Language change and evolution are phenomena that have long fascinated linguists and non-linguists alike. Any theory of grammar must account for the fact that language is, by its very nature, in a continuous process of change. The flip side of change is language evolution, when new languages appear, sometimes ex nihilo, as in the case of new sign languages in Nicaragua and Israel. In his wide-ranging study, How new languages emerge, David Lightfoot shows that the emergence of novel grammars, while part of language change, is a phenomenon of much broader scope that may shed new light on well-studied cases of morphosyntactic change. The book presents a synthesis of Lightfoot's pioneering work over the past forty years and is an engaged manifesto for a new historical linguistics, which is conversant with such diverse disciplines as ‘grammatical theory, language acquisition, discourse analysis, and social variation in grammar’ (viii).

How new languages emerge comprises a brief preface, eight somewhat essayistic chapters, a bibliography and a subject index. Addressing a readership that does not necessarily have an extensive background in language science, it is lucidly written, without entering into too much technical discussion. Yet, while the introductory chapters are very accessible, the later chapters constitute heavier reading and the frequent excursuses into learnability issues interfere with a transparent exposition of the historical facts.

Lightfoot vigorously defends a strong acquisitionist view on language evolution, according to which new languages emerge because children construct novel grammars from a constantly changing linguistic input. This is essentially a corollary of the child-centered view on grammar change, which has been a ‘fruitful way of relating historical and theoretical syntax’ (Pintzuk, Tsoulas & Warner Reference Pintzuk, Tsoulas, Warner, Pintzuk, Tsoulas and Warner2000: 1), but which is not verifiable against the historical evidence. As far as newly-emerging languages are concerned, the example of urban youth languages in Africa (and elsewhere, cf. Kießling & Mous Reference Kim2004) clearly shows that adolescent speakers are the main driving force behind language evolution and development.

The generalist approach taken in this book has its merits in contributing to the book's comprehensibility, but it also leads to an oversimplified picture of the historical reality. The greater part of the structural analysis is situated within the Principles and Parameters framework, and Lightfoot assumes a tight connection between inflection and syntax, with structural shifts in grammar being linked to prior morphological change (102). The morphology-driven approach to (historical) syntax is empirically unsustainable and theoretically unwarranted, given that syntactic computation manipulates formal features on functional heads that may or may not correspond to inflectional material (see, for example, Bobaljik Reference Bobaljik2002, van Kemenade Reference Kemenade2007). More generally, the book suffers from a considerable degree of one-sidedness, dedicating little or no space to competing views on language acquisition and diachronic changes – a point raised by Michael Arbib and Gary Miller in their critical reviews of this book (see Arbib Reference Arbib2007, Miller Reference Miller2009). In what follows, I will present an overview of each chapter, accompanied by evaluative comments on its content, occasionally dwelling on recent developments in the fields of historical linguistic and in syntactic theory that are of immediate concern for the major claims of the book.

Chapter 1, ‘Internal languages and the outside world’, introduces the topic of the book and provides an initial characterization of the main theoretical constructs, to wit, the innateness of the human language faculty (Universal Grammar) and the distinctness of I(nternal) languages (or grammars) from E(xternal) languages. The point of departure of this introductory chapter is the simple truism that (E-)languages come and go. However, language birth is much harder to pin down than language death, which can occasionally be traced back to a historical event. A case in point is the extinction of the Tangut language after the Mongolian destruction of the Xīaxià empire in 1227 ce. No such comparable date exists for the development of French and Spanish out of post-classical Latin.

As part of the outside world, E-languages are seen as amorphous entities and in constant flux (12–13). Given such rich linguistic diversity, the human language capacity represents a constant, which is common to the species and has not changed over the period for which there are recorded data. From a biological perspective, there exists just one human language; the observed differences between languages are secondary to the general, universal properties (3–5).

A different stance is taken by Baker (Reference Baker2003: 349), who defines languages as finite systems of discrete differences. Accordingly, the human language capacity is not entirely invariant, but allows for a few basic parametric choices. Languages may thus vary considerably in the actual structures formed. According to Baker, the human language faculty takes an intermediate position between the perceptual and motor systems, in which the cognitive processes at work are essentially invariant, whereas such processes vary in arbitrary ways across populations in the domains of social and cultural cognition.

Chapter 2, ‘Traditional language change’, tackles the emergence of historical linguistics as an independent discipline in the nineteenth century. The historical overview focuses on two lines of research: the comparative method employed in the study of sound change, and the charting of the Indo-European family tree of genetic relationships. To present a more complete picture of nineteenth-century linguistics, it would have been worthwhile to discuss the work of August von Schlegel, Alexander von Humboldt and August Schleicher on morphological typology, whose classic expression was the threefold division of languages into isolating, agglutinative, and inflective, with an added fourth type of incorporating or polysynthetic languages (see Greenberg Reference Greenberg1974: 35–41). Given that the comparative method (supplemented with typology) prevails in phylogenetic linguistics, Lightfoot's negative conclusion that ‘the nineteenth-century search for deterministic laws for the history of languages failed’ (36) seems unjustified.

In this chapter, only a single paragraph is dedicated to grammaticalization, i.e. the semantic bleaching and reanalysis of a lexical item as a functional category (37–38). Lightfoot insists that grammaticalization is a phenomenon that needs to be explained rather than an explanation itself (37, fn. 5). To the best of my knowledge, the claim that grammaticalization is an explanation itself is not found in the grammaticalization literature. Haspelmath (Reference Haspelmath, Fischer, Norde and Perridon2004) shows that apparent cases of degrammaticalization can be reanalyzed as instances of backward shift or retraction, in which the more grammaticalized uses of lexical items fall out of use (see Reintges Reference Reintges, Crisma and Longobardi2009 for further discussion of the question as to whether grammaticalization is a historical process distinct from reanalysis).

Chapter 3, ‘Some properties of language organs’, elaborates on language design and the notion of I-language. I-languages are represented in an individual's brain and hence finite, although they range over an infinitude of data. Grammars are equipped to produce utterances of (in principle) infinite length by applying recursive devices like relativization, complementation and coordination (40–41). The second fundamental property of I-languages is compositionality, i.e. the deconstructability of complex linguistic expressions into hierarchically organized subunits (as modeled in tree diagrams). Lightfoot's claim that decomposable clausal structure is ‘a universal property of everybody's system and not a response to particular childhood experiences’ (43) seems very plausible. It is, however, more controversial to say that certain aspects of phrasal structure are subject to parametric variation vis-à-vis a headedness parameter (42–43), given the theory of antisymmetry presented in Kayne (Reference Kayne1994).

Much ink has been spilled over ‘the poverty of stimulus argument’ for the innateness of linguistic knowledge, according to which the primary linguistic data are too impoverished to allow children to acquire the rich structure of linguistic knowledge in a relatively short period of time. From this it is concluded that data exposure represents only a trigger that activates genetically given knowledge. Lightfoot writes that ‘on the basis of exposure only to primary data and endowed with certain properties somehow encoded in the genetic material (Universal Grammar), children acquire a mature grammar, part of their phenotype that varies depending on whether their initial experiences were in Tokyo or Toronto’ (45). But the validity of the poverty of stimulus argument is controversial. Pullum & Scholz (Reference Pullum and Scholz2002) point out that claims about the inaccessibility of data have entailments about the form of corpora: to say that children could not possibly access certain evidence is to say that certain data are missing in typical corpora, rendering it an empirical issue that should be settled on the basis of extensive corpus-linguistic research.

To further illustrate the poverty of the stimulus problem, Lightfoot discusses at some length English complementizer deletion, which is sensitive to locality conditions (e.g. Peter said [_CPthat/ø Kay left] vs. *Peter said yesterday in Chicago [_CP ø Kay left]). Complementizer deletion is claimed to be regulated by a UG condition, according to which ‘something is deleted if it is (in) the complement of an adjacent, overt word’ (48). Lightfoot thus in effect resurrects the notion of government, which has been abandoned in current Minimalism. More importantly, however, there is reason to believe that the complementizer that does not really get deleted, but alternates with a phonologically null C(omplementizer), which in Pesetsky's (Reference Pesetsky1991) analysis is an affix that must move up to the matrix verb. According to the C-as-an-affix analysis (which is extended in Bošković & Lasnik Reference Bošković and Lasnik2003 and Kim Reference Kim2008), that-insertion serves as a last resort operation when an intervening element blocks affix-hopping of the null C.

Chapter 4, ‘Languages emerging in children’, addresses the language acquisition process, which is accorded a central role in initiating diachronic change. To substantiate the claim that ‘work on learnability has been dominated too much by E-language approaches’ (66), Lightfoot presents a detailed critique of the acquisition model of the 1990s, according to which children match the input from the linguistic environment as much as possible. The logic of a combinatorial explosion provides the main argument against a global evaluation of grammars (73–77). Subsequently, Lightfoot summarizes his cue-based approach to language learning, wherein children scan their linguistic environments for pieces of partially analyzed structures. Crucially, children do not attend to the full complexity of adult sentence structures, but search for structural cues only in main clauses (82–86). This is called the degree-0 learnability (125). Most of the cues postulated by Lightfoot are highly theory-laden. As pointed out by Arbib (Reference Arbib2007: 4), this raises the problem of how the child will recognize a cue. Moreover, relatively complex structures can be produced at an early age: for example, children are capable of understanding and producing long-distance wh-questions by the age of 3.5 years (Guasti Reference Guasti2002: 209–220). Hence, one might wonder why cue-based learning should be restricted to simple, unembedded clauses.

Chapters 3 and 4 reiterate the pre-Minimalist view that linguistic knowledge is shaped by our genetic endowment with a minimum of external stimulation. More recently, Chomsky (Reference Chomsky2005) proposed that the design of the human language capacity results from the interplay of three factors: (i) the ‘genetic endowment, apparently nearly uniform for the species’ (Chomsky Reference Chomsky2005: 6), (ii) ‘experience, which leads to variation, within a fairly narrow range’ (Chomsky Reference Chomsky2005: 6), and (iii) ‘language-independent principles of data processing, structural architecture, and computational efficiency’ (Chomsky Reference Chomsky2005: 9). In drastically reducing the relative weight of the linguistic genotype (Universal Grammar), the Minimalist program has rendered the strong ‘geneticism’ of the Principles and Parameters paradigm far more tenuous (see Lorenzo & Longa Reference Lorenzo and Longa2009 for a concise discussion of UG dispensability and third-factor effects).

The reader interested in language change has to wait until the second half of Lightfoot's book before the first piece of historical data is discussed. Chapters 5, ‘New E-languages cuing new I-languages’, and 6, ‘The use and variation of grammars’, present case studies from early English syntax and some related material on parallel changes in Scandinavian languages. In view of all the progress made in the generative study of Old Romance languages, the book's limited database is somewhat surprising. As it stands, there is no way of knowing how typical and generalizable the early English case is for the rise of new languages in the course of time.

Chapters 5 and 6 pursue different goals. The focus of chapter 5 is on the development of new E-languages out of new I-languages, with apparently unrelated ‘superficial’ changes taking place simultaneously or closely following each other (111). Chapter 6, on the other hand, is concerned with new E-languages that evolve ‘for different kinds of reasons, for reasons of how speakers used their grammars and how coexisting grammars may mix differently in a social context’ (138). This recalls the principled division between catastrophic and environmental changes, advocated in earlier work (see Lightfoot Reference Lightfoot1991: 169–170, 1997: 174–176). Catastrophic changes are due to the resetting of a parameter, whereas environmental changes are gradually accumulating and spring from syntactic variation or language contact.

Lightfoot adduces the well-studied history of the English modals as a paradigmatic example of a catastrophic change (93). The loss of inflectional morphology had consequences for the category membership of modal verbs, which in turn had repercussions for syntactic computation. Having lost agreement markings, modals were reanalyzed as auxiliary verbs and merged directly into the inflectional node. As a result, V(erb)-to-I(nflection) movement was no longer available (90–102). However, there are both empirical and theoretical problems with the hypothesized causal chain morphological erosion→category change→loss of verb movement. Firstly, it is merely a stipulation that the loss of inflectional morphology entails a category change. Huddleston (Reference Huddleston, Huddleston and Pullum2002) defines the class of English auxiliary verbs exclusively in terms of their syntactic distribution. In contrast to main verbs, auxiliaries display the so-called NICE properties, i.e. they appear in constructions involving negation, inversion, code, and emphasis (see, in particular, Huddleston Reference Huddleston, Huddleston and Pullum2002: 92–101, section 2.1, and 102–103, section 2.2). It is therefore not entirely clear or obvious how ‘the change in category membership of the English modals explains the catastrophic nature of the change’ (93). Secondly, it is debatable whether the history of the English modals can be insightfully interpreted as an abrupt shift. Fischer, van Kemenade, Koopman & van der Wurff (Reference Fischer, Kemenade, Koopman and Wurff2000: 16) argue that to the extent that there is an abrupt shift, the history of the English modals is not an instance of grammar change, with one parameter of grammar being reset. Rather, the development of a class of modal verbs to free-standing functional categories (auxiliaries) shows all the earmarks of a standard case of grammaticalization and is treated as such by Roberts & Roussou (Reference Roberts2003: 36–48). Thirdly, in connection with the loss of V-to-I movement, Lightfoot discusses the so-called Rich Agreement Hypothesis (103–106). In its earliest presentation, this hypothesis was typically interpreted as a causal relation between the presence of rich agreement and the availability of verb fronting to I. A diachronic corollary of this hypothesis was that the loss of rich agreement entails the loss of verb movement (see Bobaljik Reference Bobaljik2002: 130–132 and the references cited therein). It is nowadays agreed upon by most syntactians (including Lightfoot) that only a weaker one-way implication of the Rich Agreement Hypothesis is empirically sustainable: the more explicit the language's inflectional system is, the more likely it is for the language to have independent V-to-I movement. This is because verbs with impoverished agreement or no agreement at all may still undergo V-to-I movement (Bobaljik Reference Bobaljik2001, Reference Bobaljik2002; Thráinsson Reference Thráinsson and Hendrick2003). Bobaljik (Reference Bobaljik2001: 7–8) capitalizes on the fact that since the correct generalization is a one-way implication, there must be causes for movement other than rich agreement. Thus, if verb movement is a syntactic parameter that is not uniquely a consequence of morphology, then the syntax is not morphology-driven. A conceptually more attractive proposal for the correlation between rich agreement and verb movement is proposed in Bobaljik & Thráinsson (Reference Bobaljik and Höskuldur1989). The key idea is that the rich inflection corresponds to structural complexity, viz. the division of the inflectional domain into an agreement and a tense projection. The availability of additional specifier positions correlates with other significant structural properties, most notably the availability of transitive expletive constructions and additional subject and object positions.

Chapter 6 tackles the change from object–verb to verb–object order in early English. In the context of this discussion, I found the excursus on Modern Dutch word order phenomena quite dispensable and would have preferred to see a more profound description of the historical facts. An aspect that is entirely missing from the discussion of the word order shift is the productive use of verb–object orders in Old English (see Fischer, van Kemenade, Koopman & van der Wurff 2000: 139–160). A promising line of research would therefore be to relate the variation between object–verb and verb–object order in Old English to different kinds of information structure (see van Kemenade Reference Kemenade, Hinterhölzl and Petrova2009).

The historical data discussed in these chapters serve primarily as showcases for a child-centered theory of language change. Both chapters have very little to say about diglossia and language contact, which reshape the syntax through borrowing and code-mixing. Contact-induced syntax change is not a priori irregular, since borrowing and code-switching apply in a systematic fashion. This is so because the mastering of two languages is part of the linguistic competence of the bilingual speaker (see e.g. Appel & Muysken Reference Appel and Muysken1987, Myers-Scotton Reference Myers-Scotton1993 and Lanza Reference Lanza2004).

Chapter 7, ‘The eruption of new grammars’, addresses the arguably clearest cases of language genesis: pidgin and creole languages. However, the connection between creoles and emergent sign languages is much more clearly stated in Lightfoot (Reference Lightfoot1997), which groups them together as counterevidence to the input-matching model of language acquisition. Lightfoot's (Reference Lightfoot1997: 173) bold statement that, ‘if creolization for the most part mirrors adult second language learning and is not abrupt and instantiated by children, then there is little reason for the theoreticians to be interested in the phenomenon’ is not reiterated in this book, which views creoles ‘in less exceptional terms, integrating their study with what we know independently of language acquisition and change’ (141). From this perspective, ‘the restrictedness of the experience of the first creole-speaking children is not as dramatic as when one is concerned with children's total experience’ (141).

However, there are many problems that arise in comparing child language, pidgin and creoles. Perhaps the most serious one is how to define the phenomena that are being compared. As pointed out by Romaine (Reference Romaine1988: 205), one difficulty lies in ‘identifying a set of formal structural characteristics which are uniquely associated with either a pidgin or creole’. When, after a period of social disruption, a new society comes into being, as in the case of the Caribbean plantations of the colonial era, children may grow up with only a pidgin as the linguistic model. In this context, the pidgin may then be expanded into a full language. However, this historical scenario leaves open the possibility that the creole derives from a rudimentary pidgin, from a stable pidgin, or even a structurally expanded pidgin (see Appel & Muysken Reference Appel and Muysken1987: 177–178, Romaine Reference Romaine1988: 115–117). Accordingly, the evolution of a creole language from a stable or structurally expanded pidgin may be much more gradual than envisaged in Lightfoot's account, as argued for Hawai'i Creole English by Roberts (Reference Roberts1998).

To sharpen the discussion on UG biases and the creole evidence for them, Lightfoot chooses Berbice Dutch (145), a by now extinct Guyanese creole, which combines left-headedness and canonical subject–verb–object (SVO) order with structural features that are atypical of Caribbean creoles, such as postpositions (in addition to prepositions), aspectual suffixation, and clause-final negation. The SVO surface order is of particular interest, since both its substrate languages (Eastern Ijo languages, in particular Kalabari) and its superstrate language (sixteenth- and seventeenth-century Dutch) are SOV (see Kouwenberg Reference Kouwenberg1992). Lightfoot proposes an analysis of Berbice Dutch SVO order in terms of cue-based learning and degree-0 learnability, wherein the Kalabari learners of Dutch would search for cues in non-embedded contexts, which, due to the verb-second property of Dutch, display surface SVO order. Berbice Dutch has a negative marker kane ‘not’, which, as in Eastern Ijo, is enclitic to the verb. Since the verb and the negation marker form a prosodic unit, one of the indicators for the original position of the finite verb is obscured (145–148). However, the cogency of this argument from negation is debatable, since the stranding of multiple verbs and particles to the right of the object would provide sufficient cues for the verb-second property (see Goddard & Zeijlstra Reference Goddard and Zeijlstra2009 for an account that evokes a multi-stage model for the formation of Berbice Dutch, which transcends more than one generation of language learners).

Chapter 7 would have benefitted greatly from a broader empirical coverage, for instance, by referring to the developing research on African urban youth languages (see Kießling & Mous Reference Kießling and Mous2004 for an up-to-date overview). African urban youth languages challenge a child-centered view on language evolution, since the agents involved in sustaining and expanding the language are adolescent speakers. An intriguing fact about African urban youth is that to the extent that they can master two systems, speakers manage to switch from one grammar to another without effort. Due to the prevalent multilingualism, competence in several grammars is the default case.

Chapter 8, ‘A new historical linguistics’, starts by recapitulating the major claims of the book with respect to language change, cue-based learning, and abrupt creolization. Section 8.2 contains a slightly reworked version of Lightfoot (Reference Lightfoot2002), ‘Myths and the prehistory of grammars’, which addresses the limits of syntactic reconstruction, claiming that the comparative method can be applied to reconstruct the syntax of a proto-language only when the daughter languages show identity. The thus reconstructed proto-syntax does not reflect reality, but is ‘a useful means to express relationships between systems’ (Lightfoot Reference Lightfoot2002: 134–135). If, on the other hand, the daughter languages do not show identity, there is no basis for reconstructing the proto-syntax of the ancestral language, as ‘there are no principles of history and, in a sense, history is an epiphenomenon’ (Lightfoot Reference Lightfoot2002: 134). A far less defeatist view on syntactic reconstruction is taken by Harris & Campbell (Reference Harris and Campbell1995: 346–353), who argue plausibly that syntactic patterns like alignment can be placed in exact correspondence, where the corresponding structures are carefully devised minimal sentence pairs. Where cognate forms do not exist or are unreliable (due to borrowing), structural features (including clausal patterns and systems of morphological organization) provide some insight into the phylogenetic affiliation of languages (see also Dunn et al. Reference Dunn, Levinson, Lindström, Reesink and Terrill2008).

Lightfoot envisages a new historical linguistics as a truly interdisciplinary enterprise, which harmonizes work on ‘syntactic theory, language acquisition, parsing, the use of grammars, social variation, computer modeling and creoles’ (184). In this long list, closely related disciplines – philology, language typology, literary theory, anthropology, and cultural history – are conspicuously absent. It seems to me that a rejuvenated historical linguistics would benefit from a multiplicity of approaches and ought to search for points of contact with a range of linguistic subfields. In this context, Lightfoot's statement that ‘historical linguists are not aging gentlemen at the end of the departmental corridor, working on etymologies and reconstructing proto-systems for all-inclusive phyla like Nostratic’ is unnecessarily offensive (184).

Considering the interface between historical linguistics and syntactic theory, an explanation of diachronic change in terms of cue-based learning can be criticized for focusing too much on the surface properties of syntax, without explaining complexity of syntactic change on the basis of what we know about language design. An alternative theory is that the source of syntactic change is syntactic variation, where syntactic variation is defined as the co-existence of various sentence patterns (word orders) in one language to express the same basic proposition with subtle distinctions in meaning. In enhancing the expressive power of the language, syntactic variation is an integral part of the syntax, and so is syntactic change. If syntax allows for variation and variation provides the catalyst for change, we can better understand the regularity and graduality that we encounter in syntactic change (see Reintges Reference Reintges, Crisma and Longobardi2009 for a more detailed discussion).

To sum up, Lightfoot's book reflects the state-of-the-art of the late 1990s, with most of the material previously published elsewhere. Nevertheless, How new languages emerge is a rewarding read for all those interested in the interplay between language acquisition, variation and change. The main tenet of the book that language acquisition is the locus of language change is still very influential in the field of generative historical syntax. It is, however, fair to say that the book offers neither new empirical evidence nor any conceptual arguments in support of Lightfoot's cue-based learning and degree-0 learnability hypothesis of historical change. For a book on language evolution and development, surprisingly little effort has been made to integrate into the theoretical analysis the results of current research on bilingualism, language contact, and creole linguistics. These comments notwithstanding, I have no doubt that Lightfoot's book will incite discussion on the nature of language evolution and change.