Do L2 Grammars Go Beyond the L1 and L2 Input?

Under normal circumstances, acquiring a first language, compared to acquiring a second or third language, seems to come naturally and successfully. This natural process of first language acquisition (henceforth L1A) appears to be illogical. This is because what first language learners (henceforth L1ers) attain (i.e. a grammar of adult native speakers of a given language) goes beyond the positive language stimuli to which they are exposed, and the positive language stimuli seem to be erratic and chaotic. In other words, the input underdetermines the mature grammar of a given language that they attain. The puzzle is how they manage to acquire the mature grammar with comparatively insufficient and impoverished input, which is the so-called logical problem in L1A. To account for the logical problem, an innate mechanism that constrains what is possible in human languages is suggested to guide L1A. This built-in mechanism reduces the burden of the acquisition task for L1ers, and L1ers thus can successfully and rapidly attain an adult grammar of a language with deficient input (poverty of the stimulus) (Cook 1991; Gregg 1996; White 1989). In other words, this mechanism, called Universal Grammar (henceforth UG), makes L1A possible. In the 1980s, UG was characterized as the principles and parameters (e.g. Chomsky 1981, 1986); that is, UG consists of a set of principles, which are universal and invariant, and parameters with binary values, which account for cross-linguistic variations. Later, in the 1990s, the Minimalist program was developed. Accordingly, UG is argued to be composed of the lexicon, the two interpretive linguistic levels (i.e. articulatory-perceptual (PF) and conceptual-intentional (LF)), and the computational operations (i.e. Select, Merge and Move) (Chomsky 1995). A more recent Minimalism proposal (Chomsky 2001) reduces the computation system to Agree and Move. Nevertheless, Minimalism assumes that the computational operations are cross-linguistically invariant and differences between human languages reside in the morpholexicon (i.e. morphological components and lexical items). The morpholexical information is further elucidated by grammatical categories/features (i.e. functional and lexical) and represented structurally. No matter how UG is formalized, what is more important in terms of L1A is that L1ers are arguably born with the universal built-in mechanism at the initial stage S0 of their L1A. Then with exposure to the primary linguistic data of a language, L1ers only have to set proper values of permissible variations (i.e. parameters or grammatical categories and features) made available by UG as invariant principles or the computational operations do not have to be learned. Chomsky regards this language acquisition process as “a function mapping experience (primary linguistic data, PLD) to a language” (Chomsky 1995: p.169). It can be conceptualized as “proceeding from a genetically determined initial state S0 to through a sequence of stages S1, S2..., finally arriving at a steady state Ss ,” (Chomsky 1980: p.37). With the genetically built-in UG, across the board L1ers attain the mature grammar of a given language without much difficulty. As discussed above, UG is believed to constitute the initial state of L1A and makes L1A possible. However, as far as second language acquisition (henceforth L2A) is concerned, whether UG works in the same fashion in L2A as in L1A is disputable. Unlike L1A, in which L1ers start out with a clean plate along with UG, what constructs the initial state in L2A gives rise to disputation (White 2000, and many others). It is because not only the availability of UG but also a fully developed adult grammar (i.e. L1 grammar) needs to be taken into consideration. Given the assumption of Minimalism that cross-linguistic variations are rooted in morpholexicon and that morphosyntactic properties are now formalized in terms of grammatical categories (lexical and functional) and their associated features, the