Language Evolution: The Windows Approach addresses the question: "How can we unravel the evolution of language, given that there is no direct evidence about it?"
The LINGUIST List is dedicated to providing information on language and language analysis, and to providing the discipline of linguistics with the infrastructure necessary to function in the digital world. LINGUIST is a free resource, run by linguistics students and faculty, and supported primarily by your donations. Please support LINGUIST List during the 2016 Fund Drive.
AUTHOR: Lightfoot, David TITLE: How New Languages Emerge PUBLISHER: Cambridge University Press YEAR: 2006
Michael A. Arbib, Computer Science, Neuroscience and USC Brain Project, University of Southern California
David Lightfoot develops his thesis that new languages emerge because children internalize new grammars from the flux of language use around them. To do this, he offers two complementary accounts: one of language acquisition, and the other of language change. Lightfoot has aimed the book at the general reader, but perhaps a quarter of the material will be hard going for readers not used to reading standard linguistic examples. Some footnotes announce that certain portions of the book are recycled from 2 earlier volumes, ''How to set parameters: arguments from language change'' (1991) and ''The development of Language Acquisition'' (1999). I have not gone back to make an explicit comparison with these volumes but these footnotes did alert me to ask ''How much of this book reflects post-1990 scholarship?'' The answer seems to be that Lightfoot has kept abreast of the literature on historical linguistics on the history of English and related studies of Scandinavian, but has followed almost no literature on other than the Romance and Germanic languages, save for Nicaraguan Sign Language. Given that Lightfoot asserts that the new historical linguistics is excitingly interdisciplinary, it is even more striking that his discussion of language acquisition is devoid of almost any analysis of actual studies of children acquiring language.
Lightfoot reviews the valuable distinction between E-language and I-language -- between the wild variety of utterances ''out there'' and the knowledge of language inside the head of the individual speaker. The child joins the language community by constructing an I-language that approximates aspects of the ambient E-language. The key point for the discussion of language change is that the E-language is sufficiently heterogeneous and the child's sample sufficiently small that different children may end up with different I-languages. As old speakers die and new speakers arise, the set of I-languages of the community changes. Their combined effect may preserve the E-language of the previous generation, change it in minor ways, or start changes that may cumulatively yield a new E-dialect or E-language. I agree with this general framework but it leaves open (at least) two major questions:
Language Acquisition: What is the nature of an I-language and how is it acquired by the individual? A sub-question is how much of the I-language is acquired by the young child, and how much may an individual's I-language change over the lifespan? Language Change: Language may change because of innovations invented and disseminated by young and older adults, or it may occur because young language learners extract new patterns from the ambient fluctuations of E-language to create a new population of I-languages. Is this an either/or situation, or are both processes operative?
On p.7, Lightfoot states that ''children are internally endowed with certain information, what linguists call Universal Grammar (UG), and, when exposed to primary linguistic data, they develop a grammar, a mature linguistic capacity, a person's internal language or I-language. The essential properties of the eventual system are prescribed internally and are present from birth''. There is no mention here or elsewhere in the book that many linguists disagree with this, and thus there is no occasion to review actual child development data to test alternative hypotheses. Consider, for example, the work of Jane Hill (reviewed by Arbib & Hill, 1984) on language acquisition. Jane modeled data she had gathered from a 2-year old child by a process in which the child interiorizes fragments of what it hears to build what we would now call constructions, based on the child's current concerns and interests. Over time some constructions merge with others, some fall into disuse, some gain complexity, and the lexicon grows and is categorized in tandem with the development of these constructions. Such ideas are now well developed in the recent volume by Tomasello (2003). The issue here is not which approach to language acquisition is correct but, rather, the need for evaluation of alternative theories, addressing the related data on children's development.
Implicitly, the phrase ''a grammar, a mature linguistic capacity, a person's internal language or I-language'' seems to omit the person's lexicon from the I-language, and Lightfoot explicitly excludes social conventions of language use from the I-language. He seems to consider the grammar as a set of abstract rules of syntax and phonology, though the latter receives little consideration here.
Anyway, the first major unquestioned hypothesis of the book is what I will refer to as the UG/LAD hypothesis – that there is an innate UG and that the child acquires its I-language (i.e., its syntax) via a Language Acquisition Device (LAD) which scans the ambient E-language to determine which settings of the parameters of UG match the sentences the child hears. Lightfoot offers the usual Poverty of the Stimulus argument but fails to note that recent literature militates against this argument (e.g., Pullum & Scholz, 2002). Indeed, when Lightfoot provides a critique of one approach to instantiating UG/LAD, his strategy -- while offered as a better approach to UG/LAD -- may actually be more in keeping with approaches that dismiss UG/LAD. Lightfoot argues convincingly that one approach to the UG/LAD hypothesis -- setting parameters by scanning through the entire E-language (Clark, 1992; Gibson & Wexler, 1994) -- must fail. The argument is based on the logic of a combinatorial explosion, not an analysis of developmental data. He then argues -- again, I think, convincingly -- that the child does not attend to the full complexity of adult sentences but instead attends to fragments. I also agree that the child will not process complex sentences (at least at first -- Lightfoot is not explicit about the ''maturation schedule'') but will focus on simple sentences in the search for structure. Where Lightfoot and I differ is that he sees the process of ''scanning fragments'' as a search for ''cues'', which are items from UG which can structure the sentence fragments. The catch is that these cues are highly theory-laden. For example, the child is asked to recognize an empty verb slot, or recognize a V-DP structure. However, no account is given of how the young child recognizes an empty slot, or knows a DP when he sees one. In what way, then, do I see Lightfoot as suggesting how the problem of the poverty of the stimulus might be addressed without recourse to UG/LAD? It is because once one accepts that the child is not looking at the entire E-language to switch in a complete grammar from those licensed by UG, but instead is slowly building up a repertoire of ''generation strategies'' for fragments of a simplified subset of the E-language, then the way is open to test the hypothesis that an innate UG is unnecessary after all. Instead one may posit that the child builds up a set constructions that gradually extend its ability to use language both to get others to minister to its needs and to engage in a broadening range of social interactions.
Lightfoot pays essentially no attention to data on language acquisition save to note the 4 stages that Clahsen & Smolka (1986) describe in the acquisition of German. No theory is offered of how the child activates ''cues'' from the observed E-language nor -- perhaps the most crucial point -- how the child avoids activating cues that seem to match other parts of the language input. The reader seriously concerned with understanding language acquisition will find little of value in Lightfoot beyond the idea that early acquisition is driven by attending to fragments of simple sentences rather than matching parameters against a wider sample that includes complex sentences.
The reader wishing to understand the state of the art in the UG/LAD approach to first language acquisition and the subtleties that arise when one then considers language acquisition should turn to Dalila Ayoun's (2003) ''Parameter Setting in Language Acquisition''. Ayoun makes clear that many problems bedevil this work. The most fundamental problem is that there is no agreement on which principles and parameters (or cues) form UG, or when two putatively innate features of language are governed by a single parameter. Another problem is to what extent the parameters or cues must be activated according to a maturation schedule, rather than as and when a sentence is encountered with the appropriate structure. And then the issue of second language acquisition raises all sorts of problems about whether a child's brain can hold 2 settings of the same parameter and bind them to use of the appropriate language. The reader may either find this a superb reference for further work on the UG/LAD approach or (as I did, presumably contrary to Ayoun's intentions) a strong indication that UG/LAD does not solve the language acquisition problem. In either case, it is a valuable book.
As Ayoun acknowledges and Lightfoot does not, the key to work on language acquisition is to analyze data on how children acquire language. Here a useful compendium is Eve Clark's (2003) ''First Language Acquisition''. Unlike Lightfoot's view of I-language as being restricted to syntax and phonology, Clark integrates social and cognitive approaches to how the children understands and produces sounds, words and sentences -- all within the context of learning to use language to cooperate and achieve goals.
Lightfoot's Chapter 2, ''Traditional Language Change'', which ''draws heavily on chapter 2 of Lightfoot (1999)'', introduces the study of historical linguistics by presenting two lines of study that culminated in the nineteenth century -- the study of the Great Vowel Shift and related aspects of sound change, and the charting of ''family tress'' for the Indo-European languages. He emphasizes that this research, for all its success, has shortcomings because the researchers lacked the critical distinction between E-language and I-language.
After introducing UG/LAD and his views on language acquisition, Lightfoot then develops his child-centered theory of historical language change in Chapters 5 and 6, ''New E-languages cuing new I-languages'' and ''The use and variation of grammars.'' I earlier raised the question of whether languages change because of innovations invented and disseminated by adults, or because of the way young language learners extract new patterns from of E-language or both. Lightfoot again has no time to mention or debate alternative hypotheses: He insists without question that children's formation of new I-languages is the key to how languages change. There is no discussion of alternative views and thus no marshaling of support for ''his'' view. However, Croft (2000) offers a very different view of language change and even devotes Section 3.2 to a critique of ''The child-based theory of language change'', including an explicit analysis of some of Lightfoot's earlier work. Lightfoot does not mention this. Lightfoot offers what I (as an outsider) feel to be an excellent summary of material from the history of English (and some related material on changes in Scandinavian language) but offers no related data on child language. His method is to simply observe that a change occurs in the texts from date x to date y and then state without evidence that change in children's I-languages must have been the driving factor. However, he usually states that changes in E-language (e.g., due to Viking invasions) must have triggered the changes. He offers no evidence (but see below) that the changes were not made first by adults accommodating to the novel language flux, and that children's language acquisition played a secondary role. To take a contemporary example: the rapid changes in cell phone technology are driven in great part by the enthusiasm of teenagers for new devices and services, but the new devices and services are usually created by thirty-year-olds (more or less) with the consumers providing selective pressure. In the same way, it seems to me strange to privilege the language learning of infants over the effects of adult innovation in response to a whole host of historical processes including language contact.
Let's consider one example (p.131-2): ''Bean (1983) examined nine sections of the Anglo-Saxon Chronicle [and] counted four constructions that needed to be analyzed with the verb remaining in its VP. She found 50% of such verbs in the section [sic] until 755, 25% in 865-864, ..., [and] 22% in 1132-1140. ... The option of moving the verb to a higher position was exercised increasingly over a period of several hundred years. This no more reflects a change in grammars than if some speaker were shown to use a greater number of passive and imperative sentences. ... Nonetheless ... changes in E-language, representing changes in the use of grammars, if they show a slight cumulative effect, might have the consequence of changing the robustness of the expression of cues, leading to new grammars. This seems to be what happened during Middle English.'' And that's the method -- gather data showing historical change, then argue that I-languages might change as a result, and then assert without discussion that the change in I-languages explains the E-language change. In my opinion, Lightfoot could make a better case for the importance of child language in language change if he abandoned the UG/LAD theory that an I-language is just a new combination of innate parameters or cues. A construction-based view would not only support the emergence of true novelty, but would also allow the child to acquire multiple constructions even though the generative grammarian might assign them to different UG parameter settings and then have as part of her I-language the ''probability settings'' for when to apply the constructions in different settings (cf. Eve Clark's usage-based approach). One could then see the changes analyzed by Bean (1983) as marking a continuous change in I-languages.
Perhaps the most promising Chapter, Chapter 7 on ''The eruption of new grammars'', discusses creoles and signed languages but, beyond discussing the latest paper (Senghas et al., 2004) on Nicaraguan Sign Language shows little sign of reworking beyond Lightfoot's earlier books (1991, 1999). The discussion of creoles usefully stresses the interaction of substrate and superstrate languages in the formation of a creole, as against a Bioprogram Hypothesis or expression of ''unmarked'' parameters in UG. However, Lightfoot again emphasizes the children's formation of I-languages without analyzing what adult interactions formed the pidgin from which the creole emerged, or citing historical studies of the circumstances under which the substrate or superstrate language may more heavily influence the final grammar (see. e.g., Holm 2000 for a review of relevant material). Nicaraguan Sign Language is, of course, fascinating because the children who created the language over the course of some 30 years had no extant sign languages to creolize. One may thus conclude that the children's formation of I-languages was crucial, with the I-languages of each cohort providing the E-language for the innovations of the next cohort. Lightfoot also usefully cites data showing ways in which a child exposed to ''broken'' samples of E-language may end up with an I-language that allows him to produce a better version of the E-language than his models could provide. However, there is nothing here that proves that all children have an innate UG/LAD for signed as well as spoken language, and that Nicaraguan Sign Language was created by children accumulating enough E-language to allow the younger children to set parameters (or cues) for which there was hitherto no evidence. Indeed, Senghas (2003) found that changes in the grammar of Nicaraguan Sign Language first appear among preadolescent signers, soon spreading to subsequent, younger learners, but not to adults. I would thus say that older children sought to find new ways to communicate; and that when they succeeded the new pattern of signing caught on with others and so formed the E-language, with formation of I-languages playing a secondary role of partial regularization. Again, the problem with Lightfoot's book is that he never spells out the alternatives, and thus never marshals the data needed to support his hypotheses.
And so we come to the concluding chapter, ''A new historical linguistics''. We saw that Chapter 2 introduced the nineteenth century study of the Great Vowel Shift and the charting of ''family trees'' for the Indo-European languages, and emphasized that the research has shortcomings because the researchers lacked the critical distinction between E-language and I-language. One might hope then that the book would finish on a triumphal note, celebrating the new insights that have accrued from using the UG/LAD hypothesis to ground a child-centered theory of historical language change. Instead we find that the book is silent on how to move forward the study of historical phonology. As for getting insight into the historical relatedness of languages, Lightfoot only offers a counsel of despair (p.173): ''It is impossible to know what a corresponding form could be in syntax, hard to know how one could define a sentence of French that corresponds to some sentence of English [sic] and hard to see how the comparative method could have anything to work with.'' Lightfoot's closing analysis (pp.181-183) of recent research seems to suggest that historical linguistics be abandoned in favor of comparative linguistics: ''Longobardi ... has developed an approach to defining relatedness between systems by quantifying the degree of correspondence among corresponding settings, without invoking reconstructions of any kind of change.'' Such an analysis has nothing to do with how children acquire language. Moreover, Longobardi's analysis (e.g., 2003, 2005) focuses on parameters he has analyzed for the syntax of noun phrases, so that languages which appear close by this method may differ dramatically in parameters for others aspects of language. I am prepared to believe that such analysis has a role to play, but stress that this utility holds even if we view UG as a purely descriptive framework seeking to display a set of variations in the grammars of human languages and reject the view that UG is an innate capacity that guides all the essential processes of language acquisition and thus delimits the set of possible human languages.
In conclusion, Lightfoot's book offers theories of both language acquisition and language change by assertion. With minor exceptions, he does not acknowledge the existence of rival theories and thus offers no discussion of data pro or con his own theories. As a result, the case that his approach will bolster work on historical linguistics seems ill-supported.
Arbib, M.A., and Hill, J.C., 1984, Schemas, Computation, and Language Acquisition, Human Development 27:282-296.
Ayoun, D. (2003) Parameter Setting in Language Acquisition, London: Continuum.
Bean, M., 1983, The development of word order patterns in Old English, London: Croom Helm.
Clahsen, H., & Smolka, K.-D., 1986, Psycholinguistic evidence and the description of V2 in German, in Verb-second phenomena in Germanic languages (H. Haider & M. Prinzhorn, Eds.), Dordrecht: Foris, pp.137-167.
Clark, E. (2003) First Language Acquisition, Cambridge: Cambridge University Press.
Clark, R., 1992, The selection of syntactic knowledge, Language Acquisition, 2:83-149.
Croft, W., 2000, Explaining Language Change: An Evolutionary Approach, Harlow, England: Longman.
Gibson, E., & Wexler, K.,1994, Triggers, Linguistic Inquiry, 25:407-454.
Holm, J., 2000, An introduction to pidgins and creoles, Cambridge: Cambridge University Press.
Lightfoot, D.W., 1991, How to set parameters: arguments from language change, Cambridge, MA: MIT Press.
Lightfoot, D.W., 1999, The development of Language Acquisition, Oxford: Blackwell.
Longobardi, Giuseppe. 2003. Methods in Parametric Linguistics and Cognitive History. Linguistic Variation Yearbook 3, 2003, 101-138.
Longobardi, Giuseppe. 2005. A Minimalist Program for Parametric Linguistics? In Organizing Grammar: Linguistic Studies for Henk van Riemsdijk, ed. by Hans Broekhuis, Norbert Corver, Marinus Huybregts, Ursula Kleinhenz, and Jan Koster. Berlin/New York: Mouton de Gruyter, 407-414.
Pullum, G.K., & Scholz, B.C., 2002, Empirical assessment of stimulus poverty arguments, The Linguistic Review, 19:9-50.
Senghas, A, 2003, Intergenerational influence and ontogenetic development in the emergence of spatial grammar in Nicaraguan Sign Language Cognitive Development 18:511–531.
Senghas A, Kita S, Ozyurek A., 2004, Children creating core properties of language: evidence from an emerging sign language in Nicaragua. Science. 305:1779-82.
Tomasello, M., 2003, Constructing a Language: A Usage-Based Theory of Language Acquisition, Cambridge, MA: Harvard University Press.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Michael Arbib is a University Professor at the University of Southern California. Among his edited and co-edited books are The Algebraic Theory of Machines, Languages, and Semigroups, (Academic Press, 1968), Neural Models of Language Processes (Academic Press, 1982), The Handbook of Brain Theory and Neural Networks, (The MIT Press, 1995, 2003), Who Needs Emotions: The Brain Meets the Robot, (Oxford University Press, 2005), and From Action to Language via the Mirror System, (Cambridge University Press, 2006).