There are two main streams for the language acquisition theory, Chomskian approach and Tomasello's approach. The former is Generative Grammar which is the theory where language is acquired on the basis of the innate language faculty which consists of the module isolated from other cognitive systems, while the latter is based on Developmental Psychology and Cognitive Linguistics where language is inseparable from cognitive abilities in general and where language acquisition requires various factors rather than one special faculty only for language.
The aim of this paper is to give a broad overview of these two inverse approaches, and to discuss that Tomasello's approach can be more convincing than Chomsky's approach. To begin with, Chomskian approach (i.e. Generative Grammar) is introduced in 2.1. Although the basic idea of this theory has not changed, Chomsky, who is the pioneer of Generative Grammar, has renewed his theory several times; therefore, this paper briefly surveys some relevant points in order to grasp the general idea of the theory. On the other hand, Tomasello's approach (i.e. the Usage-based Theory) is considered to have its basis on Developmental Psychology and on Cognitive Linguistics, so his approach is overviewed in 2.2 with regard to these two fields. In addition, this paper shows its position to support Tomasello's approach in section 3. Finally, concluding remarks are given in the last section.
Chomskian Generative Grammar
2.1.1 Basic Ideas in Chomskian Theory
Chomsky has tried to explore the reason why children acquire language so fast even though the input data available to them are insufficient, and why its stages take much the same course for all children and all languages despite unequal quality and quantity of the input data. He is also concerned with creativity and recursion of language by which children can understand and generate infinite number and length of sentences (Chomsky 1972, 1995).
The quantity of input data is one of the most prominent issues known as the poverty of the stimulus in Generative Grammar: children can eventually acquire a complete set of knowledge of grammar despite the fact that certain kinds of sentences generated from certain types of the rules do not appear as inputs around them. Moreover, children are not always exposed to the grammatical sentences because utterances produced by people around them sometimes do not follow with the grammatical rules. This means the input data is insufficient for learning grammar. In addition, adults usually correct not children's grammar but the truth of their utterances; hence children cannot learn grammar by means of the feedback from adults. Furthermore, Chomsky presumes that, although there is the wide variation from language to language, all languages share universal features such as structure-dependent nature of language. To address these issues, Chomsky gives a psychological and biological dimension to his enterprise. He assumes the innate language faculty, Universal Grammar (UG), and considers language learning as only a mere trigger which helps children with setting the parameters within UG.
Chomsky's theory has revised several times but there are some enduring characteristics which can be observed consistently: linguistic knowledge consists of a module separated from other various mental faculties, which means it is autonomous in mind; language shows its species-specific nature in that only human beingss can acquire language; linguistic knowledge is not ready-made faculty for use but genetically programmed; the language module containing UG automatically mature according to an innate biological timetable as a child grows.
In order to study language, Chomsky (1998: 15) refers to three central concerns in his framework as follows:
a. What is knowledge of language?
How is language acquired?
How is language used?
(1a) is related to a precise description of language competence, and it can be said that his early idea, the Standard Theory (and Extended Standard Theory subsequent to it) which aims at revealing the structure of language by means of the phrase structure rules and the transformational rules (Chomsky 1965, 1972), mainly focuses on this problem. Furthermore, Minimalist Programme, the latest framework in Generative Grammar, can be seen as an attempt to elaboration of linguistic knowledge taking the question (1c) into consideration. The reason why two interface levels between competence and performance is introduced is that Chomsky may attempt to deal with the question about the language use while preserving autonomy of syntax. Finally, one interpretation is that Principles-and-Parameters approach in Government and Binding (GB) Theory which is described below is Chomsky's answer to (1b).
2.1.2 Government and Binding Theory
Given that there is a certain kind of language universals, it can be considered that vast amounts of possible options about language are wired in children because language universals is not restricted to the particular language but it is linguistic knowledge which is applied to any types of languages universally, as Chomsky (1965) stated in his Standard Theory. Chomsky (1986) evolved it into the concept of UG in his GB Theory. In this theory, he is concerned himself not with the grammatical rules which generate each linguistic phenomena but with the general principles which generate those rules.
UG is considered to be biologically-equipped linguistic module which is organized as a computational system including rich constraints and fixed operations. In other words, UG consists of various sub-modules each of which has simple principles that become complex when they interact with principles operating within other modules. Chomsky (1986) is concerned in the principles in each module and interaction among them, and claims that each module has a parameter by which a child could choose from limited options based on observation of the input data in order to set his/her language. That is to say, it is parameters that make principles in each module interact; hence "a few changes in parameters yield typologically different language" (Chomsky 1986: 152).
To sum up, children are believed to equip with UG as a mental organ, which contains (a) principles applicable to all languages and (b) parameters that are fixed properties of language whose values vary among languages, and to acquire language in terms of setting parameters which define the relevant principles within UG through observation of input data. In addition, children would be supposed to know in advance the available principles, the manner of their interactions, and parameters with regard to the principles. Chomsky (1986: 146) states that, once the values of the parameters are set, "the system as a whole becomes operative," which means that children have acquired language, and he contends that an initial state of children's language competence has highly rich content, and hence they can acquire language on the same schedule cross-linguistically in spite of the poverty of the stimulus.
2.2 Tomasello's Usage-based Theory
Owing to the poverty of the stimulus, generative grammarians persist on the view that grammar is not acquired by means of learning mechanisms. Especially in the GB Theory, the acquisition of grammar can be accounted in terms of the general principles, which capture linguistic universality, and parameters which enable linguistic variations. It, however, cannot be said that this view reflects the actual conditions of language usage by children. In response to such view, the Usage-based Theory represented by Tomasello (2003) has been remarked.
2.2.1 Prelinguistic Cognitive Development: The Nine-month Revolution
To begin with, The Usage-based Theory on language acquisition has its bases on theories of development in a child suggested by Piaget and Vygotsky. Although their theories differ from one another in some crucial ways, they also have some common features when they are contrasted with Chomskian theory. Firstly, they point out the inseparable relationship between socio-cognitive abilities and language. Secondly, they see language development in relation with development of general cognitive ability. With regard to this point, they stress the importance of prelinguistic development.
Piaget (1974) sees cognitive development in children as a process where they transform a schema, which is mental framework used for recognition of the outer world, through active interaction with environment, and proposes the cognitive development theory in which cognitive development is divided into substages characterized by a new skill children acquire. This idea about phased cognitive development leads to Tomasello's theory which emphasizes prelinguistic development of cognitive abilities. Moreover, for Piaget, although infants cannot differentiate various phenomena or concepts, they actively repeat some behaviors in response to stimuli so that they can assimilate and accommodate their own schemas to the outer world: assimilation is the process of interpretation of the world in terms of schemas while accommodation is transformation of schemas in order to adapt them to the world. This suggestion about a schema and its operations shares similarity with Tomasello's idea about the image-schema formation in language acquisition.
In the Usage-based Theory, intention-reading is the key to acquiring language, not to mention necessity for acquisition of linguistic symbols. According to Tomasello (1999, 2003), infants can typically conceptualize objects or events in the age of 4-5 months, and build up fundamental socio-cognitive abilities around 9-12 months, which form the basis for comprehending and producing language, and then start to develop language around their first birthday. Tomasello (2003) states that, during this age, infants begin to follow adult's gaze, to use adults as social reference points, and to imitate adult's act on objects, all of which are considered to be manifestations of the emergence of infants' ability, called "theory of mind" by Premack and Woodruff (1978: 515), to understand others as intentional agents as well as themselves.  With regard to this, Tomasello (2003: 21) proposes fundamental socio-cognitive abilities as bases for language acquisition: "(1) the joint attentional frame, (2) understanding communicative intentions, and (3) cultural learning in the form of role reversal imitation." The concern here is that, as Tomasello states (ibid.), children's earliest joint attentional skills significantly correlate with their earliest linguistic skills on comprehension and production, which means that linguistic skills is a special type of joint attentional skills.
Tomasello (2003: 22) defines the joint attentional frame as "the common ground â€¦ within which adult-child communication may take place." In detail, it involves the triadic relationship among a child, an adult and objects, and is intentionally defined according to the purpose of communication; hence, objects related to a purpose of communication are considered to be elements within the frame, but this is not the case with objects that are not related to a common purpose even if they are perceived by participants. Within this common ground, children begin to understand communicative intentions which are required for understanding language. Tomasello (2003) puts it this way: children can recognize a sequence of sounds as language only when they know that others are intentional agents and that they generally make sounds with certain intentions. These two abilities such as the establishment of the joint attentional frame and understanding communicative intentions within the frame are indispensable bases not only for fundamental learning of symbols but also for pragmatic skills which operate on various contexts to use symbols appropriately.
Intention-reading enables sociocultural learning which is unique to human beings. Vygotsky (1987, cited in Newman and Holzman 1993) emphasize the importance of imitation in the course of children's cognitive development. It is noteworthy that he uses this term for not a mere mimic but for one in which children know the underlying meaning of behavior. This can be interpreted that children cannot do this type of learning until they project themselves to other's position: hence this leads to Tomasello's idea about imitation learning. Tomasello (2003) observes that infants at the age of 9 months become to imitate adults' actions not only on outside objects but also on themselves, and he calls it as role reversal imitation which prompts emergence of skills of language production. In order to imitate an action on themselves, infants need to reverse their roles with those of adults: that is, "the child must not only substitute herself for the adult as actor â€¦ but also substitute the adult for herself as the target of the intentional act" (Tomasello 2003: 27).
As described above, infants around 9-12 months develop socio-cognitive abilities which serve as the important foundations for language development. Tomasello (1999: 61) calls it as "the nine-month revolution." In addition to these prelinguistic developments, infants show a pattern finding skill which is of vital importance for comprehending symbolic aspects of linguistic communication (Tomasello 2003). However, it should be noted that a pattern-finding skill itself is not sufficient for infants to deal with syntactic structures. The reason for this is that infants in this stage cannot understand symbolic aspects of structures. With regard to this point, Tomasello asserts that they need to find patterns "in the way adults use a particular form communicatively across different usage events" (Tomasello 2003: 31). It can be said that this skill requires understanding communicative intention within the joint attentional frame: that is, when infants engage themselves in the joint attentional frame with an adult and understand the adult's communicative intention, they become to comprehend linguistic symbols produced by an adult. The next section discusses pattern-finding skills required to language development.
2.2.2 Early Language Development: The Schema Formation
Another basis for Tomasello's theory is Cognitive Linguistics which is the branch of cognitive science and it tries to describe and explain various aspects of language both dynamically and statically based on a comprehensive framework which presumes that language is shaped by the way we think and conceptualize the world (Bechtel and Graham 1998). In this viewpoint, there are three major hypotheses about language (Croft and Cruse 2004: 1): "(a) language is not an autonomous cognitive faculty, (b) grammar is conceptualization and (c) knowledge of language emerges from language use." In this view, the linguistic ability is considered to be motivated by cognitive abilities in general. Croft and Cruse (ibid.) states that, in effect, the representation of linguistic knowledge is the same as the representation of other conceptual knowledge, and that the usage of that knowledge is not different from cognitive abilities used non-linguistically. In other words, language is seen not as a simple reflection of the real world which is isolated from our conceptual system but as embodiment of cognition for the outer world which is made by a language user based on bodily experience. Tomasello's theory builds especially on the Usage-Based model and the Construction Grammar in Cognitive Linguistics.
Construction Grammar has been in the spotlight particularly from Fillmore et al. (1988) downward, and has been theoretically developed by Goldberg (1995, 2006), Langacker (1987, 1999, 2008) and Croft (2001). Although a construction is the familiar unit, it is not admitted to be a theoretical unit because generative grammarians take the position that the features of constructions can be explained in terms of interaction with the general principles. By contrast, in Construction Grammar, constructions are considered to have meanings which cannot be reduced to their components. In other words, constructions are seen to be fundamentally symbolic units, that is, they carry meanings by themselves. Take, the following sentence, for example (Goldberg 1995: 199):
Frank dug his way out of the prison.
This sentence means 'Frank escaped from prison,' and this entailment of movement cannot be gained by summing its components up.
a. Frank dug his escape route out of the prison.
He knows his way around town. (ibid.: 199-200)
If way is altered to escape route as in (3a), it does not entail the meaning of movement. Moreover, the sentence as in (3b) shows that way itself does not entail movement. For these observations, Goldberg (1995) points out that the construction [SUBJi [V [POSSi way] OBL]] carries the meaning of movement. 
In the Usage-based model proposed by Langacker (1987), language is determined not as closed system of the rules but as open network of schemas with regard to how conventionalized they are in the real context of the language use. In the process, a schema is firstly extracted from concrete instances in relation to their conventionality, and then this schema is used for generalization with other instances. If there are instances which do not fit the schema, the schema can extend itself dynamically so that it can be applied to new instances. Through this process, a series of construction schemas, which vary both in frequencies of usage and in conventionality, composes a huge and complex network. According to Langacker (2008), a grammatical rule is just a schema that is abstracted from repeatedly used concrete instances in the network. In this way, the Usage-based model makes much of the aspect of the language use in the real context.
Gathercole (2009) studies the process of language acquisition based on the Usage-based model as well. He analyzes utterances of English-speaking children from 1;7-4 years, and reports that the high frequent pattern, no more X, of English expressions such as No more pieces? (2; 0.6) and Is there no more water? (2; 4.17), is observed in children around two years old (Gathercole ibid.: 357). Garthercole illustrates that no more is the fixed parts and X is a slot for some words, and that children can formulate a schema [no more X] when they repeatedly experience concrete expressions including no more. Furthermore, a schema [Y more X] is subsequently extracted from this pattern and other similar types of expressions. The following ungrammatical examples are observed three months later after appearance of this pattern (Gathercole ibid.: 358):
a. I want yes more baguette. (2;3.7)
b. I want yes more cheese! (2; 3.7)
c. Yes more room. (2; 5.14)
According to Gathercole (ibid.), the pattern in [no more X] is firstly entrenched, and then more abstract schema [Y more X] is formulated by means of frequently occurred instances including other degree expressions rather than no. As a result of the entrenchment of this schema, yes more X exemplified in (4) are produced. To put it another way, the process of schematization, where more abstract schema is formulated from pre-formulated schemas, enables children to produce creative expressions and thus further extended instances.
As seen above, In the Usage-based Theory, infants are considered to acquire language step by step: a pattern-finding follows learning concrete linguistic expressions. Once construction schemas are formulated and general features shared among such schema are abstracted, infants extend their schemas so that they creatively produce expressions which they have never heard before. In addition, infants need to engage themselves in certain constructions constantly in order to abstract its pattern. A series of studies by Tomasello show that infants in early age absorb adult's language as a whole, depending on the concrete situation of language usage, and that, by imitating them repeatedly, they extract a certain structure or pattern (Tomasello 1992, 1999, 2003).
Discussion: What is Innate about Language?
From the viewpoint of Developmental Psychology and Cognitive Linguistics, autonomy of language competence held by Generative Grammar is arguable and should be reconsidered. This paper particularly discusses the concept of innateness.
To begin with, in Generative Grammar, it is controversial that recursion in language is believed to be one important feature which enables generative grammarians to presume that human beings are born with language competence. It can hardly be said it is a valid observation because, in practical, human beings do not generate infinite length of sentence. The reason why this fact is ignored in Chomskian theory without much surprise is that linguistic competence and linguistic performance are separated on the idealized theoretical premise. Chomsky (1965) explains that, although human beings have such ability, its original function is prevented by nonlinguistic factors such as memory limitation. This means that the existence of this innate ability is not speculated even if infinite length of sentence is not actually observed. By contrast, from the viewpoint of the Usage-based Theory, there is no need to envisage such indemonstrable ability because language can be appropriately dealt with only in relation to its usage.
With regard to the above point, the most significant argument against Generative Grammar is that UG has not empirically been proved to be purely linguistic knowledge. It has been postulated that UG is genetically built-in knowledge. However, granted that human beings have such kind of innate knowledge, it does not have to be linguistic knowledge. In fact, the idea that a certain kind of innate factor is involved in language acquisition is not totally contradicted among researchers who are against Chomskian theory. It seems that the argument between Chomskian theory and Tomasello's theory is not whether there is a biological basis for language acquisition or not: rather, the difference between them is the idea about what is such biological basis. This point can be further supported by the following claims by Yamanashi (2000) that, if the knowledge in question is one that enables children to acquire grammar, it is just conceivable that it is related to general cognitive ability which is pre-symbolic rather than purely linguistic. In other words, even assuming that human brain has a certain function related to language, it is not necessarily a function peculiar only to language. Instead, there is a possibility that such function of brain is related not only to language but also to perception as well.
Take, an argument over species-specific nature of language, for further instance. Pinker (1994) points out that, although every human culture has its own language, other nonhuman animals have no such languages. It seems to be true that symbolic system we use is specific to human beings. This does not mean, however, the existence of the special innate language faculty isolated from any other cognitive abilities. In fact, Tomasello (1995) sees this argument as an oversimplification of the fact, saying that species specificity of language does not imply specific linguistic genes, and he draws an analogy by Bates (1984) that, although most of all human beings use their hands to eat, this does not mean the existence of a gene for eating with the hands. The subsequent concern here is, then, what produces species-specific universality of language. Tomasello (1995: 137) presents a compelling explanation as follows:
Universality is just as consistent with a view in which human beings all over the world are faced with similar communicative problems and have similar cognitive and physical resources with which to solve them.
As just described, the idea of Generative Grammar seems unreasonable in that evidence supporting it is not empirically proved. By contrast, the Usage-based Theory is considered to be empirically plausible because it does not isolate language from other cognitive abilities and investigates it in a manner consistent with various human activities..
Unlike Generative grammar which defines language from top down by means of closed system of the rules, Tomasello's theory derives from Cognitive Linguistics whose idea is that various language units are schematized and these units are defined according to conventionality and frequency in the real context of the language use. This means that, in the process of language acquisition as well, every language units are acquired in relation to frequency and conventionality in the actual context. Furthermore, it is possibly considered that human beings have no sensorimotor system which is specialized only for language, so sensorimotor inputs such as visual or auditory information requires to be integrated with such memory representations as concepts or images in order to actualize these inputs as meaningful language. In Tomasello's theory, human brain became to enable this sophisticated symbolic operation such as language in the course of evolution, and such operation is on the basis of biological, psychological, and environmental factors. For these points, his theory of language acquisition is viewed from the wide perspective including not only the grammatical elements but also such general cognitive processes as categorization, analogy and mental imagery. Again, as opposed to the view of Generative Grammar that input data is nothing more than a trigger for parameter setting within innate UG, Tomasello rejects innate language module and convincingly contends that children are born with a set with general cognitive abilities and cultural learning skills to acquire language through ontogeny. This approach is much more plausible in that language is viewed as a special part of socio-cognitive activities rather than over-idealized faculty isolated from any other abilities.