Chicago under revision
Download 1.36 Mb.
|
Morphological Typology and SLA Inflectio
Note, for example, that there is no single form in the Russian examples, for any particular case, or for plural marking, since these grammatical meanings are fused in the forms for the various combinations of number, case and gender. By contrast, in Kazak there is a simple plural marker followed by a simple case marker, and the form of these markers is constant, apart from regular phonological changes. The hypothesis that L2 acquisition of inflection in an agglutinative language such as Kazak would occur more readily than in a fusional language such as Russian is particularly plausible in view of other varieties of evidence. Specifically, Slobin has presented evidence that for child L1 learners, inflectional morphology is more readily acquired in agglutinative languages than inflectional languages and that inflection in agglutinative languages is more resistant to loss in aphasia than in inflectional languages (Slobin, 1985, 1991). These findings, if they are indeed general in nature, point to some representational and/or processing advantage for agglutinative morphology over fusional morphology, and we might hope that this would also benefit second language learners. Before discussing the experimental findings related to Kazak as a second language, and comparing those findings to Thomson's (2000) findings related to Russian as a second language, it is worth taking a moment to clarify the logic of the experiments reported in Thomson (2000) and here. In investigating whether or not form-function relationships are being acquired by L2 users, it is reasonable to begin by examining listening comprehension, even though by far the more common approach has been to examine spoken production data (e.g., the studies surveyed in Goldschneider & DeKeyser, 2001). Drawbacks of relying on spoken production data have long been noted (Schachter, 1974; Larsen-Freeman, 1975). The fact that consistent acquisition orders have been detected by researchers relying on production data argues for the validity of that approach. However, production data leave many puzzles. In a case such as Lardiere's (1998) Patty, where no nonnativelike instances of pronominal case were detected in a substantial corpus, the most plausible conclusion might be that the system of pronominal case has been acquired, or at least mechanisms have developed that produce totally native-appearing output. Unforturnately, many grammatical phenomena, including inflectional phenomena, have a native-like appearance in only a portion of obligatory contexts in the speech of a particular L2 user. Lardiere's Patty, for example, produced past tense inflections in relevant contexts only 34% of the time. What exactly is behind the native-like appearance of productions in those particular contexts? Are those native-like productions evidence that native-like form-function connections have been (perhaps weakly) established? Does a 34% production rate in the "fossilized steady state" have the same theoretical significance as a 34% production rate that is on the rise? Production data, in fact, may cloud the issue of what is actually taking place in the developing, or fossilized, L2 processor, since they can reflect a large variety of factors. The following are plausible sources native-appearing production in any given instance: 1) The learner only knows one inflectional form of a lexeme, and uses it in all contexts, and in the present instance this gives a native-sounding result (Pica, 1984). 2) Fortuitous "hits" on native-appearing forms based on random activation of competeing forms 3) Use of stored utterances a) Formulaic utterances may have been memorized without regard for their internal structure and may still be retrieved in that form (routines in the sense of Krashen & Scarcella, 1978). b) The L2 user at times may perseverate in using a form which the native interlocutor has just used (stored in short-term memory). At times this may result in a contextually appropriate form (and at times not--see examples in Thomson, 2000). c) The L2 user may have heard one or more native speakers utter the specific phrase or sentence at some time or times in the past, and it is still retrievable (not in the unanalyzed, formulaic manner referred to under point 3 a) . d) The L2 user may have been assisted at some time by a native interlocutor in formulating the phrase or sentence, and then stored the expression, or stored the strategy for producing it. e) On some previous occasion the L2 user may have consciously designed the specific phrase or sentence based on so-called rule-learning (see point 4, below), and then stored either the expression or the strategy for producing it. f) The L2 user may have produced the phrase or sentence by genuine, perhaps still relatively inefficient, linguistic production mechanisms (see 5 below), and then stored either the results, or the production process, in memory. 4) L2 metalinguistic strategies4 a) The L2 user may creatively apply a (relatively) descriptively adequate 'rule' in carefully planning a new utterance. b) The L2 user may have achieved fortuitous success via creative use of a descriptively inadequate 'rule of thumb' (e.g., "Use the when there is only one individual in view."). c) The L2 user may have used a translation strategy that fortuitously produced native-sounding output. d) The L2 user may have rapidly analogized from a highly familiar form (e.g. washes) to an unfamiliar form (e.g., crushes). e) The L2 user may have developed relatively rapid strategies resulting from frequent application of a), b) and/or c) (as suggested by Truscott, 1998). 5) Linguistic L2 production mechanisms a) L1 production mechanisms employed in L2 speech production may have fortuitously given a native-appearing result. b) Pre-grammatical, word-specific patterns of the sort observed in early child language (Tomasello, 2000) may have been used. c) Finally, the nativelikeness of the production in the given instance may in whole or in part result from normal production mechanisms which have developed for the L2, and which produce the nativelike forms for nativelike reasons. With the exception of 4c, productions deriving from these sources might contain native appearing forms in the absence of the corresponding nativelike functions (from the speaker's standpoint, that is—for the native listener, they would function normally, of course). This is not to deny that any of these potential sources of a particular native-appearing production might be of practical value for an L2 user desiring to appear as nativelike as possible. It should nevertheless be of theoretical interest to sort out the relative contributions of such factors (and no doubt others) in the spoken production of L2 users. It should be of considerable interest to learn the extent to which normal speech production mechanisms (see, e.g., Bock & Levelt, 1994; Levelt et al.1999) underlie the "suppliance in obligatory contexts" of particular forms. Once that were known, the relative contributions of other factors, and the way those factors affect one another, might be better sorted out. It seems reasonable that stored utterances (or stored production strategies) are a major contributor to native-appearing L2 speech (in line with findings on task-repetition; see Gass, et al., 1999; Bygate, 1996; see also Nunan, 1998). The existence of L2 metalinguistic strategies (applying "rules" in sentence planning) is also widely recognized (see DeKeyser 1998, Schmidt, 1992; Towel & Hawkings, 1994; Nunan, 1998) as is the idea that such strategies would undergo the same sort of "streamlining" and speeding up, as is the case with other cognitive problem solving strategies (Anderson, 1982, 1983; DeKeyser, 1998). All the same, there might well be some practical value in the partial development of the more familiar, L1-style speech production mechanisms for L2 use, in which case other these other, nonlinguistic factors might even be viewed as supplementary and compensatory rather than central. Normal speech production mechanisms (see e.g., Garrett, 1980; Levelt, 1989; Bock & Levelt, 1994; Levelt, et al., 1999) may be specially adapted to the demands of high-speed conversion of messages into articulation, thus placing a limit on the possible role of non-linguistic mechanisms in L2 speech.. The number of potentially confounded factors in tabulations of percentages of suppliance in obligatory contexts should encourage us to look for clearer evidence of the existence of bona fide form-function connections. A reasonable place to look is within the comprehension system. In fact, a strong argument can be made that the functions of linguistic forms are first of all comprehension functions (Thomson, 2000). A comprehension system which includes those L1-style form-function relations could be in a position to fulfill what Postma (2000) calls the tuning function of speech self-monitoring. That is, when the production system produces anomalous utterances (that is, utterances which clash with the expectations and needs of the comprehension system), the production system can itself be modified accordingly. As Stemberger (1998) pointed out, if we assume that the production system must be trained, the obvious candidate for the trainer role is the comprehension system. The comprehension system can learn to associate various aspects of form with various aspect of meaning, and with events within the meaning construction process (Thomson, 2000). The production system then merely needs to learn to supply forms according to the expectations of the comprehension system. From such a perspective (granted, a relatively recent one), it would be problematic to posit nativelike production mechanisms in the absence of nativelike comprehension mechanisms. Thus studying the effects of grammatical phenomena such as inflectional morphology within the comprehension process can provide important evidence regarding the role of genuinely linguistic mechanisms in L2 production and acquisition. If Patty's comprehension system (Lardiere, 1998) were to show no awareness of tense inflection, then 4c above could probably be ruled out as an explanation of the 34% of cases where native-appearing inflected forms occur. That is, if the comprehension system has no sensitivity to a particular form-function relationship, then the production system cannot have learned to provide the necessary forms in compliance with the needs and expectations of the comprehension system. The assumptions of this approach to L2 acquisition, and language acquisition in general, stand in contrast to other common assumptions. In particular, language acquisition is often viewed as the acquisition of "knowledge" regarding which utterances are and are not well-formed according to the requirements of an internalized grammar. To use a classic example of a non-native form in English, these assumptions would attribute the production of sentences such as I ate quickly my sandwich to the absence of the knowledge that the word order verb-adverb-direct object is not "allowed" in nativelike English (formulated in terms of some current syntactic theory). The other, processing-centered, assumptions would claim that the production of such utterances indicates that the speaker has not acquired sensitivity to the word order verb-adverb as a parsing cue, indicating that the verb is intransitive (MacDonald 1994). If the speaker had acquired sensitivity to that feature of English word order, then s/he would experience a processing anomaly in such sentences, since they contain a cue that signals the absence of a direct object to the parser, but they also contain a direct object. In a parallel sense, the view taken here is that the production of nonnativelike inflections indicates that the processor is not sensitive to the inflectional forms involved, and therefore does not notice the anomalies that such productions create for native listeners. In investigating the acquisition of inflectional form-function relationships in L2 comprehension processes, we are hampered by our limited understanding of those complex relationships, that is, of the nature of the specific processes triggered by particular inflectional forms in particular contexts. Consider the functions of grammatical gender for example. A long recognized function of gender inflection in some languages is identifying a particular agreeing noun as the subject or object. However, other functions of gender inflection are less obvious, such as lexical priming or inhibition (Friederici & Jacobsen, 1999), or constraining parsing decisions (Van Berkum , Brown & Hagoort, 1999). In the case of other inflectional categories, the functions of a form may remain largely a puzzle. Subject-verb number agreement in English robustly affects the native processor, so that agreement mismatches such as Your brother work.... cause processing difficulties for listeners and readers (Nicol, Forster & Veres, 1997; Pearlmutter, Garnsey & Bock, 1999). Nevertheless the precice nature of the processes triggered by this inflectional form has not to our knowledge been investigated.5 Furthermore, if a form has two or more simultaneous functions, an L2 user might have acquired one function but not another. Friederici et al. (1990) found evidence that for some agrammatic patients, inflectional forms that no longer functioned in connection with person and number marking, nevertheless still functioned as markers of grammatical categories (noun vs. verb). It is logically possible that a similar partial functioning might exist for an L2 user, and that the acquisition of some proper subset of the functions of a form would result in production patterns that would not be wholly nativelike. Although it may be difficult or impossible in many cases to determine whether a specific form-function connection has been acquired, it is nevertheless possible to investigate whether some function or functions have developed for a form. If the L2 processor is totally insensitive to a form that the native processor is highly sensitive to (such as the English subject-verb agreement inflection), this would indicate that the form is inert to the L2 processor (functionless). Thus even without an elaborate understanding of all of the functions of all inflectional forms, it should be possible to investigate the relative tendencies of particular forms to become active (or remain inert) in relation to L2 users' language processors. Over the long term, this line of investigation would presumably best be pursued using implicit techniques, where the experimental participants are not aware of what is being tested (see Juffs, 2001, for a review of the use of some of these techniques in L2 research). In our own research, we have so far relied on an explicit error detection paradigm, but manipulating the attentional demands and temporal processing constraints, as explained below. Participants are presented with inflectional cues that conflict with the other grammatical and semantic cues (Bates & MacWhinney, 1989; MacWhinney, 1997). For example, consider the following sentence with the apparent subject anomalously marked with instrumental case (simplified from Experiment 1 in Thomson, 2000): *Potom c&elovek-om ctal myt' ruk-i. person-instrumental.case began to.wash hand-plural. The person began to wash his hands. (Nonanomalous form: Potom c&elovek ctal myt' ruk-i.) The contsraints imposed by the discourse context, by knowledge of the world, by the accompanying pictures, and by the verbal agreement all require that the c&elovek, 'person' be understood as the subject (the individual at the focal center of the mental model of the discourse at that point) and also as the agent (volitional performer of the action). However, the instrumental case imposes conflicting constraints requiring either that the c&elovek, 'person' not be the agent, or not be the subject (in the latter case, passive verbal morphology must also be present). Assuming the L2 user can retrieve the lexical meanings, and make use of knowledge of the world, etc., then we can predict that if the listener is processing the instrumental case form in accordance with some relatively nativelike functions of that form, s/he will be unable to process this sentence in any way that successfully satisfies the requirements of all processing cues. Thus the L2 user's processor will encounter an anomaly, and react to it as such. One or both of two factors will cause the listener to experience the clash of comprehension cues in this example. First, the listener's language processor may expect nominative case whenever all other factors require that the noun in question be construed as subject. Second, the instrumental case-marking may itself trigger processes which try to assign some role to the noun in question, such as the instrument role. For either of these factors to come into play, the case form of the noun must be processed. Thus, if the form is perceived, but the listener experiences no anomaly, we can conclude that the form was not processed, but was rather inert to the processor, at least on that occasion, and possibly in general. Thomson's (2000) study of sensitivity to inflection in L2 Russian investigated four noun case values, along with verbal aspect, verbal gender agreement, and verbal person agreement. The number of exemplars of each category was small (one or two). However, in order to illustrate how a even single exemplar of a category can be revealing, and also to illustrate how comprehension data can provide an important corrective to conculsions based on production data alone, we will consider the findings related to the verbal perfective-aspect marking prefix system. The chosen stimulus was a verb, stojat', 'stand' that Brown (1996) ranks as 104 in his frequency ranking of 10,000 high frequency Russian words. Thus, it was assumed that this word would be highly familiar to most participants (all of whom used Russian extensively as a means of communication). In the experiment, this stem was anomalously perfectivized using an extremely common perfectivizing prefix po-. When that prefix is added to that particular verb stem, the meaning becomes 'stand for a proscribed period of time', carrying an implication of volition on the part of the one standing. In the experiment, this perfective form was substituted for the imperfective form in a sentence about the furniture standing in a room. Native Russians uniformly react to the sentence as anomalous, and tend to find it humorous (as one reacted by adding, "And then the furniture walked away."). Under the most demanding listening conditions of the experiment (intended to permit only highly automatic inflectional processing) only two out of thirty-seven non-native participants reacted to the anomaly (unlike thirteen out of thirteen native control participants in the same task). Under less demanding listening conditions nine additional nonnative participants (close to 30% altogether) reacted to the anomaly. When presented with the sentence in written form, and without time constraints, ten more (altogether then, nearly 60%) were able to locate the anomaly and correct it. Before locating the anomaly in the written modality, participants were informed that one and only one mistake was present in the sentence. Thus, there was maximum scope for applying L2 metalinguistic, analytical knowledge. It is safe to say that most, if not all, of the learners had been repeatedly taught about the perfective-imperfective distinction as part of their formal Russian training. The fact that such a common aspectual prefix, attached to such a familiar verb, attracted no reaction from more than two thirds of L2 listeners suggests that the prefix was largely inert to the L2 comprehension systems of most participants. Relying on production data in an investigation of the acquisition of L2 Russian aspect, Relying on extensive production data from thirty-two L2 Russian users, Boots-Ebenfield (1995) attempted to find evidence for acquisition of the aspectual system. His complex argument was based on the distribution of different morphological varieties of perfective and imperfective verb stems in the spoken production data. He took the fact that the distribution of these varieties became more like that of native corpora following a study abroad experience as possible evidence of acquisition in process. However, with one item in one experiment, Thomson (2000) was able to raise serious doubts about the robustness of acquisition of the aspectual system in L2 Russian. What should be a highly salient perfective inflection attached to an extremely high frequency verb, was inert to the processors of a substantial portion of L2 Russian users at the proficiency levels examined by Boots-Ebenfield, and even well beyond those levels. That is, it appeared that listeners treated the prefixed form as equivalent to its unprefixed counterpart, in essence, filtering out the prefix. If a formal distinction such as the one involved in Russian prefixal perfectivization is simply filtered out by listeners' L2 processors , there is reason to question the extent to which its form and function are being connected in the language processing systems of L2 learners, i.e., the extent to which it is being acquired in a truly linguistic sense.6 As noted, the overall picture that emerged in Thomson (2000) suggested that the acquisition of inflection was in progress for typical learners, but that the progress might be so gradual that full acquisition would be unlikely for many learners. This is represented graphically in Figure 1. Compare the performance of the lower group of L2 users (less than four years since onset of learning) and the higher group (more than four years) in Figure 1, especially in the Dual Task (a divided attention task demanding relatively automatic anomaly detection) and the Listening Only task (in which the processing conditions were less demanding). Note that the native controls achieved a ceiling level in the Dual Task, and thus did not perform the other tasks. It can be seen that the more advanced L2 users performed substantially better than the less advanced ones, leading to the conclusion that growth occurs. Yet the difference between the more advanced learners and native controls is far more striking than the difference between the two groups of L2 users, leading to the conclusion that this growth is inadequate for the achievement of complete acquisition in the course of several years. Figure 1: Mean percentage of inflectional anomaly detections by L2 Russian users and native controls in Experiment 1 on Thomson (2000). (< 4 years, mean 2.15 years, n = 17; > 4 years, mean = 5.79 years, n = 20; native controls, n = 13). As noted this protracted, and presumably generally incomplete, acquisition of inflectional form-function connections in L2 Russian was attributed to the complexity of the Russian form-function relations, which are considerably more complex than we indicated when discussing the fusional nature of Russian inflection above. Consider the inflectional suffix -u. Depending on the noun declension, the ending -u can mark nearly any of the non-nominative cases: mal'chiku, 'to the boy' (dative); devochku, 'the girl' (accusative); risu, 'some rice' (partitive genitive); na polu, 'on the floor' (locative); dverju, 'with the door' (instrumental). Looking at matters from the opposite direction, a single case value, again depending on the noun declension, can be marked by a variety of forms. For example the accusative case can be marked by the endings -Ø, -u, -a, and -o. (To further complicate matters, in the latter case, and with some instances of -Ø, the nominative-accusative distinction is neutralized, while in the case of -a, the accusative-genitive contrast is neutralized for animate nouns only.) Besides such inflectional homonymy and arbitrary allomorphy, each case value is used in a variety of functions. For example, instrumental case can signal the semantic roles of instrument, agent, time, or path. Individual verbs may also control the case form of their complement nominals (for example, the verb pravit', 'direct' requires that its complement be marked with instrumental case, rather than usual accusative case). Some of the semantic roles marked by instrumental case, e.g. agent and time, can also be marked by other cases (nominative and accusative respectively). In addition, case marking interacts with prepositions to express a variety of specific oblique noun roles. We already noted that two or three grammatical meanings may be fused in a single affix. Add to this other possible functions of case forms, such as word demarcation, lexical category identification, and hence, syntactic parsing-related functions, and it should be abundantly clear that the relationship of form to function in Russian case inflection is a many-to-many relation in a variety of senses. If deviation from one-to-one relationships makes acquisition more difficult, than it should not be surprising that complete L2 acquisition of Russian inflectional morphology is extremely prolonged, and probably commonly incomplete. Turning to Kazak, from a descriptive standpoint, the case system must also strike many learners as complex, due primarily to the phonological alternations involved. Many Turkic languages are traditionally said to have vowel harmony. In the case of Kazak, this is better termed place-harmony, since it involves both vowels and consonants, and both tongue position and lip rounding. In addition, various consonant assimilations and dissimilations occur related to voice, sonority and nasality. Thus the genitive case ending has the forms: -nyN ~ -niN ~ -dyN ~ -diN ~ -tyN ~ -tiN while the dative case ending has the forms: -ƒa ~ -qa ~ -ge ~ -ke Thus, each noun case (except the nominative) has multiple forms. However, there is not the sort of arbitrariness that is seen in Russian. The forms are obviously phonologically related, and the system is totally regular. Place-harmony occurs more generally than just with case suffixes, applying to a wide variety of suffixes. Like Russian case endings, Kazak case endings can also have more than one function. For example, the oblative case can express the point of origin of motion ('from the house'), and also notions including temporal origin ('from 6:00'), and nonvolitional cause (as in 'uncomfortable from the heat'). Instrumental case can express instrument, accompaniment and path (as in 'walked along the road'). However the extended functions appear to be semantically based, and there is little of the sort of lexical government of case that is seen in Russian.7 The extent of interactions between case and adpositions also differs in the two languages. Russian employs a large variety of case forms and functions for nouns governed by prepositions, and the functions can be arbitrary to varying degrees. In Kazak, postpositions uniformly govern a particular case (with some postpositions governing nouns which are not even case-marked, e.g., the postposition üs&in, which marks benefactees and purposes) and there is typically a semantic basis (e.g., the postposition qaraj, 'toward/to' governs the dative/directional case). In general, then, the number of functions of each case-affix in Kazak is smaller than in Russian, and with less semantic arbitrariness. This picture of Kazak is somewhat complicated by the fact that there is a second set of case forms for possessed nouns. In the possessive construction Kazak marks both the head and dependent nouns (in the sense of Nichols, 1986). The dependent noun (possessor) is marked with genitive case, and the head noun (possessed) is marked with possessive morphology. First and second person possessors are marked with person endings that are also used with verbs following certain verbal suffixes. Those person endings are then followed by the case ending indicating the case of possessive NP as a whole. This is illustrated by the dative possessive phrase meaning 'to my house' in the following example: men-iN üj-im-ge first.person.pronoun-genitive.case house-first.person-dative.case to my house Compare this with the simple dative form: üj-ge house-dative.case to a/the house With third person possessors (and with dative possessed singular forms in general), matters are more complicated. With third person possessors, the dative form has -ne ~ -na rather than the general -ƒa ~ -qa ~ -ge ~ -ke; the accusative case has the form -n, rather than the general -ny ~ -ni ~ -dy ~ -di ~ -ty ~ -ti; and the locative case has -nde ~ -nda in place of the general -de ~ da~ -te ~ -ta). An example is the form meaning 'to his house': on-iN üj-i-ne third.person.pronoun-genitive.case house-third.person-dative.case to his house Since there is no obvious natural phonological basis for these alternate case forms which might relate them to the more general case forms, this can be considered partially arbitrary allomorphy. That is, there are two separate sets of forms for marking accusative, dative and locative cases. Note that this is not the sort of lexically based, abritrary allomorphy that is seen in Russian, but nevertheless it introduces what could be called inflectional homonymy. In fact, these forms are rather widespread. There is a tendency to use them without overt possessive nouns or pronouns (for example, s&es&e-si, mother-possessed, 'the mother', banka-sy, can-possessed, 'the can', in a context where someone is holding the can) and for heads of many noun-noun compounds (kitab dükan-i, book shop-possessed, 'bookshop'). A final complication in Kazak involves the fact that direct objects are not always case-marked. Although this further diminishes the extent to which form-function relations can be called one-to-one in Kazak, the accusative case itself plays no direct role in the current study. It should be clear that Kazak morphology is not descriptively simple. To language learners being given a descriptive explanation of the allomorphy, it seems doubtful that Kazak, with phonological place harmony involving vowels and consonants, would come across as that much less complicated than Russian. For example, learners might feel intimidated to hear that there are six forms of the plural suffix (-lar ~ -ler ~ -dar ~ -der ~ -tar ~ -ter8. However, there is a basic difference in the nature of the complexity in the two languages, with Russian complexity resulting from arbitrary multiple forms for each case, and Kazak having linguistically more natural and regular sets of forms for each case value, in line with overarching aspects of the phonology that apply to almost all suffixes. In addition to having simpler sets of forms, the set of functions signaled by the Kazak forms is smaller, and arguably more semantically motivated than is the case with the Russian forms. Therefore, any advantage of Kazak over Russian in L2 acquisition of the case system would arguably reflect the relative advantage of the agglutinative form-function connections over fusional form-function connections, and not simply reflect a greater ease of L2 metalinguistic "rule learning". In addition to differences between Kazak and Russian related to the ease or difficulty of inflectional acquisition, we might expect to find similarities in the pattern of acquisition, since the factors proposed by Thomson (2000) to explain the apparent pattern of case acquisition in Russian would conceivably be of cross-linguistic generality. The comparison of L2 Russian with L2 Kazak reported here is based on the first experiment reported by Thomson (2000), since that experiment provides excellent parallels between the two languages9. For the most part, only the case-marking inflections are involved in this cross-language comparison, since it was possible to adapt the Russian experiment to Kazak fairly readily in the domain of case-marking. In Experiment 1 of Thomson 2000, it was argued that Task 1 (here called the Dual Task) detected too little sensitivity to anomalies for a clear pattern to emerge, while in Task 3 (here called Printed Form), the unbridled application of L2 metalinguistic analytical reasoning also obscured any systematicity that might have resulted from ordinary language processing mechanisms. However, regarding the pattern of sensitivity to case anomalies in Task 2 (here called Listening Only), the following observations were made : The largest number of reactions resulted from anomalous oblique case forms (locative or instrumental) in a nominative (subject) contexts.10 The smallest number of reactions resulted from anomalous nominative case-marking in oblique (instrumental or locative) contexts. The anomalous substitution of one oblique case for the other (instrumental for locative or locative for instrumental) provoked an intermediate number of reactions. This general pattern can be seen in Figure 2. Figure 2: Sensitivity to case anomalies, as reflected in percentage of participants detecting errors in Experiment 1 of Thomson (2000). The following set of assumptions would have such consequences:11 First, the function of oblique case-marking as a signal of nonsubjecthood (indicating that the referent is not the central entity in the narrative model at the moment) develops more readily than other functions of case marking (hence the relative detectability of anomalously oblique-marked subjects). Second, the function of nominative case-marking as a signal of subjecthood does not develop particularly readily (hence the relative non-detectability of anomalously nominative-marked oblique nouns). Third, subjecthood is strongly determined by factors other than case (allowing the anomalous oblique marking to clearly clash for many L2 users). Fourth, oblique case marking as a signal of particular semantic roles, while not developing as readily as the function referred to in 1), does have some tendency to develop (accounting for the moderate tendency of L2 users to react to anomalous oblique-for-oblique substitutions). Fifth, although oblique case-marking may develop the functions referred to in 1) and 4), the expectation of oblique case-forms in appropriate oblique contexts is slower to develop (further accounting for the low detectability of nominative-for-oblique substitutions). Such considerations will be important in comparing the patterns of sensitivity to anomalies in L2 Russian and L2 Kazak. Turning then to the Kazak experiment, and to the comparison of the Kazak experiment with the first Russian experiment of Thomson (2000), we make the following predictions. In relation to the Kazak experiment by itself, the predictions are the same as those of the Russian experiment: 1) There will be a significant effect of group, as defined in terms of time since onset of learning, indicating that learning related to inflectional processing in L2 Kazak does occur. That is, most learners' L2 processors do not tend to totally and permanently ignore inflectional form as redundant and uncecessary to comprehension. 2) There will be a significant effect of task, indicating that decreased processing demands and increased opportunity to apply L2 metalinguistic knowledge will result in increased error detections. 3) The sensitivity to the various inflectional anomalies will be distributed in a non-random manner. That is, there will be evidence of patterning in the development of sensitivity to various types of inflections and inflectional contexts. The central aim of this paper, however, is to compare the L2 Kazak findings with the L2 Russian findings in order to test the following predictions: 4) Kazak learners at different developmental levels will show a clear advantage over Russian learners at comparable developmental levels in terms of sensitivity to anomalous case marking (and by implication, in the level of development of the L2 case processing system). 5) There will be a positive correlation between the general patterns of development of sensitivity to case forms (reflected by the distribution of anomaly detections) in the two L2s. Hypothesis (4) reflects the predicted effect of morphological typology on ease/difficulty of acquisition, while hypothesis (5) reflects the expected effect of universal principles. These principles could work against one another, in that a strong effect of cross-linguistic principles could diminish the effect of differences due to morphological typology. Since this is the first research of its kind related to Kazak as a second language, other aspects of the experiment will be of interest in addition to the comparison with the Russian experimental results. In the discussion below some of these will be explored together with findings directly related to the specific hypotheses. Method Kazak Experiment Participants Thirty-five non-native speakers of Kazak participated along with fifteen native Kazak control participants. The experimental participants were divided into a low group and a high group on the basis of the time since onset of Kazak learning, the dividing point being four years. This was found to be a practical and statistically viable way of dividing participants for a pseudolongitudinal comparison in Thomson (2000). The mean time for the low group was one year, nine months (1;9) (range 0;8 to 1;0, median 1;10), and for the high group, the mean was 7;8 (4;0 to 11;3, median 6;8). The overall mean time since onset of learning was 4;8 (median 3;6). By comparison, in Experiment 1 of Thomson (2000), the average time reported for the low group was 2;2 (0;6 to 3;6, median 2;0). The average time reported for the high group was 5;9 (4;0 to 15;0, median 5;0). The overall mean years was 4;1 (median 4;0). It would have been desirable to better match the time since onset of learning for the two languages. That was simply not possible. In both cases, all willing volunteers who could be located within the time constraints of the research participated. In fact, even if the groups were perfectly matched for time since onset of learning, other important variables would be uncontrolled, such as the total amount of experience using the languages in comprehension and production. It should be kept in mind, then, that the comparability of the Russian groups and Kazak groups in terms of true developmental stages cannot be terribly precise, and therefore a clear effect of language morphological typology will be needed to support hypothesis 4) above. Recall that the two languages are claimed to differ considerably in the extent to which they approximate or deviate from one-one form-function relationships. We are predicting a strong effect of this cross-linguistic difference, so that the high and low groups in the two experiments can be considered similar enough for the difference to be detected. In the current, Kazak experiment, the L1s of the learners were Korean, 6 (3 in the low group, and 3 in the high group); English, 21 (11 low, 10 high); Japanese, 1 (high); German, 2 (high); Norwegian, 2 (low); Chinese, 3 (2 low, 1 high). In the corresponding Russian experiment, the L1s were English: 17 (9 low, 8 high); Chinese: 8 (3 low, 5 high), Finnish: 3 (2 low, 1 high), Spanish: 2 (high), German: 2 (1 low, 1 high), Farsi: 1 (low), Italian: 1 (high), Japanese: 1 (low), Korean: 1 (high) and Turkish: 1 (low). Twenty-one males and fourteen females participated in the Kazak experiment. Seventeen males and twenty females particpated in the corresponding Russian experiment. All of the participants, either by self-report or based on our observations, use the target language regularly as a major means of communication in their daily lives. However, twenty-seven of the thirty-seven participants in the Russian experiment were full-time international students in Russia, with heavy daily immersion in the target language, while most of the Kazak learners work with non-governmental organizations in Kazakstan, and the extent of their daily exposure varies depending on whether they live with Kazak speakers, employ Kazaks in their homes, and whether their work involves them in regular interaction with Kazak colleagues or clients.12 Again, it was predicted that the effect size of morphological typology would be strong enough to override the differences in total amount of exposure between L2 Russian users and L2 Kazak users. That is, whereas the Russian inflectional system appeared to be highly resistant to full L2 acquisition by adult learners, the Kazak inflectional system was predicted to be substantially more readily acquirable. Download 1.36 Mb. Do'stlaringiz bilan baham: 
|