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.
Review of Variation and Universals in Biolinguistics
Date: Wed, 13 Jul 2005 18:09:34 +0200 From: Judit Gervain Subject: Variation and Universals in Biolinguistics
EDITOR: Jenkins, Lyle TITLE: Variation and Universals in Biolinguistics SERIES: North-Holland Linguistic Series, Linguistic Variations 62 PUBLISHER: Elsevier Ltd. YEAR: 2004
Judit Gervain, Cognitive Neuroscience Sector, Scuola Internazionale di Studi Avanzati, Trieste, Italy
1. Introduction Had Eric Lenneberg written his seminal book "Biological Foundations of Language" (1967) today, he most probably would have written something quite similar to the present volume. As the very frequent references suggest, the authors of this collective volume themselves were also very well aware of the Lennebergian heritage. Although many of Lenneberg's original proposals have been refuted or reformulated since, the general spirit of his ideas still continue to shape the way we think about the biological aspects of language. His ideas gave rise to what has become known today as the biolinguistic enterprise.
The volume offers a very fortunate selection of papers from the biolinguistic field, succeeding in blending general and easily accessible introductions to more biological or formal research areas that linguists and psychologists are less familiar with, such as the correlation between genetic and linguistic variation, or the state of the art in the study of human evolution, with in-depth analyses and detailed empirical reviews of controversial issues, such as the exact nature of the deficit underlying Specific Language Impairment (SLI), the mechanisms involved in language acquisition or the evolution of language. Many of the papers offer a rethinking of some of the received views in the domain. A common underlying theme running through all the articles is the attempt to achieve a complex understanding of the key issues, with an eye towards an ultimate unification with the biological sciences. This genuinely multidisciplinary approach gives a unique appeal to the volume.
2. What form should linguistic theory take? Most of the papers explicitly or implicitly assume, as their theoretical background, the Principles and Parameters (P&P) model as most recently reformulated in the Minimalist Program (MP; Chomsky 1995). This model represents a radical attempt at creating a "minimal" theory of grammar, i.e. the most parsimonious possible, based on the underlying assumption that syntactic computations, i.e. the core of language, is a nearly perfect design to mediate between meaning and sound. Most of its properties are imposed by these two external systems, and syntax itself introduces as little of its own material as possible. This intuition, as it will become clear, has paved the way for a broad range of novel investigations, especially into the evolution of language. Therefore, it is crucial to understand the motivations behind and the evidence in favor of it.
This is precisely what Noam Chomsky's contribution "Language and Mind: Current Thoughts on Ancient Problems" does. First, in a historical perspective, he shows how the generative enterprise has proceeded from the empirical descriptions of particular constructions in particular languages to a more general, abstract and explanatory theory, the MP, ridding itself of unnecessary formal machinery. Although the idea of a nearly perfect syntax is, as Chomsky himself concedes, surprising, he also claims that the theory has not lost in empirical adequacy in the attempt, which constitutes a reason to prefer it over the previous, more redundant, less parsimonious models. This, however, is not uncontroversial. Pinker and Jackendoff (2005), for instance, has recently argued to the contrary, citing a certain range of phenomena that remained unaccounted for in the new framework. Ultimately, it is, of course, an empirical question whether the MP will succeed in equaling the former theories in empirical scope.
One relevant issue in this regard is what precise form the correct model should take. Building on his previous work, Richard Kayne sets out, in his present article "Antisymmetry and Japanese", to provide further empirical support for his "antisymmetry thesis". His main proposal (Kayne 1994) has been that contrary to apparent cross-linguistic variation, word order is underlyingly identical in all languages, conforming to the basic SVO (in more technical terms Specifier-Head-Complement) order. The reason why some languages exhibit surface orders different from this, e.g. SOV, is because several displacement operations apply to the underlying order in their syntactic derivations. If true, this surprising and empirically not uncontroversial claim provides substantial support to the idea that basic syntax is universal and conforms to the needs of the interfaces, in this case to the needs of the conceptual-intentional interface by reflecting the underlying order of elementary predication. Therefore, in his current contribution, Kayne investigates word order phenomena in Japanese, the paradigmatic example of a language that shows SOV order on the surface, and convincingly argues, by deriving Noun-Postposition and Subordinate clause-Complementizer orders through movement, that word order in these constructions is epiphenomenal.
3. Language acquisition from a biological perspective The issue of distinguishing linguistic universals from cross- linguistically varying properties bears direct relevance to the problem of language acquisition. One of the most fundamental commitments, indeed one of the original motivations, of the generative enterprise has been the claim that language cannot be learnt exclusively from the input (the logical problem of induction, also known as the argument from the poverty of the stimulus), therefore humans must possess a language learning device with a considerable amount of linguistic knowledge already hard-wired in it. This approach has been theoretically formulated in the P&P model, according to which the principles are universal and need no further specification during acquisition, while parameters define (binary) choices, the adequate values of which the learner has to set to model the target grammar. It has been proposed (e.g. Pinker 1984, Gibson and Wexler 1994) that the exact mechanism by which parameters are set can be thought of as some kind direct triggering from the input, in which certain cues direct the learner's choice. As a corollary, it has been argued that triggering is fast and effortless, therefore parameters are set very early on (Wexler 1998). Under this view, young children's syntactic errors are attributable to maturational or "performance" factors (e.g. memory limitations in the case of long or complicated sentences etc.).
In the current volume, three articles address the issue of language acquisition, two of them, Wexler and Avrutin, proposing accounts that are compatible with this classical view, while the third one by Yang challenges some of the traditional assumptions from a computational/statistical point of view. In the chapter entitled "Lenneberg's Dream", Ken Wexler offers a comprehensive overview of the theoretical arguments and the empirical evidence for his Optional Infinitive (OI) stage theory. The original proposal derives from the observation that children's early production contains both finite, inflected adult-like verbal morphology and infinitival, non-inflected forms that alternate in the same sentential environments. Initially, the infinitival forms are dominant, but become less and less frequent with time. When inflected morphology appears, it is always grammatical; it is never used when an infinitive or a participle would be required, and the different inflected forms never replace each other randomly or ungrammatically. All these facts taken together imply, argues Wexler, that in the developmental phase that he calls the Optional Infinitive stage, children do possess the relevant grammatical knowledge, otherwise they would use the inflected forms randomly. However, they have difficulties exhibiting this knowledge in constructions that require the application of certain grammatical operations (movement/feature checking) twice. In English, this is precisely what is required to get inflection right. As a good control case for the hypothesis, Wexler presents data from languages such as Italian and Spanish, in which the same operation is not required to apply twice, only once. As the hypothesis predicts, in these languages, children never fail to mark the main verb of a sentence for the correct person and tense inflection. The constraint that children cannot perform certain operations twice, while adults can, can thus be seen as a maturational, i.e. performance limitation on an otherwise adult-like grammar. In this respect, then, Wexler's OI theory follows the traditional path of attributing children's errors to extra-linguistic factors. However, the scope of his hypothesis is enlarged by encompassing data from behavioral genetics and language pathologies. In accordance with the idea that the limitations underlying the OI stage are genetically determined maturational constraints, Wexler and colleagues found that the difference in the time of onset and the duration of the OI stage is only 3 weeks in monozygotic twins, as compared to 13 weeks in dizygotic ones. Moreover, investigating the errors English- and Dutch-speaking children with SLI make, Wexler observed that their patterns are very similar to those of normally developing children in the OI stage, only with poorer overall performance. Therefore, Wexler argues that in SLI children, the OI stage is maintained for a much longer period of time, sometimes not even fully obviated in adulthood. Under this view, then, the SLI grammar is claimed to be intact, and what is affected is the performance system.
A similar line of reasoning comparing the linguistic abilities of normally developing children and language pathology underlies Sergey Avrutin's contribution, entitled "Beyond Narrow Syntax". Starting out from the observation that young children interpret the use and distribution of pronominal forms differently from adults, Avrutin argues that the behavior of these referential elements (pronominals, anaphoric expressions, nouns preceded by determiners, introducing individuals into the discourse, as well as tense marking, introducing events into the discourse etc.) is governed not by syntax, but by what he calls `discourse', and defines as the interface between pure syntactic computation, i.e. narrow syntax and the conceptual-intentional (C-I) system. One empirical motivation to delegate referential elements to the C-I interface as opposed to syntax proper is the existence, even in normal adult language, of special registers such as `diary language' where the behavior of referential elements and tense marking is different from normal usage, e.g. omissions are possible in contexts were they are otherwise ungrammatical etc. Therefore, even the fully mature language faculty allows for the introduction of novel discourse entities through non-syntactic means. Interestingly, as Avrutin notes, determiners, pronominals and tense marking are precisely those grammatical elements that young children and aphasic patients very frequently omit in their linguistic production. Importantly, however, when they do employ them, their usage is grammatical. Consequently, Avrutin conjectures that the narrow syntax of language learners and aphasics is intact, but when the production of a certain structure becomes too resource-intensive, their computational system breaks down, and the discursive means of expression take over, giving rise, as in adult special registers, to characteristic omission patterns. Avrutin's model is another example of a theory that takes errors to be the symptoms of performance limitations on an otherwise normal grammar.
This view is challenged from a formal perspective in the chapter entitled "Toward a theory of language growth" by Charles Yang. He argues that if the triggering/parameter-setting view of acquisition is correct, children's production should change abruptly and categorically as a parameter is set in one way or in the other. Nevertheless, says Yang, this is not what we find in their productions. Rather, at least in the case of certain syntactic constructions, though not all, error rates are initially high, and decrease only gradually, without any sign of a dramatic drop that would be predicted by the (correct) setting of the relevant parameters. The author thus argues that the Very Early Parameter-Setting (VEPS) view proposed by Wexler (1998, current volume), according to which parameters are set correctly from the earliest observable age (around 18 months), cannot hold true for all parameters (e.g. the second position of the verb in Germanic languages, the obligatory presence of the subject in English etc.), although he acknowledges that for some, it does (e.g. word order in English, verb raising in French etc.). Instead, he proposes to account for the gradual disappearance of errors as a selectional process at the beginning of which infants start out with the set of all humanly possible grammars, probabilistically choosing one of them for parsing each time a sentence comes in from the input. If the chosen grammar is compatible with the input sentence, its probability is increased, otherwise it is decreased. Over time, grammars with low probabilities are eliminated, and learners converge on the target grammar. In this model, children simultaneously entertain multiple grammars, and eliminate them gradually, which explains the observed error patterns.
As empirical support for the model, Yang discusses how the obligatory presence of the subject in English (the non pro-drop property) is acquired. The world's languages exhibit three major patterns with respect to pro-drop. In languages like English, pro-drop is not allowed. In many other languages, such as Italian or Spanish, pronouns can be dropped if they are in subject position and agree with the inflected verb. In languages, like Chinese, subjects can be dropped if they are topicalized. While it is easy for an English-learning child to exclude the Italian-like pro-drop option, since it is readily visible in already a few input sentences that inflectional morphology is poor in English, the Chinese- like grammar takes much more input to rule out. Therefore, the empirical prediction is that English-learning children's initial pro-drop errors, before they disappear gradually, will show a Chinese-like pattern. This is confirmed by empirical data. Let us note, however, that more evidence is needed before Yang's suggestions can be accepted, since a probabilistic selectional learning model has no way to explain why all children present with highly similar developmental paths (e.g. all make similar kinds of errors and produce the same kinds of constructions in the same chronological sequence), not only within one language, but also cross- linguistically. If, at least at the beginning, children were to choose randomly from the thousands of possible grammars, wouldn't considerable divergence be expected? Also, would the development not heavily depend on the nature and the amount of input in such a framework? However, empirical evidence (Gleitman and Newport 1995, Wexler 1998, current volume) points to the contrary.
4. Language growth: Language change and language evolution Models of language acquisition determine how language is transmitted from one generation to the next. Therefore, they are implicated in the account one can offer for how language changes over time, or even how language came about during human evolution. Recently, with the advent of the minimalist architecture of the language faculty, which has made empirical inquiry more simple, thus more feasible, the issues of language change and evolution have received increasing attention (e.g. Hauser, Chomsky and Fitch 2002).
Since evolutionary and even historical sources are scarce, Partha Niyogi in his "Phase Transition in Language Evolution" puts forth a computational approach formally modeling language change, with potential implications for language evolution. (As it might have already become apparent, Nigoyi's title is somewhat misleading, as he is not very strict about distinguishing language change from language evolution.) He proposes to treat language in time as a dynamical system. In each generation, grammatical variation is determined by a certain distribution of all the possible grammars, e.g. in a homogeneous population all speakers have to same grammar, in a heterogeneous population, some have grammar G1, others grammar G2, yet others maybe simultaneously both of them, etc. Language arises in the next generation through language acquisition, which is driven (within the logical space defined by UG) by the input received from the previous generation. This, in turn, is derived from the grammar distribution characterizing that generation. In such a model, grammatical change depends on two factors, the probability with which speakers of the older generation produce examples for a given construction, and the overall amount of input to the new generation.
Crucially, argues Niyogi, gradual changes in these parameters do not produce linear changes in the dynamics of the system, rather as certain threshold values are crossed, phase transitions occur. Transitions that eliminate variation (e.g. a population going from having two grammars to having only one) always produce stable states, while transitions in the other direction are not probable, and happen only at vary high values of the two parameters, which may happen following external influence to the system, e.g. a massive influx of speakers with a different grammar. Thus, overall, variation seems to be eliminated, when present, and practically never emerges, if originally absent. Although this conclusion constitutes an important contribution to the on-going debate about the origin of language change (internal or external), it needs further confirmation for at least two reasons. Firstly, it goes against the initial variationist assumptions of the model. Secondly, and more importantly, it is not self- evident how to interpret it in the light of known empirical observations of historical change, since variation does not always seem to get eliminated over time.
Another way of overcoming the lack of historical and evolutionary evidence is to look at cases where language emerges from nothing in the present times. Although such evens are rare, the creolization of pidgins and the emergence of sign languages are good cases in point. Judy Kegl's article, "Language Emergence in the Language-Ready Brain" gives a detailed description of the lexicon and grammar of American Sign Language and Nicaraguan Sign Language (Idioma de Senas de Nicaragua, ISN), and offers a comprehensive report about the emergence and current state of ISN. This is particularly interesting from a biolinguistic point of view, because ISN is a genuine case of language emergence de novo. The language was created by deaf children who came together in a special school in the early 1980s. Kegl's paper describes the Nicaraguan signer population in detail, showing that signers who learn ISN early, as a first language (L1), gain access to all the production rules encoded in lexical items, while late learners memorize them as frozen, monolithic elements. (In most sign languages, lexicalized signs are complex and reflect the composition rules of the language, a little bit like derivational morphology in spoken language reflects certain productive patterns.) In addition, Kegl analyzes ISN to show that elements come from one of three different sources. Some derive directly from simple home-signs and gestures used in the surrounding spoken language. Others are taken over from these sources, but are subsequently modified and put to a different use in ISN. Yet others do not have external origins, and constitute real innovations of ISN. Interestingly, as Kegl points out, these are precisely the elements used for grammatical purposes. Thus, she concludes that ISN constitutes strong evidence in favor of a genetic endowment for language, which can be triggered by minimal external input (home-signs and gestures from the spoken languages) to create a full language, because the missing grammatical system can be filled in by innate content.
Massimo Piatelli-Palmarini and Juan Uriagereka, in their paper entitled "Immune Syntax", explore the emergence of language from a more evolutionary, but at the same time more speculative perspective. Starting out from the minimalist idea of the perfection of syntax, they use the analogy of the immune system to claim that the imperfections of syntax, which are introduced by lexical and morphological variation, can be seen as some kind of a `virus', and syntactic computations work to eliminate them as fast as possible. On a biological level, they argue, the virus metaphor can be interpreted as syntax evolving through `horizontal' mechanisms, a typical example of which is when viral genetic material gets integrated into the DNA of the host, and then gets transmitted by regular vertical procedures to future generations. Examples of this are rare, but attested in our natural history, and the advantage of such a `horizontal' scenario, the authors argue, is that it operates on exactly the right time scale to explain the emergence of language, whereas most known vertical mechanisms cannot account for the recency, as well as the complexity of language. Nevertheless, as the authors themselves acknowledge, their hypothesis remains, in its current formulation, an evolutionary just-so- story.
Isabelle Dupanloup, in her contribution entitled "Genetic Differences and Language Affinities", explores a much less speculative approach to language evolution. She starts out by reviewing what is known about the evolution of the human line from the common primate ancestors up to the peopling of Europe during the Neolithic spread about 10 000 years ago. Her state-of-the art report is surely very useful for the non-biologically trained readers to better understand what issues are at stake, and what evidence we currently have that bears on these issues. This background set, she asks the question whether current genetic diversity correlates with linguistic variation. She finds that when genetic diversity is measured in terms of blood groups and protein variability, close correlations obtain (at least, in the populations of Africa, Asia, and Europe), but similarities between linguistic and genetic variation are much weaker, when genetic variation is measured at the DNA level. This suggests, she argues, that there is a common pattern of genetic and linguistic divergence during human history, indicated by proteins, blood groups and other allelic variants affected by genetic fluctuations as rapid as linguistic change. DNA sequences, on the other hand, are subject to slower evolutionary change, and still reflect an evolutionary stage in which languages had not yet diverged.
6. Specific Language Impairment under the microscope One of the focal topics in relating language to the genetic endowment is the study of genetically-based language pathologies, among which SLI is of particular interest. The present volume contains several contributions on the issue.
Heather van der Lely ("Evidence for and Implications of a Domain-Specific Grammatical Deficit") sets the stage by arguing that G(rammatical-)SLI is a domain-specific deficit, since the affected children present with problems in the grammatical domain only, but importantly, not in audition, general cognition or non-grammatical language skills (pragmatics etc.). According to van der Lely's Representational Deficit for Dependent Relations (RDDR) account, G-SLI children have a pervasive deficit representing hierarchical structure, which surfaces in phonology, morphology and syntax alike, all three linguistic levels relying on hierarchical structural representations. More formally, in a minimalist framework, G-SLI children can be characterized as treating the movement operation as optional in cases where it is obligatory in the adult grammar. She presents experimental evidence from wh-movement (question formation) in English, Greek and Hebrew confirming the predictions. In her view, then, SLI affects the core of language.
The idea of the optionality of certain grammatical operations in SLI also underlie the other accounts of this pathology in the volume. However, these papers focus more closely on inflectional morphology. Nevertheless, they offer very different explanations.
Ken Wexler, as already mentioned, claims that SLI children are stuck in the Optional Infinitive stage during development. Since this phase is brought about by performance limitations on an otherwise adult-like grammatical system, Wexler, unlike van der Lely, argues that SLI does not affect the core of grammar.
Following up on Wexler's OI hypothesis, Laurence Leonard ("Exploring the Phenotype of Specific Language Impairment") observes that SLI children do indeed omit inflectional morphology. However, as he points out, they do it more often than linguistically matched normally developing children. Therefore, he argues that SLI children are more sensitive to the resource demands of sentence formation and the processing loads imposed by complex constructions. Supporting this view, he presents experimental results showing that children with SLI produce more uninflected forms in ditransitive than in simple transitive constructions.
Myrna Gopnik ("The Investigation of Genetic Dysphasia") gives a different analysis of optional morphology in SLI, or, in her terminology, genetic dysphasia. In her view, SLI children lack sublexical, that is morphological features in their linguistic representations. Therefore, they cannot decompose polymorphemic forms, they can only memorize them as unsegmented chunks. So whenever they produce a morphologically complex inflected form, they have retrieved it from memory. This view, she argues, is further confirmed by the observation that in morphologically rich languages such as Greek or Japanese, SLI children have problems not only with inflections, but also with allomorphoic variations of the roots.
7. Language and the brain Bridging the gap between genetics and behavior, some of the contributions in the volume investigate the brain representations and functions underlying language.
Alfonso Caramazza and Kevin Shapiro report the case studies of aphasic patients who present with a classical dissociation between verbs and nouns, but, interestingly, in different modalities. One patient shows poor performance on verbs in the the oral modality, while another patient shows a similar deficit for verbs, but only in writing. The authors take the dissociation of modalities to be strong evidence in favor of a specifically grammatical deficit, since the underlying semantics (and all other non-grammatical factors such as imagibility etc.) are the same for both modalities, e.g. the semantic complexity of the verb to `give' is the same in speech and writing. Nevertheless, they don't make it clear why the same argument doesn't apply to the grammatical properties. In addition to naturally occurring pathologies, an experimental technique called Transcranial Magnetic Stimulation (TMS) is available to induce temporary `deficits' by knocking out the fuctions of the stimulated brain area. Using this technique, similar dissociations could be obtained in linguistic performance. On the basis of the patient and the TMS data, Caramazza and Shapiro conclude that verbs and nouns are represented in the brain by distinct circuits, the former probably in the superior, anterior, the latter in the inferior, posterior part of the left frontal cortex.
Another very common issue in neurolinguistics is the localization of syntax in the brain. As Yosef Grodzinsky, in his contribution "Variation in Broca's Region", points out, the old view according to which syntax is in Broca's region is untenable, because, on the one hand, Broca's region is also implicated in non-linguistic functions, and, on the other hand, syntax has been shown to recruit other brain areas, too. However, as he argues, presenting evidence from patient and brain imaging studies, it seems to be the case Broca's are is indeed responsable for the syntactic operation of the movement of phrasal constituents. Therefore, Grodzinsky concludes, the idea that language is modular at the neural level is supported.
8. Unifying biology and language: still a long way to go? Beyond their empirical and theoretical merits, the papers are also of interest from a foundational or epistemological point of view. Implicitly or explicitly, they are all attempts to bring formal linguistic theory closer to the study of the more biological aspects of language. The idea of such a unification was first raised by Chomsky (e.g. 1968) and has been gaining popularity ever since. In some domains, such as psycholinguistics and neurolinguistics, progress has been considerable. Other areas, like the investigation of the evolution of language, have only begun to emerge (possibly due to the recent changes in linguistic theory brought about by the MP). Today, the possibility of a unifications might seem so imminent that the editor of the volume, Lyle Jenkins fully dedicates his contribution ("Unification in Biolinguistics") to this issue. Unfortunately, he doesn't go beyond briefly introducing some cases of unification from the history of the natural sciences, and some rough analogies between the evolution of species and the emergence of language diversity.
Although biolinguistics has gone a long way towards unification, so that many questions have turned from unsolvable puzzles into formulable scientific problems, to use Chomsky's terminology, a lot remains to be done on both sides of the gap before we completely understand the path from the genome to linguistic behavior in its full complexity.
Chomsky, N. 1968. Language and Mind. New York: Harcourt, Brace and World.
Chomsky, N. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press.
Gibson, E. and Wexler, K. 1994. "Triggers". Linguistic Inquiry 25(3):407- 454.
Gleitman, L and Newport, E. 1995. "The invention of language by children". In: Gleitman, L. and Liberman, M (eds.): Invitation to Cognitive Science, Vol. 1.: Language. Cambridge, Mass.: MIT Press.
Hauser, M., Chomsky, N. and Fitch, T. 2002. "The Faculty of Language: What Is It, Who Has It, and How Did It Evolve?" Science 298:1569-1579.
Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press.
Lenneberg, E. 1967. The Biological Foundations of Language. New York: Wiley.
Pinker, S. 1984. Language Learnability and Language Development. Cambridge, Mass.: Harvard UP.
Pinker, S. and Jackendoff, R. 2005. "The Faculty of Language: What is Special about it?". Cognition 95(2):201-236.
Wexler, K. 1998. "Very early parameter setting and the unique checking constraint". Lingua 106:23-79.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Judit Gervain is currently a 3rd year PhD student at the Language, Cognition and Development Lab, Cognitive Neuroscience Sector, SISSA, Trieste, Italy under the supervision of Prof. Jacques Mehler. Her first degree is in English Philology, French Philology and Theoretical Linguistics from the University of Szeged, Hungary. She wrote her MA theses in English Philology and in Theoretical Linguistics about focus- raising phenomena in Hungarian. She has published several papers in this topic. Currently, she is working on language acquisition. Her precise research topic is the acquisition of the foundations of syntax in the first year of life. At the same time, she continues to do research in theoretical linguistics, in the area of left peripheral phenomena (focus and wh-raising etc.).