Chapter Two: Analysing IL Grammars
“The fundamental problem for a theory of syntax is to characterize the mapping between semantic predicate-argument relationships and surface word order and phrase configurations by which they are expressed.” (Kaplan & Bresnan,1982, p. 174)
Kaplan and Bresnan (1982) suggest that the fundamental problem for syntactic theory is the characterisation of mappings between semantic and syntactic structures. If so, then a fundamental problem for a theory of the acquisition of syntax must be a characterisation of how such a mapping process develops. This chapter discusses past research into the relationship between emergence orders and syntactic processing.
Section 2.1 discusses how the morpheme studies of the 1970s gave rise to the idea of a natural acquisition order, and so to the developmental problem, the need to explain the acquisition orders observed. It then outlines problems inherent in the use of an eclectic selection of morphs to represent an IL grammar and the move to viewing learner language as Interlanguage: a series of transitional but comprehensive grammatical systems in their own right. It also discusses the ‘comparative fallacy’ (Bley-Vroman, 1983) , the mistaken idea that similar phonetic sequences in learner and target languages necessarily serve similar functions and represent similar underlying structures and processes.
Section 2.2 discusses subsequent moves to relate language-specific emergence orders to universal processes. If learner languages are treated as the product of universal processes of language production, then analysis of IL syntax can proceed on the basis of IL data, without reference to the TL. This significantly reduces the risk of being influenced by any comparative fallacy.
Theoretical perspectives on universal linguistic processes fall into three main schools: one views language acquisition as an instance of general learning processes, another as a process specific to language, the third sees it as a process of transition from universal non-linguistic to language-specific processing systems. The focus of this thesis is on the latter two types. This section introduces theories which exemplify the transitional viewpoint, with particular emphasis on Processability Theory (Pienemann, 1980, 1984, 1998). Section 2.2.1 describes the six developmental stages that Pienemann suggests arise in second language acquisition (SLA), as a consequence of gradual automatisation of syntactic processing. This allows increasing independence from general cognitive processes related to thematic structure and perceptual saliency. Section 2.2.2 discusses the advantages of this theory over its predecessor, a strategy-based approach proposed by Clahsen, Meisel and Pienemann (1983), where constraints are gradually shed to reveal an underlying grammar, and over an alternative three – stage development proposed by Perdue (1993), where internal conflicts within a basic universal syntactic system provide the pressure to develop more grammatical structures.
Section 2.3 introduces theories that exemplify the nativist or purely linguistic viewpoint, including the minimalist programme (Chomsky, 1995, 1999, 2000). First it argues that parameter-setting addresses the logical problem of why acquisition occurs, but not the developmental problem of why it follows the course it does. Then it discusses two views of acquisition as a gradual extension of phrase structure from lexical through various functional levels, and introduces the minimalist programme (Chomsky, 1995, 1999, 2000). The discussion of minimalist theory reviews the generative structure-building operations of minimalism, from the construction of lexical items to the determination of word order, then reviews economy conditions held to determine choices between alternative derivations for the same surface string. Herschensohn’s constructionism is discussed briefly (Herschensohn, 2000), but shown to be inconsistent with the version of feature theory proposed in Chomsky’s “Derivation by Phase” (Chomsky, 1999).
Because it makes explicit links between derivational efficiency and grammaticality, minimalist theory provides the basis for an account of acquisition order in terms of relative economies. This is outlined on page 42f.
Section 2.4 outlines a new minimalist theory of acquisition which views minimalist operations as contributors to processing demands.
MORPHEME STUDIES AND NATURAL ACQUISITION ORDERS
As early as 1972, Selinker had proposed that learner languages could and should be treated as independent objects of linguistic analysis. His proposal that learners pass through a series of interlanguages as they move from an initial state towards a target system, opened the door to attempts to explain IL development in terms of general, and possibly universal processes of language production. Since 1972, the analysis of learner language has progressed from descriptions of ILs viewed as transient ideolects to descriptions of IL systems viewed as instantiations of universal grammar.
The earliest characterisations of acquisition orders were couched in terms of grammatical functors or morphemes. In 1973, Brown observed a marked similarity in the order in which three children acquired certain grammatical morphemes as part of their L1, English (Brown, 1973). Soon after, Dulay and Burt demonstrated common but different patterns in the acquisition of English as a second language (ESL) (Dulay & Burt, 1974) and it was subsequently shown that this order is relatively unaffected by the linguistic environment (Fathman, 1978; Makino 1979); the learners’ L1s (de Villiers & de Villiers, 1973; Dulay & Burt, 1974; Larsen-Freeman, 1975); or the measure of proficiency employed (Pica, 1983).
Though not all these morpheme studies found precisely the same order, Krashen (1975, 1976, 1977) argued that the morphs could be allocated to stages whose order does remain constant, and on this basis he proposed the existence of a natural order of acquisition for English. He also presented evidence that this natural order applies equally to instructed and naturalistic learners. In fact, he claimed that teaching and conscious study are quite unable to affect the natural acquisition order. A controlled experiment conducted by Pienemann later confirmed that even intensive instruction could not affect the order in which morphs emerge in spontaneous speech, though it can improve declarative knowledge that can be employed in more slowly paced tasks, like writing, or grammatical judgements (Pienemann, 1987).
Similar studies have found consistent emergence orders for other languages, including Spanish (van Naerssen, 1980) German, (Clahsen, 1980) and Japanese (Huter, 1996, 1998). Other studies have also revealed a stable order in the emergence of syntactic structures within a given pragmatic domain. Spanish, Chinese, and Norwegian speaking children acquiring English all follow a similar sequence in the development of interrogative forms (Adams, 1978; Cazden et al., 1975, cited in Larsen-Freeman & Long, 1991; Huang, 1970; Ravem, 1974; Wagner-Gough, 1975), and pre-verbal negation occurs consistently in ESL, regardless of the form of negation in the learners’ L1s (Hyltenstam, 1977; Schumann, 1979; Stauble, 1984).
The observation of these similarities gave rise to a growing conviction that there is a profound cross-linguistic homogeneity of language production and acquisition processes. However the language-specific and apparently arbitrary nature of the morphemes studied and the stages they fell into made it difficult to grasp just what those universal underlying processes might be, and how they might relate to the form or function of the selected morphs.
During the ‘80s and ‘90s, a theoretical framework was sought that might unify and explain this disparate data. Attention turned to complete syntactic structures, and whole interlanguage systems rather than to isolated morphs, and a continuum of views emerged, ranging from, at one extreme, those who saw language production as arising out of general cognitive processes (Bever, 1970; MacWhinney & Anderson,1986; Rumelhart & Mclelland 1986; N. Ellis, 1994), and at the other extreme, those who view human languages as products of innate language-specific cognitive processes (Chomsky 1986, 1995, 1999, 2000; Bresnan, 1982; Haegemann, 1995,1996; Radford, 1990; Vainikka, 1993; Vainikka & Young-Scholten, 1994, 1995, 1996a,b, 1998; White 1990, 1992, 1996).
The notion of language-specific processes has two different, but equally important senses. First productive processes may be language-specific in the sense of applying only to language, and not, for example, to vision, or mathematical deduction etc; secondly, processes may be language-specific in the sense of applying only to language A, and not to language B. The idea of processes that are language-specific in the first sense makes it possible to explain phenomena that are language-specific in the second sense in terms of constraints on and freedom within the language-processing system generally.
Between these two extremes is a view I will label ‘transitional’. This holds that mature languages are the product of specifically linguistic processes, but in immature languages, or in times of cognitive stress, words can be manipulated by more general cognitive processes (Clahsen, 1980, 1984; Clahsen & Muysken 1986; Pienemann, 1984, 1989, 1998c; Perdue, 1993). The cognitive processes seen as particularly relevant relate to semantic and pragmatic knowledge: in particular, the application of knowledge about the world, or the immediate context, to the interpretation of words, alone or in sequence. Semantic knowledge includes understanding thematic relations – the relationships between entities in an event; pragmatic knowledge includes an understanding of pre-supposition – inferences about another’s state of knowledge at the time of speech.
I will be concerned only with the transitional and language-specific views of language acquisition. In the next section I consider three different proposals about the nature of the transition of language systems from a pre-syntactic to a syntactic state. The first two prove to be problematic because they do not adequately address the nature of the developing syntax; nor are they readily applicable to the analysis of nominal structures. The third, Processability theory, incorporates a complete theory of generative grammar. This makes it applicable in principle to the nominal structures of Mandarin, the focus of this study. However, as will be shown, that applicability is significantly limited by the theory’s treatment of phrasal structures.
It is well-established that semantic and pragmatic knowledge have an important relationship with word order. Native-speaker choices of word order reflect the relative agentivity, topicality, salience, and familiarity of the referents within an utterance (Dubois, 1987; Lambrecht, 1987; La Polla, 1995).
Pragmatic knowledge also forms part of the information content expressed in many languages by nominal morpho-syntax such as the choice of articles, determiners, or inflections to indicate definiteness, specificity, number etc. Because of this relationship between word order, semantics and pragmatics, alternative syntactic structures can be understood as a means by which to arrive at a word order that is desirable for semantic or pragmatic reasons. However, since word order serves multiple functions, indicating thematic relations – who does what to whom- as well as pragmatic information-what is known to whom – and discourse structure – what is being talked about, each variation in word order motivated by factors in one domain, threatens the preservation of meaning in the other domains. In this context, morpho-syntactic markers can be seen as a means of avoiding loss of information: they express information that, for whatever reason, cannot be expressed by word order alone.
A central idea within the transitional models of language acquisition is that pragmatic or thematic information can be mapped to word order, before it can be expressed by morphological means. Thus, language acquisition is a process of transition from a primitive reliance on semantic and pragmatic linear ordering strategies, through a stage where abstract hierarchical linguistic units are discovered or constructed, to a mature stage where those units are manipulated, primarily by processes that are language-specific, in both senses.
The multi-dimensional model: shedding strategies to reveal syntax
One of the earliest transitional accounts of SLA was the multi-dimensional model (MDM) (Clahsen, 1980, 1984, 1986; Clahsen, Meisel & Pienemann, 1983; Clahsen & Muysken, 1986; Meisel, Clahsen & Pienemann, 1981; Pienemann, 1985). This integrated psychological research on memory and on processing under stress into a model proposed to account for patterns in German SLA. According to this model, learners gradually bring into play, and then abandon, three processing strategies, the ‘Canonical Order Strategy’ (COS), the ‘Initialisation- Finalisation Strategy’ (IFS) and the ‘Subordinate Clause Strategy’ (SCS). COS is based on work by Bever (1970) who demonstrated that adults processing under stress tend to interpret NVN strings as agent-action-patient, and ascribed this to universal cognitive processes. COS constrains the order of the main predicate and its arguments to a stable canonical order, initially thought to be SVO. IFS is based on primacy and recency effects (Murdock, 1962); items at the end-points of a linear sequence are more memorable than intermediate items, and the IFS excludes movement, or the addition of items to central positions in an SVO string. The last strategy, SCS, is an extension of this and also based on direct observations of SLA, especially of German, where word order of Subordinate clauses (SCs) is different from that in main clauses. The SCS prohibits permutations in subordinate clauses, reflecting the fact that SC word order is mastered late in German SLA (GSL).
One important contribution of this model was the move towards a formal definition of a “stage”. Stages in development were defined in terms of the combination of strategies in use at a given time. To count as a ‘theoretically interesting’ stage, Meisel, Clahsen and Pienemann (1981) required a set of structures to be a) obligatory, and b) ordered with respect to other sets. So the earliest stage is one where word order is fixed by COS, and the next stage is one where IFS allows the use of initial or final adverbs in addition to the COS string. Larsen-Freeman and Long (1991) point out that prior to this a stage was traditionally identified in terms of the most frequent structures used, not in terms of a definable class of structure.
Also significant was the idea that strategies function to extend or limit the psychological difficulty of the structures that learners produce. In Pienemann’s words, “the psychological complexity of a structure is dependent on the degree of re-ordering and rearrangement of linguistic material involved in the process of mapping underlying semantics onto surface forms” (Pienemann, 1998c, p. 46). When COS is abandoned, in the third stage, core items, S, V, and O can be moved to a terminal position, which accounts for the emergence of Subject-auxiliary inversion in ESL questions (as reported in Larsen-Freeman and Long, 1991), and verb-final order (SEP) in GSL. However SCS still prevents reordering in subordinate clauses. Abandonment of IFS then allows string-internal movement, contributing to the V2 effect in GSL, and Do-support in English Wh-questions. Finally, abandonment of SCS, allows the verb to take clause-final position in German subordinate clauses.
PROBLEMS WITH THE STRATEGIES APPROACH
Weaknesses in this account are immediately apparent. Larson-Freeman and Long (1991) raise three questions of falsifiability. Firstly, formulaic or “chunked” data can easily mislead in the assessment of productive capacity; secondly, there are no clear grounds on which to identify structures which are variational, and hence properly excluded from the model1, and thirdly, it is not clear to what extent structures from different developmental levels can be permitted to co-occur without invalidating the model.
Pienemann (1998c) raises concerns about the structure of the language processing system represented by the MDM, pointing out that the strategies do not generate linguistic structures; they only constrain them. Given this, “the specification –COS, +IFS, +SCS needs to be complemented by an explicit grammatical rule or system of rules” (Pienemann, 1998c, p. 49). In other words, the acquisition of the grammar that supersedes or underlies production still remains to be explained.
Pienemann also maintains that early lexical items “cannot be shown to be indexed to particular syntactic categories” (1998c, p. 50). If so, then learners cannot initially implement strategies like COS, that refer to N or V. Likewise, learners cannot identify linguistic units like a clause, until they have acquired knowledge of language-specific collocational possibilities, so strategies that refer to clauses, like SCS, cannot initially constrain production. Crucially, Pienemann (1998c) also rejects the centrality of syntactic movement, claiming that psychological experiments have shown passive sentences, theoretically involving movement, to be no slower to process than active sentences, theoretically more basic (Altmann, 1990; Horrocks, 1987; Levelt 1989) 2.
Finally, Pienemann echoes White’s (1990) criticism that the strategies applied to production are derived from studies of comprehension. As comprehension routinely anticipates production, constraints on the latter must relate to processes other than those employed for interpretation.
More could be said about the relationship between the proposed psychological constraints and productive capacity. For example, the IFS refers to spans of only three items: “in underlying sequences [X Y Z] permutations are blocked which move X between Y and Z]” (Pienemann, 1998c, p. 46). Yet, the psychological salience of terminal positions is based on recency and primacy effects for lists of 7 2 random items (Murdock, 1962). The items may themselves be ‘chunks’, complex items that for Structures whose emergence order varies considerably are properly excluded because Processability theory aims to understand the factors that constrain emergence order, where it is constrained. Pienemann suggests that some variation is inevitable, because any structure can emerge at a given stage as long as it does not exceed the processing capacities that define that stage. The nature of this variation cannot be explained in terms of processing demands. It requires an alternative account.
Though in later transformational analyses, active sentences also involve movement of the Subject out of VP whatever reason are readily memorable in isolation (like important dates in history, familiar birthdays etc). Thus, an ordered set of just three words or phrases should fall well below the threshold at which recency and primacy effects arise. Thus the IFS cannot be said to follow directly from the effects discovered in psychological research.
Moreover, it is unclear how a violation of the IFS could even arise in a string of just three items. Given an underlying string [XYZ], all six permutations of X, Y, and Z should be legitimate in the– COS, +IFS stage since all can be derived by moving items only to terminal positions, as indicated by the following, where tx represents the initial position of X etc.: [tx Y Z X] [X tY Z Y] [Y X tY Z ] [Z X Y tz] [Z txY tz X]. As the same orders can also be derived by prohibited movements, it is impossible in principle to distinguish licit strings of three items from illicit ones, and impossible for the learner to establish, on the basis of input, which order is basic, and which derived.
Though it is plausible to think that general psychological constraints might limit the number of linguistic units that can be combined in early processing, this entails the existence of linguistic units to begin with.
CHAPTER ONE: INTRODUCTION
1.0 The Research Context
1.1 The Research Approach
1.2 Establishing a natural emergence order
1.3 Ascertaining Processing demands
1.4 Correlating processing demands and emergence orders
CHAPTER TWO: ANALYSING IL GRAMMARS
2.1 Morpheme studies and natural acquisition orders
2.2 Transitional models
2.3 Nativist /structural models
2.4 A new Minimalist account of acquisition order
CHAPTER THREE: MANDARIN SLA
3.1 Morpheme studies in Mandarin
3.2 A Processability account
3.3 Some unresolved issues
CHAPTER FOUR: METHODOLOGY
4.1 Research Design
4.3 Data Collection
CHAPTER FIVE: A NATURAL EMERGENCE ORDER
5.1 The Corpus
5.2 Lexical categories in the ILs
5.3 The emergence of constituents in three ILs
5.4 An Average acquisition order
5.5 Comparison to Zhangs results
CHAPTER SIX: PROCESSING PHRASES
6. 0 Introduction
6.1 Feature Unification
6.2 Abstract vs conceptual information
6.3 A revised framework: from local to long-distance relations
6.4 Conclusions: Outside-In development
CHAPTER SEVEN: STRUCTURES AND PROCESSING DEMANDS
7.1 Determining Lexical F-structure
7.2 Deriving IL Constituent structures
7.3 Processing demands of IL nominal structures
7.4 Processing demands and emergence orders
CHAPTER EIGHT: EXPLAINING EMERGENCE ORDERS
8.1 Processing demands and emergence times
8.2 The Minimalist account and inter-learner variability
8.3 The Emergence of Grammatical Functions
CHAPTER NINE: CONCLUSIONS
9.1 The natural order
9.2 Processing demands correlate with emergence times
9.3 Expanding functional capacity but no clear stages
9.4 Implicit Processes that are not task-specific?
9.5 Which theory is better?
GET THE COMPLETE PROJECT