Chicago under revision

Morphological Typology and SLA: Inflectional processes in L2 Russian vs. L2 Kazakh
Greg Thomson & Bayan Uataeva
Al Farabi Kazakh National University
Presented at the Conference on Form-Meaning Connections in Second Language Acquisition
University of Illinois at Chicago
February 24, 2002
Abstract
Recent research into the acquisition of inflection in L2 Russian (Thomson, 2000), as reflected in sensitivity to inflectional anomalies in spoken texts, pointed to an extremely lengthy, and probably typically incomplete pattern of acquisition. This was attributed to the many-many nature of form-function relationships in Russian, a language of the fusional morphological type. This research is extended here to an agglutinative language, Kazak, where the form-function relationships are farther down the continuum in the direction of one-one form-function relationships. Two central predictions were that 1) in terms of the pace of acquisition, L2 Kazak would have an advantage over L2 Russian, and 2) in terms of the pattern of acquisition, there would be a detectable similarity between L2 Kazak and L2 Russian. The first prediction was not supported, while the second was supported. Implications for our general understanding of the nature of the acquisition of inflectional form-function connections in second languages are explored.

Inflectional morphosyntax is a domain of language in which second language learners may tend not to become nativelike (Selinker, 1972; Schumann, 1978; Johnson & Newport, 1989), with some varieties of inflection possibly presenting more resistance to acquisition than others. It has been argued that both in English (Zobl & Liceras, 1994), and German (Vainikka & Young-Scholten, 1998), whereas child L1 learners acquire inflectional morphosyntax more readily than they acquire free function words, the opposite is true for adult language learners. The domain of inflectional morphology thus raises interesting theoretical and practical questions, answers to which might well expand our currently limited understanding of the nature of inflectional processing in general, along with our understanding of the factors influencing the relative learnability of various features of second languages.

Interest in second language grammatical morphology (including inflectional morphology and function words) was especially stimulated by the oft-cited Morpheme Studies (e.g., Bailey, Madden & Krashen, 1974; Dulay & Burt, 1973, 1974; Larsen Freeman, 1975; Krashen, et al., 1976; see Goldschneider & DeKeyser, 2001, for a recent meta-analysis of twelve of these studies). Some of the interest evoked was related to the belief that a constant order of L2 acquisition, independent of the learners' mother tongues, ages and learning contexts, counts as evidence for a built-in mental mechanism responsible for SLA (Dulay, Burt & Krashen, 1982).
More recently, attempts have been made to provide formal theoretical bases for the relative ease versus difficulty of acquisition of various grammatical morphemes in SLA (Eubank, 1994; Lardiere, 1998; Pienemann, 1999; Zobl & Liceras, 1994; Vainikka, & Young-Scholten, 1998). There have also been more informal proposals, such as Andersen's (Andersen & Shirai, 1994) operating principles (following Slobin's [1973] proposals for first language acquisition), Goldschneider & DeKeyser's (2001) proposed factors related to the English L2 acquisition order of grammatical morphemes, and VanPatten's (1996, 2000) input processing principles.
Generative approaches to L2 inflectional morphology (e.g., Epstein et al. 1996; Lardiere 1998, 2000; Prévost & White, 2000; Vainikka & Young-Scholten, 1998; Zobl & Liceras, 1994) sometimes appear to accept as a axiomatic the hypothesized relationship between abstract inflectional elements and surface syntactic configurations (Pollock, 1989; Chomsky, 1995), in spite of the fact that evidence for such a relationship in L2s is largely absent. For example, Prévost & White (2000) describe the speech Lardiere's (1998) subject, Patty, as demonstrating "complete knowledge of the fact that English verbs do not raise". The fact that this "knowledge" exists in the absence of nativelike inflectional expression of tense and agreement is taken as evidence that the inflectional elements are nevertheless present and properly parameterized. Lardiere (1998) herself suggests that although major aspects of the syntax depend on the acquisition of the inflectional categories, "the productive acquisition of the spell-outs may in fact be largely irrelevant to the syntax" (p. 20). From this perspective, the question of why certain inflectional form-function relationships are more readily acquired than others would appear to become a question of why the mappings of some categories onto their phonetic forms (say, plural -s, or possessive -s) are more readily acquired than the mappings of other categories (say third person agreement -s) onto their phonetic forms.
Pienemman (1999) proposes an account of the relative difficulty of acquisition of L2 inflectional morphosyntax in terms of the level of phrase structure at which each inflectional category functions. On this account, word-level inflection (e.g., number) ought to be acquired more readily than phrase-level inflection (e.g., noun-adjective agreement), which in turn should be acquired more readily than subject-verb agreement, which in turn ought to be acquired more readily than grammatical morphology involving interclausal relationships (mood in complement clauses, perhaps). These proposals are supported with evidence from L2 English and German.¹
Goldschneider & DeKeyser (2001) found that the best predictors of "accuracy scores" in L2 English speech were "grammatical status" (understood in terms of the contrast between free versus bound status and between lexical versus functional status—r = .68) and phonological salience (r = .63). Of course, as they note, such properties are not entirely independent of one another. The amount of phonetic substance is generally related to the grammatical status (free function words having more phonetic substance than affixes), and also, it has been argued, to the degree of semantic concreteness versus abstractness (Bybee, Pagliuca & Perkins, 1994).
VanPatten (1990, 1994, 2000; VanPatten & Cadierno, 1993; VanPatten & Sanz, 1995) has proposed a notion of communicative value as a predictor of ease of acquisition of grammatical morphology. Communicative value is taken to be a function of redundancy and inherent semantic value. Grammatical morphemes with low redundancy and high semantic value should be more readily acquired than those with high redundancy and low semantic value. Associated with this proposal is a set of principles governing which aspects of grammatical form learners are relatively less or more likely to process: content words first; lexical items in preference over grammatical items as a means of conveying some specific information; and more meaningful morphology over less meaningful morphology. Forms that are less likely to be processed in general can become more readily processable once other aspects of the input come to be processed with increasing automaticity (decreasing attentional demands).
In this paper, we want to explore the influence of a particular property of inflectional form-function relationships on their ease versus difficulty of acquisition: the extent to which an inflectional system approximates or deviates from a one-to-one relation between form and function. While testing the prediction that this property affects ease of acquisition, we also want to look for common patterns of acquisition across languages that differ in terms of this property, since acquisition patterns that are common across languages that differ in their morphological typology may reflect universals of second language acquisition. In spite of the obvious importance of inflectional morphology to our overall understanding of SLA, to date there has been little research comparing the acquisition of inflectional morphology in L2s of contrasting morphological types (in particular comparing fusional languages with agglutinative languages).
Goldschneider & DeKeyser (2001) suggest that inflectional homonymy and allomorphy (and, citing Peters, 1995, fusion) lower the salience (in a broad sense of that term) of inflectional forms. Such features also violate Andersen's (1994) one-to-one principle, according to which learners attempt to assign a single meaning to a single form. In tasks much simpler than language processing (i.e., pushing a button in response to the appearance of particular visual arrays), Schiffren and Schneider (1977; Schneider & Schiffren, 1977) found that the development of automatic responses was hindered when there was not a “consistent mapping” of stimulus sets to responses. Thus even at a very general level of learning principles, the existence of a large number of many-many relations in an inflectional system might interfere with learning.
With this in mind, Thomson (2000) examined aspects of the acquisition of certain Russian inflectional categories. The general picture that emerged in that study was one of extremely prolonged acquisition across a range of inflectional phenomena, with some apparent differences in the rates of acquisition of different functions of individual inflectional forms. It was suggested that the apparently slow rate of acquisition (probably rarely leading to near-native functioning) resulted, at least in part, from the complexity of the overall system in terms of the number and degree of many-many relations of form to function, and in part from the essential nature of inflectional form-function relationships in general (being associated with brief events in the comprehension process, rather than with stable components of conceptual form as is the case with content words). It was further speculated that if the degree of deviation from the one-one principle is the major cause of slow acquisition, then L2 acquisition of comparable systems of inflectional categories should occur more readily in languages where the relationships are farther along the continuum toward one-to-one relationships.
In terms of classical morphological typology, the relevant difference is that between fusional languages (also, somewhat confusingly, referred to as inflectional languages), such as Slavic languages, and agglutinative languages, such as Turkic languages. In fusional languages multiple categories can be combined within a single inflectional expression (what Matthews, 1991 called cumulative exponence) and inflectional expression may be subject to a large amount of arbitrary allomorphy and homonymy. By contrast, in agglutinative languages there tends to be one affix (with phonologically regular variants) for each grammatical meaning, and allomorphy involves general phonological processes. This typological difference is illustrated by the following parallel examples from Russian and Kazak: