Structure dependence, English Grammar and Universal Grammar

FACULTADE DE FILOLOXÍA DEPARTAMENTO DE FILOLOXÍA INGLESA E ALEMÁ Structure dependence, English Grammar and Universal Grammar Beatriz Gómez Vidal Supervisado por Dr. Víctor M. Longa Martínez Grao en Lingua e Literatura Inglesas Liña temática: Lingüística e procesamento das linguas. Aplicacións 2016/2017 FACULTADE DE FILOLOXÍA DEPARTAMENTO DE FILOLOXÍA INGLESA E ALEMÁ Structure dependence, English Grammar and Universal Grammar Beatriz Gómez Vidal Supervisado por Dr. Víctor M. Longa Martínez Grao en Lingua e Literatura Inglesas Liña temática: Lingüística e procesamento das linguas. Aplicacións 2016/2017 Table of Contents INTRODUCTION ........................................................................................................................................... 1 1. SCIENTIFIC AND PHILOSOPHICAL BACKGROUND ...................................................................................... 3 1.1 RATIONALISM VS. EMPIRICISM ......................................................................................................................... 3 1.2 COGNITIVE AND BIOLOGICAL UPDATE: MENTAL ARCHITECTURE MODELS, INNATENESS AND LEARNING ............................. 5 2. LINGUISTIC INNATENESS ........................................................................................................................ 11 2.1 THE PRINCIPLES OF LINGUISTIC INNATENESS THEORY ........................................................................................... 11 2.2 MODULARITY, THE LANGUAGE FACULTY AND THE APS ....................................................................................... 12 2.3 UNIVERSAL GRAMMAR ................................................................................................................................ 16 3. THE APS AND STRUCTURE DEPENDENCE ................................................................................................ 21 3.1 THE THREEFOLD NATURE OF THE APS.............................................................................................................. 21 3.2 STRUCTURE DEPENDENCE AND ENGLISH GRAMMAR ........................................................................................... 25 3.2.1 Structure dependence: definition and relation with the APS and UG ............................................ 26 3.2.2 Yes-no question formation: experiments and results ..................................................................... 29 4. DEBATE ON STRUCTURE DEPENDENCE AND THE APS ............................................................................. 33 4.1 CRITIQUES ON INNATENESS HYPOTHESIS: A FIRM REJECTION OF THE THIRD DATA DEFICIENCY ...................................... 33 4.2 POSSIBLE CONFIRMATION OF THE THIRD DATA DEFICIENCY ................................................................................... 38 5. INNATENESS REVISED: A CRITIQUE ON LINGUISTIC INNATENESS HYPOTHESIS BASED ON DEVELOPMENTAL BIOLOGY ....................................................................................................................... 41 5.1 THE RATIONALE BEHIND THE APS ................................................................................................................... 41 5.2 THE (INDIRECT) RELATIONSHIP BETWEEN GENES AND TRAITS ................................................................................ 42 CONCLUSION ............................................................................................................................................. 47 WORKS CITED ............................................................................................................................................ 49 1 Introduction It is a widely accepted fact that many things about linguistic behavior are yet to be satisfactorily explained by linguists. Among these, the discussion on the possible sources of linguistic knowledge is to be highlighted. This debate, which is still on-going in contemporary linguistics, has long been of interest to me. Whether certain aspects of linguistic knowledge are innate or whether they are learned from the environment is a complex question which still goes unanswered, despite the many academic efforts that have been made to support one view or the other. Linguistic innateness hypothesis argues for the existence of innate linguistic knowledge in the human species (Homo sapiens). This linguistic theory is closely linked to Generative Grammar – one of the most influential formal approaches in linguistics. Being highly intrigued by the enormous potential that formal approaches have for grammatical analysis, my intention was to explore the relation between English Grammar and the theory of linguistic innateness. I was able to achieve this by focusing my argument on one crucial aspect of syntax known as structure dependence, which has been considered by linguistic innateness theory to be an innate linguistic principle. Structure dependence is a key property of all grammars, including English, as it accounts for the hierarchical nature of language. The aim of this work is twofold: (i) to present a thorough revision of structure dependence in relation to English Grammar from the point of view of linguistic innateness hypothesis, and (ii) to offer relevant criticism of such a theory, with the intention of providing a constructive and critical evaluation of the methodology and claims of linguistic innateness hypothesis. For this purpose, I have consulted numerous works from a wide array of authors, each of them notorious in their corresponding field of inquiry. Out of the many scholars whose works I have referred to, Noam Chomsky is to be highlighted, as he is the main theoretical defender of linguistic innateness theory. This revision of linguistic innateness and English Grammar is organized in five sections. In section 1, I set the grounds for the introduction of linguistic innateness theory by focusing on the presentation of the ‘rationalism vs. empiricism’ dichotomy, since rationalism is the undeniable predecessor of linguistic innateness hypothesis. I also provide a scientific 2 update of the aforementioned philosophical trends by resorting to current biology and the cognitive sciences. In section 2, I provide an ample description of linguistic innateness hypothesis by presenting its basic tenets, its biological and cognitive assumptions, and its main arguments. The formulation of Universal Grammar (i.e. the set of innate principles that linguistic innateness theory argues for) will be likewise presented. The introduction of Universal Grammar is of crucial importance, since one of the principles which is supposedly contained by it is precisely structure dependence. In section 3, I offer a description of the most important argument utilized to support claims of innateness, known as the Argument from the Poverty of the Stimulus. This argument is critical in the postulation of structure dependence as an innate principle. Afterwards, a thorough definition of structure dependence in relation to English Grammar (and more precisely, English syntax) is introduced. For this purpose, examples and analyses of yes-no question formation will be offered. Subsequently, I also present a review of some experiments which tested children’s supposed innate knowledge of structure dependence through yes-no question production. Their results, which tentatively support the view that structure dependence is an innate principle of human language, will be likewise presented. In section 4, I introduce criticism on both the methodology and the claims of linguistic innateness theory. In this manner, I present a brief review on the conclusions reached by several authors who, by resorting to evidence found in English language corpora, reject the premise that structure dependence is innate. Contrary to these conclusions, I subsequently offer a possible confirmation of the claim of innateness concerning structure dependence, which is also based on empirical data withdrawn from corpora. Finally, in section 5, I advance a brief critique on the notion of ‘innateness’ that linguistic innateness hypothesis relies on by resorting to current findings within the field of developmental biology. Thus, I challenge the notion of ‘innateness’ that linguistic innateness depends upon by explaining its flaws from a biological perspective, while also defending the view that a more biologically accurate notion of innateness is possible and certainly applicable to linguistic innateness hypothesis. 3 1. Scientific and philosophical background In this section, I will provide a brief overview of the bases of both rationalism and empiricism, the two philosophical trends which dominated Western thought during most of the Modern era. These two major – and for the most part, opposing – theories are the philosophical predecessors of current nativist and empiricist scientific approaches. As such, a short exposition of their principles is crucial for the comprehension of the state-of-the-art in contemporary linguistics, especially regarding the theory of linguistic innateness that will be explored in sections 2 and 3. Subsequently, I will also offer an update of this traditional opposition by resorting to current general biology and the cognitive sciences. The presentation of current scientific correlations of traditional rationalism and empiricism is of great importance when it comes to the consideration of language, since these trends dominate contemporary scientific studies. Furthermore, a thorough and accurate exploration of language and its properties cannot but found itself on facts drawn from studies of the brain; for the brain/mind is the ultimate base of language. With this purpose, the most recent developments in the fields of cognitive science and biology will be presented, especially in relation to three pivotal concepts: learning, innateness and models of mental architecture. Lastly, it should be noted that the content provided within this section will have a very broad span, not referring to linguistic matters in particular, but to issues regarding knowledge and cognition in general.1 1.1 Rationalism vs. empiricism Firstly, it should be noted that ‘rationalism’ and ‘empiricism’ are not labels that denote specific and unified theories, but rather names that refer to two broad philosophical trends (Schwartz 1999:703). It must also be taken into account that, despite being mainly opposing theories, rationalism and empiricism did share some common ground. 1 These, in turn, yield consequences that also apply to language. 10 11 2. Linguistic innateness In this section, I will present a detailed depiction of the basic tenets of linguistic innateness. With this purpose in mind, a presentation of its theoretical standpoints will be offered, along with a discussion of the biological and cognitive presuppositions that this theory relies on. Finally, the logical argumentation that comes from the nativist treatment of the APS will lead to the exposition of the most central aspect within innateness theory: the formulation of the set of allegedly innate principles it argues for, known as Universal Grammar. 2.1 The principles of linguistic innateness theory As we have seen in the previous section, the central assumption of a nativist program is that the environment in itself has no structure; as a consequence, all laws of order – whether cognitive, biological or linguistic – must come from within (Piattelli-Palmarini 1980:10). The foundation of knowledge, then, relies on “universal inborn structures” which allow the organism to impose order upon perceptual data, instead of deriving order from it – which is what empiricist theories typically hold (Piattelli-Palmarini 1980:10). These “laws of order” are considered by nativists to be invariable over time and across individuals; in other words, they are viewed as species-specific properties and, therefore, as genetically determined (Piattelli-Palmarini 1980:10). This is the base of any nativist program, and linguistic innateness is no exception. Linguistic innateness hypothesis is one of the most notable actualizations of current nativism. Its proposition relies largely on the work of revolutionary linguist Noam Chomsky, who formulated most of the theoretical components of this hypothesis, along with many of the pivotal arguments that support it (Wexler 1999:408). The core of linguistic innateness relies on the premise that very specific information, i.e. linguistic principles or patterns, are already imprinted in the organism before the environment can reveal these structures (Piattelli-Palmarini 1980:12). To put it simply, innateness theory proposes that linguistic knowledge is “wired in” (Anderson & Lightfoot 2000:698), especially when it comes to structure of language. So, defenders of this hypothesis attribute implicit linguistic knowledge 12 to the human species, maintaining that certain structural aspects of our linguistic capacity are innate, genetically determined, part of our biological endowment and, as such, a true speciesspecific property (Chomsky 1988:2–3; Hoekstra & Kooij 1988:52). As a consequence, the learning of a language would result from a specific innate capacity rather than by inductive observation of that language (Anderson & Lightfoot 2000:698). Furthermore, renowned nativists such as Crain & Pietroski (2001:178) have postulated the following: [M]any aspects of adult grammar are innate and in place at a very early age. These innate linguistic principles define a space of possible human language – a space the child explores, influenced by her environment, until she stabilizes on a grammar equivalent to that of adults in her linguistic community. In this manner, language acquisition is thought to be determined by the same genetically encoded principles which delineate the possibilities of human language; these derive from the genetic make-up of the human species (Hoekstra & Kooij 1988). These principles, according to nativists, are abstract in the way that they allow the child to learn any language, yet precise in the fact that they meet several restrictions which ultimately characterize the possibilities of human language (Hoekstra & Kooij 1988:45). In other words, innateness hypothesis maintains that our (innate) linguistic capacity is what shapes human languages, and places such constraints on its possibilities for grammar. Therefore, arguments for linguistic innateness cannot be separated from the fact that grammars have idiosyncratic constraints (Crain & Pietroski 2001:159). The logical outcome of this is that properties of specific languages could be derived from (for they would be determined by) those of our innate linguistic knowledge (Chomsky 1988:40). This specific aspect of linguistic innateness theory is stressed particularly in the Chomskyan nativist argument, as Chomsky emphasizes the fact that grammatical operations are governed by specific innate constraints (Crain & Pietroski 2001:164). 2.2 Modularity, the Language Faculty and the APS Nativist arguments rely on a number of cognitive and biological assumptions that are fundamental to the conception of language acquisition as ‘growth’ rather than ‘learning’ in its inductive sense (Anderson & Lightfoot 2000:710–711). One of them is modularity, that 13 is: the theory of a specialized, compartment-like brain. Modularity supports linguistic innateness because, as we have seen, the theory of a biologically specialized brain fits well with the hypothesis of having an innate set of specialized linguistic information (Hoekstra & Kooij 1988:42). This is so for the following reasons. Firstly, the nativist argument directly identifies language with grammar, and, more specifically, with syntax (Chomsky 1980a:55– 59), thus determining the type of linguistic information that is allegedly innate. Secondly, this grammar is said to be one module or subcomponent of the mind, which cooperates with other modules or cognitive capacities in the brain (Anderson & Lightfoot 2000:712). Thirdly, modularity maintains that any module, including the grammar, is likely to have clear initial and mature states, thus accounting for the supposed a priori linguistic principles (the initial state) which develop in time and through experience into the mature state (Anderson & Lightfoot 2000:712).6 This definition of grammar as one of the brain’s modules must not be confounded with the grammar of a language – henceforth GL – which is defined as a system of rules that explain the empirical data and which we use to communicate with other members of our linguistic community (Chomsky 1980a:75; Hoekstra & Kooij 1988:35). Similarly, ‘grammar’ as a mental module should not be mistaken by Chomsky’s concept of grammar as individual knowledge.7 Lastly, it is not to be confounded with a generative grammar either, which is the theory or detailed description of the grammar of a language.8 With the purpose of maintaining clarity, we will from now on refer to the concept of grammar as a mental module by using Chomsky’s own label: the Language Faculty – henceforth LF. Abiding strictly to modularity, innateness hypothesis maintains that the LF has a physiological base, to the point that it could be considered an “organ of language” (Anderson 6 According to Chomsky (1986:xxvi), linguistic principles do not generalize with respect to other cognitive domains, but are specific to language. Such a supposition is tentative; what’s more, the idea that innately given constraints are specific to language has been recently challenged (Marcus 1999:661; Tettamanti & Perani 2012:241). 7 Chomsky defined ‘grammar’ as a set of mental rules and representations of a given language; that is, a stable linguistic (i.e. syntactic) competence. It is, therefore, the mature system of knowledge an individual has of a particular language (Chomsky 1988:15, 36–37). This concept is referred to by Crain & Pietroski (2001:139) as ‘adult grammar’ in a perhaps more elucidating manner. 8 A generative grammar is a sufficiently explicit theory of a language so as to “predict an unbounded range of structured expressions [which] can be tested for empirical adequacy by investigating the accuracy of these predictions” (Chomsky 1988:61). 14 & Lightfoot 2000; Chomsky 1980a:60; Frazier 1999:557–558). Chomsky (1988:60) describes the LF as: [A] component of the mind/brain, part of the human biological endowment. Presented with data, the child, or, more specifically, the child’s language faculty, forms a language, a computational system of some kind that provides structured representations of linguistic expressions that determine their sound and meaning. According to Chomsky, the LF is said to incorporate “quite specific principles that lie well beyond any ‘general learning mechanisms’” (1988:47). It is also thought to be the hypothetical housing of all grammatical operations in the brain, during and after the language acquisition process. Lastly, the consideration of the LF as an “innate property of the human mind” accounts for linguistic innateness’ approach to language in terms of human biology (Chomsky 1975:34; Boeckx 2006:67). Following Chomsky, innate linguistic rules are equated merely to syntactic rules, which are in turn thought to be predetermined by the broad innate principles housed in the LF module. Chomsky’s reduction of language to syntax was not only a theoretical standpoint, but also justified empirically by him (1980a:55–59). Through the use of evidence, Chomsky (1980a:59) argued that his theoretical partition paralleled the functional independence that the grammatical competence has with respect to the pragmatic competence. Moreover, other scholars have suggested that the initial separation Chomsky sketched might be corroborated by studies of the brain, supposedly proving that syntax is a specialized and autonomous component of a modular language processor (Anderson & Lightfoot 2000:717; Frazier 1999:557; Smith 1999:24, 133). Consequently, innateness theory proposes a theoretical separation between linguistic rules and other cognitive abilities – even with those that are also concerned with language, as it is the case with the conceptual and pragmatic systems (Hoekstra & Kooij 1988:42–45). In this manner, Chomsky established a necessary correlation between cognitive function (the computational aspect of language, i.e. syntax), its physiological base (the language organ) and the genetic code of the human species (the human genome) (Wexler 1999:408).9 9 According to Chomsky, native principles are expected to be found “physically represented in the genetic code and the adult brain, respectively, with the properties discovered in our theory of the mind” (1980a:82–83). Other authors offer a less drastic viewpoint by stating that linguistic innateness “seeks to identify information which must be available independently of experience, in order for a grammar to emerge in a child” (Anderson & Lightfoot 2000:703). Concerning genetics, they do not claim that such information is encoded in the human genome, since it could also result from epigenetic 15 Precisely due to its claim of universality, the scientific purpose of linguistic innateness is to explain the internal structure of the universal subject by formulating the set of very intricate and highly specific linguistic structures accessible to the organism (Piattelli- Palmarini 1980:11).10 To put it simply, this hypothesis intends to define exactly what is innate in our linguistic knowledge. The linguist’s aim is, therefore, to analyze in depth and characterize in detail an ‘ideal speaker’ through appropriate abstractions from empirical givens (Piattelli-Palmarini 1980:10). Problems arise due to the consideration of the LF as a mental organ, since it poses several difficulties in its physiological or empirical study. Nevertheless, nativists affirm that, contrary to common belief, innateness is not beyond investigation (Hoekstra & Kooij 1988:47). In theory, linguistic innateness is supposedly supported by linguistic research. Such inquiries are carried out by studying thoroughly the logical possibilities of linguistic facts, which are drawn from linguistic data. In fact, innateness hypothesis maintains that only by extensive and detailed linguistic analysis about language structure – the principles of human grammars – can linguists realistically address the question of whether these principles can or cannot be learned or drawn from experience, or whether they must be innately specified (Crain & Pietroski 2001:140). This is precisely the theoretical standpoint that is behind the experimental study of structure dependence, an aspect of English Grammar which will be presented in section 3. So proponents of innateness have attempted to demonstrate that the stable states children achieve – adult grammar – are underdetermined in theoretically important respects by the linguistic input they receive (Crain & Pietroski 2001:139). Nativists hold that the primary linguistic data – henceforth PLD – available to children are insufficient for them to acquire the mature knowledge of language that they do gain once the process of language acquisition is over. This would still remain true, apparently, even given optimistic assumptions about the nonlinguistic capacities of children, especially the capacities to form and test generalizations based on their experience (Crain & Pietroski 2001:139, 153). This form of inductive indeterminacy is a “problem of under-determination of theory by data, applied to language learning” (Marcus 1999:660). Within linguistic innateness theory, the and developmental properties of the organism; notwithstanding, despite various doubts regarding domain specificity or genetic determination, nativists have no doubt that such innate constraints exist (Anderson & Lightfoot 2000:703; Crain 1991:597; Lightfoot 1999:62; Marcus 1999:661). 10 A universal subject is a prototypical individual of the human species. 16 insufficiency of the PLD is commonly referred to as the APS or Plato’s problem. This is a pivotal and necessary argument that would, if proven to be true, provide irrefutable support to linguistic innateness. If we accept that there is indeed poverty in the linguistic data, the crucial but limited role of experience in language acquisition would be logically derived from this argument. In this case scenario, experience would provide fundamental data in the process of language acquisition; it simply would not be the source of all linguistic knowledge (Chomsky 1980a:66). According to nativists, the child’s PLD does not provide the basis for attaining certain aspects of linguistic knowledge; therefore, “another source must exist for that knowledge” (Anderson & Lightfoot 2000:710). In order to fill the empirical gap that the APS proposes, theoretical linguists have formulated a rich system of linguistic principles which would constitute the innate aspects of our linguistic knowledge, thus making language acquisition possible in spite of impoverished data. This is Universal Grammar. 2.3 Universal Grammar Universal Grammar – henceforth UG – is the theory that explains the internal organization of the LF at the initial state (Crain 1991:598). It is a structural characterization of the genetically-determined LF that includes its specific components (Chomsky 1986:3, 1988:61). The content of UG is the “body of ‘hard-wired knowledge’” that takes part into the operation of the LF, reflecting its general computational properties (Chomsky 1988:25, 61; Pesetsky 1999:476). To put it simply, UG is the set of ‘unlearnable’ native grammatical principles that govern human language; these constraints are broad and universal (Chomsky 1988:3; Crain 1991:598, 611; Hoekstra & Kooij 1988:45).11 The properties of UG impose severe restrictions on the type of possible human grammars, thus accounting for the fact that many of its principles are thought of as constraints (Chomsky 1980a:66). UG is a system of conditions on possible grammars; the principles of UG provide both a restrictive and schematic model to which languages do conform, as well 11 These principles are characterized as ‘unlearnable’ by innateness theory due to the claims that nativists make concerning the APS: as we have seen, some aspects of language cannot be learned from the PLD, and as such, they are ‘unlearnable.’ 17 as certain heavy constraints that determine how grammars can function (Chomsky 1968:63). In this manner, possibilities of linguistic diversity are narrowly limited by the principles of UG, so that no natural language can exist if it violates these conditions (Anderson & Lightfoot 2000:706). This is so because properties of the UG “reflect innate features of human grammars” which hold universally (Crain & Pietroski 2001:152). Additionally, although it may seem contradictory at first, the systems that make up UG also account for linguistic diversity and variability in structure (Chomsky 1980a:66–67). There are several theories which justify how variation is incorporated into UG; the proposal that is closest to Chomskyan linguistics involves the abidance to a binary parameter-setting model (Pesetsky 1999:477). This kind of parameter-setting model understands UG as a system composed of unifying principles or constraints, and a number of parameters (Chomsky 1980a:66). On the one hand, parameters are made up by a limited and ordered series of hypotheses or values, representing a purported universal syntactic principle; according to a binary parameter-setting theory, parameters are constituted by both an unmarked and a marked value (Crain 1991:601; Crain & Nakayama 1987:541). During language acquisition, the child’s task is to set the parameters with the correct value of the target grammar in the view of the positive data that they receive (Crain & Nakayama 1987:541). On the other hand, constraints are not parameterized because they represent only one possibility, as opposed to parameters, which represent at least two; in this manner, constraints strongly delimit certain aspects of grammar which are said to be completely universal (MacWhinney 2004:888).12 The structure of Chomsky’s parameter-setting model of UG allows for the fact that setting a particular parameter may, in turn, set other parameters, thus having repercussions throughout the child’s grammatical system (Crain & Nakayama 1987:541). The content of UG, therefore, is considerably rich in its deductive structure, but with: (i) parameters that remain open to be fixed by triggering experience, and (ii) constraints that delimit the possibilities of human language (Chomsky 1980a:66; Hoekstra & Kooij 1988:39). Notwithstanding, the principles of UG are not derivable from the data, for although UG does specify a series of parameters, the values in which they can be set vary across languages (Hoekstra & Kooij 1988:39). In other words, UG is made up by general properties 12 One purported constraint contained by the theory of UG is structure dependence, a property which will be explored in detail in section 3. 18 of language, although the concrete realization of those abstract forms differs from language to language (Chomsky 1988:14). Logically, it follows that the grammars of all natural languages “share a common core” (Hoekstra and Kooij 1988:35). This necessary plasticity imposes quite a strong empirical demand on the theory of UG, yet it is supported by the supposed deep similarity among the world’s languages (Anderson & Lightfoot 2000:703, 715). Anderson & Lightfoot (2000:703–704) define UG as a ‘biological entity’ for, in fact, its characterization within innateness hypothesis cannot be attained without resorting to general biological terminology. Since UG is the initial state of the language learner (i.e. the initial state of the LF, before it has been presented with linguistic data), it must also be the basis on which language develops (Chomsky 1980a:69, 1988:61). As such, the set of innate properties that constitute UG must be genetically determined in the human genome, i.e. the entire genetic material or DNA of the species (Homo sapiens) (Wexler 1999:408). Logically, these genetically-determined principles should also be part of every individual’s genotype, i.e. the organism’s particular DNA (Jablonka & Lamb 2005:28). In turn, the genotype delineates the possibilities or potential of the organism’s development into the phenotype, i.e. the individual’s specific characteristics (Jablonka & Lamb 2005:28). That means that the same genotype could potentially develop into several phenotypes. However, these possibilities are not endless, for they are prepared for and limited by the genotype of the organism. As Anderson & Lightfoot explain, “[e]ach individual’s genotype determines the potential range of functional adaptations to the environment” (2000:703). Were UG to be part of the species genome, all prototypical humans would have the same linguistic potential for functional adaptations into any particular grammar, depending entirely on the PLD offered to them and not at all on variation in their genetic make-up (Anderson & Lightfoot 2000:703). To sum up, innateness hypothesis assumes that the linguistic genotype or UG is uniform across the species, given the absence of severe and specific pathology (Anderson & Lightfoot 2000:703; Boeckx 2006:68). Thus, UG allegedly determines the possible paths of maturation from the initial state to a specific mature state (the actual grammar of a particular language, or GL). UG, then, is a genotypic determination within the organism; environmental differences bring about phenotypic variation which account for linguistic diversity and the possibility that any person 19 might acquire any natural language (Hoekstra & Kooij 1988:38). As Chomsky points out (1988:65), [T]he languages of the world appear in all sorts of respects, but we know that they must be cast from the same mold, that their essential properties must be determined by the fixed principles of universal grammar. If that were not so, it would not be possible for the child to learn any one of them. At this point, it should be stressed that experience does provide relevant data for the choice of values in the parameters, assuming a fundamental but passive role in language acquisition; to put it simply, “environmental stimulus is just a trigger” (Lightfoot 1999:64). What’s more: it is assumed that if the UG or initial state is rich enough, even limited data will produce a highly articulated system of rules in the organism, i.e. an adult grammar (Chomsky 1980a:66). In conclusion, linguistic innateness holds the view that human language arises “in biologically based ways that are quite comparable to those directing other aspects of the structure of the organism” (Anderson & Lightfoot 2000:700; Chomsky 1986:2).13 Therefore, the construction of a theory of UG would be the solution to the APS (Chomsky 1988:61). There is quite a number of empirical evidence that plausibly validates the existence of UG (Pesetsky 1999:476–478). The most obvious one would be identification of linguistic universals through cross-linguistic research. Linguistic universals are principles or patterns that appear recurrently in the world’s languages and are, quite logically, predicted by the theory of UG (Hoekstra & Kooij 1988:45). Another possible evidence of UG would be the rapid rate at which language acquisition takes place, along with the existence of a critical period for language learning (Anderson & Lightfoot 2000:710; Hoekstra & Kooij 1988:36; Meisel 2013). Moreover, innateness hypothesis in general seems to be supported by evidence that some language deficits “show a clear distribution within families that epidemiological and other studies show to be just what would be predicted of relatively simple heritable traits” (Anderson & Lightfoot 2000:699). In addition, the development of structurally poor pidgins into creoles also appears to be another source of evidence of the existence of UG, since this reorganization supposedly involves the implementation of principles found precisely in this theory (Anderson & Lightfoot 2000:716; Pesetsky 1999:478). In relation to this, the fact that there exist deep structural similarities between oral and signed languages could also be 13 It is in this vein that Lightfoot boldly characterizes language acquisition as “language growth” (1999:64; my italics). 26 3.2.1 Structure dependence: definition and relation with the APS and UG As its own name indicates, structure dependence is a property that deals with the structure or computational characteristics of language. The rules of language are structure dependent because they depend on the structure of constructions and their constituents in order to operate (Chomsky 1988:45). This means that the structure of language has a hierarchical (i.e. non-linear) organization (Chomsky 1988:45). Contrary to structure dependence, systems which rely upon structure-independent rules include operations that are contingent on linear order, lacking hierarchical organization (Crain & Nakayama 1987:522; Crain & Thornton 1998:165). So, whereas a structure-independent rule may rely on arithmetic concepts such as ‘first,’ ‘third’ or ‘last,’ a structure-dependent rule relies on concepts such as ‘main constituent,’ ‘subordinated constituent,’ and so on. Structure dependence is a relevant, non-trivial property of grammars; in fact, it is not a logical necessity, but rather a fact, about human language (Chomsky 1968:61, 1975:32, 1988:47, 59).17 By any reasonable standards, structure-dependent rules are far more computationally complex – and, as such, more unlikely to be hypothesized by children – than structure-independent rules (Chomsky 1968:62, 1975:32). This is so because, on the one hand, the use of structure-dependent rules implicates a complex computational analysis of the structure of linguistic data, which requires the classification of elements into grammatical categories, the recognition of boundaries between such elements, the combination of these into abstract phrases, and the knowledge of the structural position that they have within their hierarchical disposition (Chomsky 1975:32, 1988:43; Tettamanti & Perani 2012:234).18 This would require the language learner to have early access to abstract structural notions (Crain 17 It should be noted that hierarchical structuring principles or structure-dependent operations are present at all levels of the structure of language (Chomsky 1968:63; Tettamanti & Perani 2012:230, 232); however, this work focuses on syntactic structure dependence only. 18 As Chomsky (1975:32) points out, these phrases are ‘abstract’ in the sense that their boundaries or categorical classification (noun phrase, verb phrase, and so on) need not employ marks of any kind, neither in production nor in perception. In this manner, language is perceived as a “phonetic stream of words” (Crain & Nakayama 1987:522; Crain & Thornton 1998:166). It should also be considered that, even though structure dependence is present at all linguistic levels, only part of these hierarchical properties and relations are evident at the level of production and perception (Tettamanti & Perani 2012:232). What’s more, “[m]ost hierarchical relations are established at a deep abstract level and are not obviously mirrored by surface structures, [which] poses intriguing questions about the mechanisms underlying language acquisition” (Tettamanti & Perani 2012:232). 27 & Thornton 1998:165; Tettamanti & Perani 2012:234).19 On the other hand, the application of the computationally simpler structure-independent rules only requires the speaker to be able to identify the units which conform a sequence and their linear order (Chomsky 1988:43). Consider the following examples of yes-no question formation proposed by Chomsky (1975:30–31):20 (1) The man is tall. (2) Is the man tall? Chomsky’s proposed sentences (1–5; see below) are considered to be the type of most obvious examples of structure dependence in English Grammar. Example (1) is a simple declarative clause, while example (2) is its interrogative counterpart. In order to render an interrogative clause out of an initial declarative one, the verb from the main clause in example (1) needs to be moved to the front of the sentence, thus yielding Subject/Auxiliary inversion, as illustrated in (2) (Berwick et al. 2012:25; Chomsky 1988:43; Crain & Nakayama 1987:526–527). In simple cases like (1), a linear rule such as “Move the first verbal element to the front of the clause” would definitely produce a grammatical interrogative like the one in example (2). However, the same linear rule would fail utterly at generating grammatical results for complex cases such as (3), which is illustrated in (4): (3) The man who is tall is in the room. (4) *Is the man who tall is in the room? This is so because such a structure-independent rule would rely on ‘counting’ the occurrences of verb forms while ignoring their structural position, thus selecting the auxiliary verb from the subordinate clause for Subject/Auxiliary inversion, instead of the verb from the main clause. So, the only correct rule is the structure-dependent one, which contemplates the structure of its constituents, and not their order, as shown in (5) (Berwick et al. 2012:25): (5) Is the man who is tall in the room? 19 Cf. section 1.2, p. 7, where it is stated that nativist models rely on the learner’s innate knowledge of abstract principles. 20 Yes-no question formation is the formulation of a closed interrogative clause (or yes-no question, or polar interrogative) out of an initial declarative one. In many of the interrogative examples proposed in this work, an underscore ( ) will be introduced to mark the structural position that the verb would have if it were not fronted, in order to better illustrate the workings of structure-dependent rules. 28 Contrary to what may seem logical, there is no a priori necessity for human grammars to use exclusively structure-dependent rules instead of structure-independent ones (Chomsky 1968:61–62, 1988:46). Codes which use the simpler linear rules – thus relying upon ‘counting’ operations – can be easily constructed (Berwick et al. 2012:26). Notwithstanding, any such code – no matter how complex, or how well it fulfilled the purposes of communication – would never be an instance of human language, because language invariably lacks ‘counting’ rules (Berwick et al. 2012:26; Chomsky 1968:62, 1988:46). Consequently, linguistic innateness hypothesizes that the LF is different in kind from other cognitive abilities (Smith 1999:133). Whereas other mental processes require ‘counting’ operations and pattern recognition, this approach at processing supposedly fails at explaining how language works, because “grammars can’t count: rather, all linguistic operations are structure dependent” (Berwick et al. 2012:26; Smith 1999:133–134).21 As a result, linguistic rules invariably consider structure-dependent operations, despite the fact that their production is merely sequential (Chomsky 1988:43, 45; Tettamanti & Perani 2012:231). The powerful statements generated by the APS stream from the fact that both linear and structure-dependent rules are available for human mental processes, yet only structuredependent rules apply to grammar (Chomsky 1975:32, 1988:48). Apparently, children make many mistakes during language acquisition, but never those which involve the use of structure-independent rules (Boeckx 2006:73; Chomsky 1975:31, 1980b:40). It seems that 21 Evidence from Christopher, the polyglot savant who has been extensively documented over the years, supports such a claim (Smith 1999:24, 134). He is considered to be a paramount example of how linguistic competence and general intelligence dissociate (Smith 1999:24). Christopher, along with a control group of undergraduates, was charged with the task of learning a ‘language’ which contained several structure-independent rules (Smith 1999:134). Such rules would be impossible in any natural language and, therefore, they would also be incompatible with the allegedly innate principles of UG. Unsurprisingly, and despite his extraordinary capacities for second-language learning, Christopher proved unable to learn such structure-independent rules (Smith 1999:134). What is more striking is that neither did the control group, regardless of their greater intelligence (Smith 1999:134). Consequently, it was hypothesized that the structure-independent rules were so linguistically alien to the undergraduates that they could not use their general-intelligence mechanisms to apply them to language learning (Smith 1999:134). They could, however, identify and solve computational problems of equal complexity in cases which did not entail linguistic processing (Smith 1999:134). Chomsky (1988:46–47) had previously predicted this by stating that, compared to structure-dependent rules, and in spite of being computationally simpler, learning structureindependent rules would require much more effort from children. Supposedly, children rely on automatic computational operations housed in the LF to process structure-dependent rules (Chomsky 1988:47). However, for structure-independent rules, children would need to perform such linear operations consciously, which is more costly (Chomsky 1988:47). 29 infants select the computationally more complex structure-dependent rules invariably and without error at all stages of language learning, eschewing the equally available structureindependent rules – even though these could be compatible with the majority of their PLD (Boeckx 2006:71; Chomsky 1988:48; Crain & Nakayama 1987:522; Crain & Pietroski 2001:162–163). This fact is even more striking considering that, according to the third level of data deficiency from the APS, children have no data to make such an unerring choice in opting for structure dependence (Chomsky 1988:50; Lightfoot 1999:51). This is the kind of error-free learning that nativists have postulated when it comes to innate constraints (MacWhinney 2004:888). As a result, nativists have claimed structure dependence to be a property of the LF, i.e. an innate constraint contained by UG (Chomsky 1980b:40, 1988:48). In conclusion, the knowledge of structure dependence seems to be an a priori condition of language, since children never hypothesize structure-independent rules in the first place (Chomsky 1968:62, 1975:32, 1986:7–8, 1988:44–45; Crain & Nakayama 1987:522). The proposal that structure dependence is an innate principle of the LF or UG, which operates upon the child’s judgment of the PLD and yields structure-dependent rules as the only candidates, has been subjected to empirical test (Chomsky 1975:32–33, 1988:44– 45). According to Crain & Pietroski (2001:162–163), every child that has been experimentally studied has adhered themselves to the structure-dependence hypothesis. These experiments and their corresponding results will now be presented. 3.2.2 Yes-no question formation: experiments and results Crain & Nakayama (1987) carried out a series of experiments to test children’s knowledge of structure dependence at several stages of language acquisition. For this purpose, the researchers designed a study which required children to engage in yes-no question formation; subsequently, they analyzed the kinds of hypotheses that children entertained for Subject/Auxiliary inversion. The reason for choosing this aspect of English Grammar is that, besides being a quintessential case of structure dependence, the formation of complex yes- 30 no questions in English is also thought of as the strongest case of stimulus absence (Pullum & Scholz 2002:36).22 In these experiments, the children’s task was to apply their grammatical knowledge to the formation of a complex yes-no question out of a given declarative sentence, a task which they were likely to not have attempted before (Crain & Nakayama 1987:527). Most of the complex targeted sentences contained a subordinated clause modifying the Subject noun phrase, in the manner of Chomsky’s example (5) (Chomsky 1975:31; Crain & Thornton 1998:165, 170). Additionally, most test sentences also contained one complex verb phrase in the subordinated clause – with one auxiliary and one main verb –, while the verb phrase in the main clause was simple. An example of the test sentences utilized in Crain & Nakayama’s experiments (1987) is given in (6): (6) The boy who is watching Mickey Mouse is happy. To construct complex polar interrogatives, children would need to know that the verb from the main clause – and not the one from the subordinate clause – is the one that must be moved to the front of the sentence. This operation is shown in (7): (7) Is the boy who is watching Mickey Mouse happy? Notwithstanding, it is of crucial importance to note once again that yes-no questions could also be subjected to structure-independent analyses in many cases; for, as we have seen, structure-independent hypothesis are compatible with most of children’s PLD (Crain & Nakayama 1987:522, 524; Crain & Thornton 1998:167). This has been previously demonstrated in Chomsky’s examples (1) and (2): to construct yes-no questions with simple clauses, a linear rule – which is arguably simpler in computational terms, and therefore more likely to be hypothesized by children – would indeed render grammatical results (Chomsky 1975:30, 1988:43; Crain & Nakayama 1987:526). Additionally, it should be noted that a high number of the test sentences in Crain & Nakayama’s study (1987) used the same verb, i.e. the verb to be, in both verb phrases, as it is illustrated in example (6). Apparently, the fact that a part of both verb phrases were identical could potentially encourage children to confuse the two and use a linear rule to move the auxiliary verb from the subordinate clause to the front, instead of moving the main verb from the main clause. Such an error is shown in (8): 22 Subject/Auxiliary inversion is used in English Grammar to mark interrogative sentence type, thus involving a purely syntactic issue (Pullum & Scholz 2002:38). 31 (8) *Is the boy who watching Mickey Mouse is happy? Surely, the structure-independent rule fails at producing complex yes-no questions in complex cases. Notions such as ‘main clause,’ ‘subject’ or ‘subordinate clause’ are necessary to carry out the correct analysis of the data and produce the correct hypothesis (i.e. a structuredependent rule), which in turn generates grammatical yes-no questions in all cases (Crain & Nakayama 1987:525–526). According to the theory of UG, this is the kind of knowledge that is native to children, which is what Crain & Nakayama (1987) put to the test. The thirty children who were tested in this study were aged 3 to 5 years old, an age range at which children have been observed to hypothesize wrong generalizations regarding grammatical rules (Crain & Nakayama 1987:527). Again, the formation of a complex yes-no question such as (7) is a task which children were likely to not have ever attempted before this experiment (Crain & Nakayama 1987:527). Since the structure-independent hypothesis is consistent with the subset of their PLD, if children lacked the structure-dependent constraint contained by UG, it would be expected that at least some of them might entertain structure-independent hypotheses in eliciting the required yes-no questions (Crain & Nakayama 1987:526; Crain & Thornton 1998:171). This, however, did not occur. Even though children produced many ungrammatical constructions, thus hypothesizing the ‘wrong’ rules for their target grammar, none of these were errors which specifically implicated a structure-independent hypothesis (Crain & Nakayama 1987:530). In other words, it seems that all children abided by structure dependence, thus lending presumptive support to Chomsky’s contention that children never consider structureindependent rules in the first place (Crain & Nakayama 1987:522, 530).23 According to the theory of UG, children never entertain structure-independent hypotheses even if their PLD is in accordance with such rules (Crain & Nakayama 1987:527, 533). On the contrary: they attempt to form complex yes-no questions by invoking their allegedly innate knowledge of structure dependence, no matter how computationally complex in comparison with linear 23 Moreover, the evidence from these experiments supports Chomsky’s prediction concerning the autonomy of syntax with respect to semantics (Crain & Nakayama 1987:522, 542). Smith (1999:126– 129) has also argued for the autonomy of syntax, in the light of Christopher’s disassociated abilities as well as other pieces of evidence which also seem to support modularity and domain specificity. Correlations between structure dependence, modularity and domain specificity are also apparently supported by evidence from neurophysiological investigations of the brain (with special attention on Broca’s area) explored in Tettamanti & Perani (2012). 32 rules (Crain & Nakayama 1987:527). As a result, the evidence gathered from these experiments corroborate that children’s grammatical hypotheses seem to be highly constrained by the principle of structure dependence (Crain & Nakayama 1987:542; Crain & Thornton 1998:174). Notwithstanding, it must be taken into account that evidence from these experiments are not conclusive, but merely supportive, of the claims of UG made by Chomsky; critical or decisive experiments remain to be reported (Crain & Nakayama 1987:524). In fact, Crain & Thornton (1998:172) emphasize the fact that Crain & Nakayama’s study (1987) could only argue for a weaker claim concerning the innate knowledge of structure dependence, by relying on the fact that blatantly structure-independent rules were never tested by children. In this manner, it is assumed that because children did not in fact produce any yes-no questions that violated the structure-dependence constraint, such a constraint is indeed innate and in place in their grammars, especially taking into account the fact that many ungrammatical questions were elicited, and none were of them followed a linear hypothesis (Crain & Thornton 1998:175). In conclusion, it is true that the formulation of structure dependence as an innate constraint would effectively solve the problem of the APS in all grammars, yielding a solution to the acquisition puzzle (Crain 1991:599; Lightfoot 1999:58). Nevertheless, as we have just seen, results from the yes-no question formation experiments did not indubitably attest for the innateness of structure dependence. In fact, many linguists have criticized nativists’ use of the APS, as well as their formulation of UG and the consideration of structure dependence as one of its innate constraints. Such criticism will be explored in the following section (4). 33 4. Debate on structure dependence and the APS In this section, I will offer a brief review of the current debate regarding the consideration of structure dependence as a quintessential case of the third data deficiency, which is the only decisive level of the APS. As we have seen in section 3, innateness hypothesis uses this argument to defend the thesis that structure dependence must be an innate constraint. In order to show the other side of the discussion, I will now present several critiques on nativist arguments applied to the case of structure dependence. Even though this issue has been addressed by many authors, due to space limitations I have made a short selection of the most relevant researchers in this matter. Firstly, these authors’ conclusions will be presented, along with the implications that their findings have regarding nativists’ claims of innate linguistic knowledge. Subsequently, I will contrast such findings with those of other scholars (Legate & Yang 2002) who have shown dissimilar results that support linguistic innateness. 4.1 Critiques on innateness hypothesis: a firm rejection of the third data deficiency As we have seen in section 3, the APS maintains that structure dependence is ‘underivable’ from children’s PLD (Chomsky 1988:47). Structure dependence is a typical example of the third data deficiency; this level of the APS is not based on stimulus poverty, but rather on stimulus absence (Pullum & Scholz 2002:16). Chomsky (1980c:114–115) has been extensively criticized for maintaining that children do not encounter direct evidence of structure-dependent rules; his claim is based on the assumption that complex cases in which the structure-dependent rule is the only possible hypothesis rarely arise. Following Chomsky (1980b:40), A person might go through much or all of his life without ever having been exposed to relevant evidence, but he will nevertheless unerringly employ [the structure-dependent hypothesis], never [the structure-independent hypothesis], on the first relevant occasion. Hornstein & Lightfoot (1981:9) have also argued that there is total lack of evidence for some aspects of grammar, especially for knowledge of structure dependence. This is quite a strong 34 nativist claim which has been heavily refuted over the years. As Sampson (1997:43) points out, Chomsky’s arguments were theoretically-based statements which he never validated with data withdrawn from corpora or studies about language acquisition. Many authors believe this to be the biggest flaw in Chomsky’s formulation of the APS (Pullum & Scholz 2002:45). After all, this argument – especially focusing on its third and decisive level – relies on the supposed absence of the stimulus for it to hold; thus, examining children’s PLD to check whether this lack of evidence can be confirmed empirically seems strictly necessary (Pullum & Scholz 2002:21). The importance of Chomsky’s lack of empirical corroboration is not to be underestimated, since it has thought to be enough of a reason to reject his premises until proven otherwise (Sampson 1997:43). In fact, authors such as Sampson (1997), Pullum & Scholz (2002) and MacWhinney (2004) have done exactly this: they have tentatively assumed that children do not lack information about the structure dependence of syntax. These scholars strongly discard the formulation of the APS, especially its third level of data deficiency. They suppose that the data are indeed rich enough for children to have clear evidence of structure-dependent rules, a thesis which they have attempted to demonstrate empirically (MacWhinney 2004:884; Pullum & Scholz 2002:11–17, 39–40; Sampson 1997:43).24 After all, there are constructions in English Grammar for which structure-dependent rules are the only possible hypotheses; these could potentially be part of a child’s PLD, and thus provide decisive evidence in choosing structuredependent over structure-independent rules in grammar formation (Boeckx 2006:70). This view, which challenges linguistic innateness hypothesis, is consistent with the position which Piaget originally assumed in relation to Chomsky. Despite the fact that both authors did agree on some important points regarding their conception of language and thought, Piaget (1980b:57) has made a strong critique of Chomsky’s defense of innate linguistic knowledge, basing his criticism on biology and psychology. Piaget (1980a:23) defended the thesis that no a priori or innate cognitive structures exist in the human species. As the canonical representative of constructivism, Piaget has postulated that language learning is an exemplary case of learning by induction (1980a:23–24, 1980b:58–59). He has 24 Although Pullum & Scholz (2002) do question nativist claims of innate knowledge of the structuredependence constraint (among others), they clearly advise against reading their paper as an instance of anti-innateness. According to them, their research simply challenges nativist statements that, to their mind, have not been properly ascertained with data (Pullum & Scholz 2002:10). 35 stressed the active role that the language learner supposedly has during language acquisition; in the light of experience, and thanks to their growth and general cognitive schemes, children construct knowledge about their language (Piaget 1980a:23–25). These learning mechanisms are completely general and operative since the moment of birth, and ultimately rely on the richness of the stimulus (Piaget 1980a:24).25 In this manner, although the child plays an active role in constructing their linguistic knowledge, only a PLD that is rich enough will provide the necessary inductive base for language acquisition. This is precisely what Sampson (1997), Pullum & Scholz (2002) and MacWhinney (2004) argue for in the case of structure dependence. The authors mentioned above maintain that it is impossible for English speakers to not have encountered any relevant sentences for structure dependence during their lifespan (Pullum & Scholz 2002:41). Notwithstanding, adult’s linguistic experience is unimportant to our case, because the question that we are concerned with is how children come to acquire knowledge about structure dependence; in this manner, the only relevant linguistic experience is that which is received during childhood. Although the ideal manner of testing the presence of direct evidence for structure dependence in children’s PLD would be to record the entire PLD of at least one child, this task is currently unattainable for obvious reasons (Pullum & Scholz 2002:21). As a consequence, linguists have turned to corpora instead, under the assumption that a property which is supposedly absent – as nativists argue – from children’s PLD will also be absent from any corpus that records child speech as well as speech addressed to children at the age of language acquisition (Pullum & Scholz 2002:21). Pullum & Scholz (2002:41–42) have made a very interesting case by expanding the type of sentence that would offer direct evidence of structure dependence. Besides Chomsky’s canonical example of polar interrogatives (1975:31) which have the same verb both within the main clause and the relative clause – as shown in example (5) –, Pullum & Scholz (2002:41–42) argue that any interrogative sentence which has an auxiliary verb within the Subject noun phrase also counts as crucial evidence for structure dependence. Therefore, 25 The general learning mechanisms that Piaget (1980a:24) argues for do not imply any sort of knowledge on behalf of the organism, since they are purely inductive. Thus, Piaget’s proposal differs greatly from a nativist one. 42 As we have seen in section 4, the nativist conclusion which is derived from premises 1–4 is reasonable, since it is arguable that there is not sufficient evidence for the learnability of structure dependence in the linguistic input. However, according to developmental biology, the content of premise 5 is false. The reasons for such a statement will be explored in the following section (5.2). 5.2 The (indirect) relationship between genes and traits To put it simply, premise 5 comes to state that a phenotypic trait is innate if its existence is determined by genetic factors (Mameli & Bateson 2006:158; Samuels 2004:137).28 This perspective, which was adopted by linguistic innateness hypothesis, was initially defended by one early 20th century framework within the field of biology known as Neo-Darwinism. This was an extremely genocentric trend which attributed the central role as causal agents to genes when it came to explaining innateness (Longa & Lorenzo 2012:56). As Longa & Lorenzo (2012:55) indicate, [G]enes were given priority over other developmental levels and resources, in such a way that they were considered the only possessors of the essential information guiding the growth and maturation of organic studies. From the 1960s onwards, molecular Neo-Darwinism came to argue for the existence of genetic programs or blueprints contained by the genes of the organism, thus explaining the relationship between genotype and phenotype in terms of plan and product while also reinforcing the primacy of the genetic level (Jablonka & Lamb 2005:33). In this manner, the development of the organism would merely consists on the unfolding of the information already contained in the genes (Longa & Lorenzo 2012:56). This genocentric perspective, which is highly problematic, is currently being challenged by a reacting framework known as developmental biology (Longa & Lorenzo 2012:57, 63).29 The developmental framework rejects genocentric premises, since genocentrism “ignores the contribution of many factors 28 Cf. section 3.2, p. 22 for a brief definition of genome, genotype and phenotype. The distinction between genotype and phenotype is crucial in genetics (Jablonka & Lamb 2005:28). 29 In fact, Oyama et al. (2001:5) affirm that there is no scientific evidence that proves the existence of genetic programs. 43 and resources located between genes and traits, without which development would not take place” (Longa & Lorenzo 2012:58). According to Neo-Darwinism, genes are hereditary units which carry information about the organism’s traits (Jablonka & Lamb 2005:24–29). In order to defend the view that genes contain information about specific traits, Neo-Darwinian biologists follow one of two possible approaches: either a ‘causal’ or a ‘representational’ account of genetic information (Samuels 2004:137–138). Both the causal and representational accounts deal with how the genetic information contained by the genes is supposedly accountable for the development of specific traits. With the purpose of introducing critiques from a developmental perspective, the two approaches that were traditionally adopted to defend the traditional notion of innateness will be presented. Firstly, the representational approach maintains that “traits are genetically determined if represented in (or encoded by) the genes” (Samuels 2004:137). This statement somehow implies that traits are pre-specified in the information encoded by genes. However, it has been noted throughout the years that the only information that genes or DNA sequences directly contain is the information which is involved in the production of proteins (Godfrey- Smith 2007:110; Jablonka & Lamb 2005:30; Longa & Lorenzo 2012:60; Mameli & Bateson 2006:159; Samuels 2004:138). Evidently, the claim that genes contain information about the phenotype is also accepted, but with certain limitations that must be taken into account. According to Godfrey-Smith (2007:106), [W]e can say that genes contain information about the proteins they make, and also that genes contain information about the whole-organism phenotype. But when we say that, we are saying no more than what we are saying when we say that there is an informational connection between smoke and fire, or between tree rings and a tree’s age. So, the notion of innateness defended by Neo-Darwinism claims that genes can represent the traits of the phenotype directly; according to developmental biology, “no such notion of [genetic] representation appears to exist” (Samuels 2004:138). As Wimsatt maintains, “[e]quating ‘innate’ with ‘genetic’ is a kind of functional localization fallacy,” which wrongly implies “assuming that the function of a larger system or subsystem is realized completely in a part of that system” (1999:160). Since the only genetic information directly contained by genes is that of constructing proteins, identifying innateness with the coding of the phenotype is plainly incorrect (Mameli & Bateson 2006:159). In other words, genes are 44 not part of the phenotype – at least directly – nor representations of it (Jablonka & Lamb 2005:28). Secondly, the causal account of innateness according to a Neo-Darwinist perspective maintains that “traits are genetically determined if caused (in the appropriate way) by genetic factors” (Samuels 2004:137). In this manner, a trait is hypothetically innate if it is caused entirely by genetic factors, so that the only thing required for its development are genes (Mameli & Bateson 2006:158; Samuels 2004:138). However, these claims are inaccurate. To begin with, there is no clear explanation as to what the ‘appropriate’ relation between genes and innate traits is (Samuels 2004:137). And most importantly, there is no such phenotypic trait for which only genes are needed for its development (Mameli & Bateson 2006:158). In other words, traits are not caused by genes alone: they also depend on non-genetic sources for the development of the phenotype (Jablonka & Lamb 2005:28; Mameli & Bateson 2006:158; Samuels 2004:138). Therefore, genetic information and innateness cannot be equated, because the genetic level is just one of the many biological levels that intervene in a given organism’s development. To prove this point, Johnston & Edwards (2002) designed a model of behavioral development constituted by 13 levels of interacting factors, only one of which corresponds to genetic activity.30 A list of the proposed factors goes as follows (Johnston & Edwards 2002:28): 1. Sensory Stimulation 2. Patterned Neural Activity 3. Neural Connectivity 4. Neural Growth 5. Non-neural Structures 6. Non-neural Growth 7. Individual Nerve Cell Activity 8. Extracellular Biochemistry 9. Cell Membrane 10. Intracellular Biochemistry 30 In spite of being fairly recent, this model (Johnston & Edwards 2002) has already become a classic in its field. 45 11. Protein Synthesis 12. Physical Influences 13. Genetic Activity It is stated that all biological levels, not just the genetic one, are needed for development, because “every trait is produced by the interaction of many developmental resources” (Oyama et al. 2001:2). So, while it is true that genes do influence the phenotype, they are not the only factor involved in development (Johnston & Edwards 2002:26). Furthermore, the route between genotype and phenotype is complex, and the effects of genes on the phenotype are very indirect because, among other reasons, the immediate consequences of genetic activity are confined to the cell (Johnston & Edwards 2002:28; Longa & Lorenzo 2012:64). Consequently, without appropriate environmental context, genes make nothing happen (Johnston & Edwards 2002:26; Longa & Lorenzo 2012:60). In fact, “the bare genes in isolation are among the most impotent and useless materials imaginable” (West-Eberhard 2003:93). Finally, it should be noted that findings within the field of developmental biology do not undermine the notion of ‘innateness’ at all; they simply argue for a different formulation of ‘innateness.’ According to developmental biology, innate traits are simply those “that reliably appear at certain points of a species-typical path of development” (Longa & Lorenzo 2012:65). Thus, the thesis that innate traits exist is still defended; these, however, do not need to be genetically determined to be considered innate. In this manner, it is not the logical argumentation of the APS what developmental biology challenges; in fact, premises 1–4 are accepted. Notwithstanding, this is not the case with premise 5, since this step equates ‘innate’ with ‘genetically determined,’ which is not biologically accurate according to the developmental framework. In conclusion, the result of development cannot simply be reduced to genetic activity because many equally important non-genetic factors are involved in this process (Longa & Lorenzo 2012:53, 58–60). This ultimately means that an innate trait need not be present in the genes of the organism, because (i) genes do not contain traits or programs, and (ii) genes are not the only causal agents of development (Godfrey-Smith 2007:114; Longa & Lorenzo 2012:53, 58, 65; Samuels 2004:137–138). Moreover, there is non-genetic inheritance, which means that genes are not the only recipients of hereditary information; consequently, 46 correlating ‘genetic’ with ‘innate’ is plainly false (Jablonka & Lamb 2005:32; Longa & Lorenzo 2012:63). For these reasons, it is evident that linguistic innateness hypothesis and the theory of innateness in general requires greater biological credibility, since its claim for genetic innateness has been dismantled by developmental biology. Moreover, I agree with those who encourage linguistic innateness theory supporters to avoid “the temptation to attribute magical powers to the genes” in arguing for the innateness of language (Johnston 1987:175). 47 Conclusion The study of structure dependence has generated a significant body of non-trivial linguistic knowledge about the syntax of the English language, which has allowed linguists to make tentative assumptions about what the source of linguistic knowledge may be. Both positions in this debate – i.e. arguing for innateness, as well as arguing against it – are still provisional. As it has been extensively shown, there is not yet sufficient evidence to conclude the argument. Admittedly, up-to-date controversial issues such as the claim of innateness and universality of structure dependence are yet to be fully attested. Consequently, it seems evident that additional research on this topic is needed. However, even though conclusive results remain to be reported, what has been achieved up to this point is not to be underestimated. The literature that has been created thus far conforms a secure base which will prompt future lines of research concerning structure dependence, English Grammar and Universal Grammar. In this respect, serious consideration of the criticism on the notion of ‘innateness’ offered by developmental biology could prove to be very useful in helping the theory of linguistic innateness progress realistically and in biologically accurate ways. In addition, we have also considered several authors’ critiques concerning linguistic innateness’ lack of empirical corroboration. Such criticism could also provide an appropriate base for progress and change within the theory of linguistic innateness, encouraging its development towards a linguistic theory which sought systematic empirical confirmation of its theoretical predictions. Finally, it is evident that the study and analysis of structure dependence is a crucial piece of evidence in the acquisition puzzle that will allow researchers to answer the question of whether some aspects of language are innate or not. Thus, the study of English syntax from a formal perspective continues to be of interest within the field of linguistics in general, as well as within the theory of linguistic innateness in particular. As I intend on furthering my studies on linguistics, it is possible that I will consider tackling this line of research in the future. However, for the time being, my only hope is that this work will be recognized as a 48 humble attempt at producing a worthy bibliographical revision on the topic of structure dependence, English Grammar and Universal Grammar. 49 Works Cited Anderson, S. & D. Lightfoot (2000): “The human language faculty as an organ.” Annual Review in Physiology 62/1: 697–722. Berwick, R.C., N. Chomsky & M. Piattelli-Palmarini (2012): “Poverty of the stimulus stands: Why recent challenges fail.” In M. Piattelli-Palmarini & R. Berwick (eds.): Rich languages from poor inputs. Oxford: Oxford University Press, 19–42. Boeckx, C. (2006): “Universals in a generative setting.” In R. Mairal & J. Gil (eds.): Linguistic universals. Cambridge: Cambridge University Press, 67–79. Chomsky, N. (1968): Language and mind. New York: Harcourt Brace. ---. (1975): Reflections on language. New York: Pantheon. ---. (1980a): Rules and Representations. Oxford: Blackwell. ---. (1980b): “On Cognitive Structures and Their Development: A Reply to Piaget.” In M. Piattelli-Palmarini (ed.): Language and Learning. The Debate between Jean Piaget and Noam Chomsky. London, Melbourne and Henley: Routledge & Kegan Paul, 35– 54. ---. (1980c): “The Linguistic Approach.” In M. Piattelli-Palmarini (ed.): Language and Learning. The Debate between Jean Piaget and Noam Chomsky. London, Melbourne and Henley: Routledge & Kegan Paul, 109–117. ---. (1986): Knowledge of language. Its nature, origins, and use. New York: Praeger. ---. (1988): Language and Problems of Knowledge. Cambridge, MA and London, England: MIT Press. Crain, S. (1991): “Language acquisition in the absence of experience.” Behavioral and Brain Sciences 14/4: 597–612. Crain, S. & M. Nakayama (1987): “Structure dependence in grammar formation.” Language 63/3: 522–543. Crain, S. & P. Pietroski (2001): “Nature, nurture, and Universal Grammar.” Linguistics & Philosophy 24/2: 139–186. Crain, S. & R. Thornton (1998): Investigations in Universal Grammar: A guide to experiments on the acquisition of syntax and semantics. Cambridge, MA: MIT Press. 50 Frazier, L. (1999): “Modularity and Language.” In R. Wilson & F. Keil (eds.): The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press, 557–558. Gelman, R. (1999): “Cognitive development.” In R. Wilson & F. Keil (eds.): The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press, 128–130. Gelman, S. (1999): “Domain specificity.” In R. Wilson & F. Keil (eds.): The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press, 238–240. Godfrey-Smith, P. (2007): “Information in biology.” In D. Hull & M. Ruse (eds.): The Cambridge Companion to the Philosophy of Biology. Cambridge: Cambridge University Press, 103–119. Hoekstra, T. & J. Kooij (1988): “The innateness hypothesis.” In J. Hawkins (ed.): Explaining language universals. Oxford: Basil Blackwell, 31–55. Hornstein, N. & D. Lightfoot (1981): “Introduction.” In N. Hornstein & D. Lightfoot (eds.): Explanation in Linguistics. The Logical Problem of Language Acquisition. London: Longman, 9–31. Jablonka, E. & M.J. Lamb (2005): Evolution in four dimensions. Genetic, epigenetic, behavioral and symbolic variation in the history of life. Cambridge, MA: MIT Press. Johnston, T. (1987): “The persistence of dichotomies in the study of behavioral development.” Developmental Review 7: 149–182. Johnston, T. & L. Edwards (2002): “Genes, interactions, and the development of behavior.” Psychological Review 109: 26–34. Karmiloff-Smith, A. (1999): “Modularity of mind.” In R. Wilson & F. Keil (eds.): The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press, 558–560. Keil, F. (1999): “Nativism.” In R. Wilson & F. Keil (eds.): The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press, 583–586. Legate, J. & C. Yang (2002): “Empirical reassessment of the poverty stimulus argument.” The Linguistic Review 19/1–2: 151–162. Lightfoot, D. (1999): The development of language. Acquisition, change and evolution. Malden: Blackwell, 49–76. Longa, V.M. & G. Lorenzo (2012): “Theoretical linguistics meets development.” In C. Boeckx et al. (eds.): Language, from a biological point of view. Current issues in biolinguistics. Newcastle upon Tyne: Cambridge Scholars Publishing, 52–84. 51 MacWhinney, B. (2004): “A multiple process solution to the logical problem of language acquisition.” Journal of Child language 31: 883–914. Mameli, M. & P. Bateson (2006): “Innateness and the sciences.” Biology and Philosphy 21: 155–188. Marcus, G. (1999): “Poverty of the Stimulus Argument.” In R. Wilson & F. Keil (eds.): The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press, 660–661. Meisel, J.M. (2013): “Sensitive phases in successive language acquisition: The critical period hypothesis revised.” In C. Boeckx & K.K. Grohmann (eds.): The Cambridge Handbook of Biolinguistics. New York, NY: Cambridge University Press, 69–85. Oyama, S., P. Griffiths & R. Gray (2001): “Introduction.” In S. Oyama et al. (eds.): Cycles of contingency. Developmental systems and evolution. Cambridge, MA: MIT Press, 1–11. Pesetsky, D. (1999): “Linguistic universals and Universal Grammar.” In R. Wilson & F. Keil (eds.): The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press, 476–478. Piaget, J. (1980a): “The Psychogenesis of Knowledge and Its Epistemological Significance.” In M. Piattelli-Palmarini (ed.): Language and Learning. The Debate between Jean Piaget and Noam Chomsky. London, Melbourne and Henley: Routledge & Kegan Paul, 23–34. ---. (1980b): “Introductory Remarks.” In M. Piattelli-Palmarini (ed.): Language and Learning. The Debate between Jean Piaget and Noam Chomsky. London, Melbourne and Henley: Routledge & Kegan Paul, 57–61. Piattelli-Palmarini, M. (1980): “The Chomskian Hard Core.” In M. Piattelli-Palmarini (ed.): Language and Learning. The Debate between Jean Piaget and Noam Chomsky. London, Melbourne and Henley: Routledge & Kegan Paul, 9–12. ---. (1989): “Evolution, selection and cognition: from ‘learning’ to parameter setting in biology and in the study of language.” Cognition 31/1: 1–44. Pullum, G. & B. Scholz (2002): “Empirical assessment of stimulus poverty arguments.” The Linguistic Review 19/1–2: 9–50. Reisberg, D. (1999): “Learning.” In R. Wilson & F. Keil (eds.): The MIT Encyclopedia of the Cognitive Sciences. Cambridge, MA: MIT Press, 460–461.