thing that is not wholly clear from this last comment is where he stands with regards to linguistic theories, in particular, generative and structural linguistics. He appears to equivocate. On the one hand, he writes that the methods of studying complex systems do not replace other linguistic theories (p. 2), but on the other hand, he writes that ‘our paradigm for how to think about language has changed’ (p. 4). And, he takes the LSA to task for choosing ‘a scientific model that does not fit the facts of language’ (p. 43). Then, too, we find it curious that he writes, ‘[t]he Five Graces Group was wrong to make the tacit assumption of universal grammar, just as the generativists were wrong to do so’ (pp. 95–6). In fact, the usage-based position of the Five Graces Group is diametrically opposed to the notion of a universal grammar.
Chapter 5 investigates the history of the English language from a complex systems perspective. ‘For historians of the language, the key question is how to accommodate Hopper's sense of continual movement, and still be able to describe the grammar of the language at any moment in time’ (p. 109). We feel that this statement is the crux of the matter, and yet one that is perhaps not unlike the awareness in quantum mechanics that light behaves simultaneously as a particle and as a wave. The truth is that grammaring can be a process and grammar a product, depending on the perspective one takes.
Kretzschmar advises that ‘[f]or historians of the language, . . . “change” in the complex system of speech will consist of an alternation in frequency of any particular feature’ (p. 113), and he faults Labov for his ‘drive towards abstraction, to make generalizations in keeping with the Neogrammarian program, [which] must necessarily color his observation of real-time evidence and its difficulties of interpretation’ (p. 118). ‘The A-curve, on the other hand, offers historians of the language a way to judge their data when it is too sparse to construct its own frequency profile’ (p. 119). While the A-curve may well be more inclusive of features, even sparse ones, and is less abstract, it seems to us it still fails the process test since it is a representation at one point in time. Indeed, Kretzschmar himself acknowledges that we would need ‘a sequence of A-curves that showed the change in frequency of each realization over time’ (p. 129). He concludes the chapter with a discussion of Gould's punctuated equilibrium in evolution and offers the Great Vowel Shift in English as an example of the same complex system process in biology operating in language over time (although it seems to us that without specifying the evolutionary mechanisms responsible for the Shift, Kretzschmar is applying evolution metaphorically, something he has faulted others in the past for doing).
Chapter 6 addresses neural networks and complex systems. Kretzschmar writes ‘[w]e will thus be interested not in some state of the system after application of a rule or the evaluation of algorithms, but in the emergent pattern of the state space as it continues to change over time’ (p. 137). He discounts ‘taste’ for different solutions which arise from our socially conventional subjective perceptions, favoring ones that are consistent with the distributional patterns in the evidence. A prime example of this he suggests is our perception that there are separate dialects in which certain features cluster. According to him, the existence of separate dialects is not supported by the evidence available. He claims that dialects are ‘observational artifacts’, which cannot be validated by neural networks or other statistical processing of speech data (p. 150). After considering neural networks, self-organizing maps and dynamical systems, he concludes that connectionist models with massive parallel processing (neural network model or dynamical systems model) fit the evidence well, whereas the physical symbol system and neurolinguistics models do not.
Chapter 7 deals centrally with sociolinguistics and is where Kretzschmar makes his strongest case in the book, not surprisingly, given his professional identity. He argues that a few simple principles can give rise to frequency-based patterns (which Kretzschmar says are not the same as what we usually call grammar). While lauding Labov for his contribution to an understanding of variation in language, Kretzschmar opines that Labov's Class and Style Model is limited in that ‘no small, fixed set of coexistent systems could account for the multidimensional patterning of language in use from complex systems’ (p. 163). Instead, ‘[i]t is the nature of complex systems to create the same A-curve frequencies profiles for variants of speech for any population of speakers within the complex system’ (p. 166). We, too, have found Zipfian-like distributions to operate at various levels of scale in a given text and among texts of various sizes, even though the members that figure in the distribution are not the same (Larsen-Freeman Reference Larsen-Freeman2017b). It is the distributional pattern in the flux that endures. In any case, from a complex systems perspective, one would expect a dynamic equilibrium, balancing between accretion and attrition.
In discussing Schneider's (2007) book on the emergence of new varieties of English, he observes that ‘there is no fixed relationship between input frequencies from settler, indigenous, or adstrate languages and what will become most common in a new variety’ (p. 169). Again, we are in agreement, as we are with his warning earlier in the book concerning relying on large-scale corpora to inform cognitive development (p. 79), and for the same reason. What is available in a corpus is attested language captured at one period of time or over a period of time from diverse sources. Although frequency undoubtedly has a role to play in language change and acquisition, any frequentist argument also represents a reductionist simplification. Frequency may well be the starting point, but it is not the be all and end all. Users/learners of a language have the capacity to go beyond attested language. After all, language is a semiotic system (Herdina Reference Herdina and Schmid2017), not merely a compilation of variants already in circulation. Accordingly, Kretzschmar's remark (p. 170) that new varieties are emerging all around us every day makes sense as does his demonstration that the reason that speakers tend to think about variants in categorical terms is the perceptual salience of the top-ranked variant (p. 179). We are also in accord with the idea that by adopting a complex systems perspective, we can connect with researchers in other disciplines who also subscribe to the principles and practices of complex systems, by, for example using the Gini Coefficient from economics and public health to model linguistic data (pp. 181ff.).
In chapter 8, Kretzschmar makes the case for a postmodernist approach. ‘If we begin with the idea that speech around us is not necessarily well-ordered and coherent, we can look, in the postmodern mode, for those aspects of it that seem to show order and coherence for particular times and situations’ (p. 211). He goes on to write: ‘By using complex systems in our work we can build bridges from the humanities back to the natural and physical sciences, and that will be a welcome step toward the resolution of the problem of two cultures on our campuses and in our communities’ (p. 212).
While we cannot fault this claim, we do feel compelled to point out that post the arrival of quantum physics, the nature of the hard sciences themselves has been transformed such that the gap between the humanities and natural sciences can no longer be perceived as two different cultures (Hayles Reference Hayles1991). It also appears strangely inconsistent to refer to the obsolete two cultures argument (Snow 1959) in view of the fact that Kretzschmar cites Gould's (Reference Gould2003) The Hedgehog, the Fox and the Magister's Pox: Mending the Gap between Science and the Humanities, which dismisses the two cultures argument as misleading and unhelpful.
While we have written this review to point out what we feel are inconsistencies, we end as we started. This is an important book, weaving together many areas of linguistics usually thought of as separate, and doing so under the banner of complex systems, a cause to which we are also dedicated. It is a worthy cause indeed and one that has helped us and others to think differently about language. Nevertheless, we are still left with the impression that many a subject has just been touched upon and much still awaits exploration and explication.
