Chicago under revision
Download 1.36 Mb.
|
Morphological Typology and SLA Inflectio
Table 3: Responses to specific anomalies. Total possible = 35. Based on the same theoretical predictions as those of the Russian experiment in Thomson (2000), the order of case anomalies should be as shown in Table 4. Although the rank orders for the two listening tasks does not correlate significantly with the predicted order, the rank orders of the two tasks do correlate highly with one another (= .986, p = .028).14
Table 4. Theoretically predicted order of sensitivity to anomalies, and observed orders. It is also possible to combine the anomalies as in Table 5. This was done in Thomson (2000). However, given that the pairs of anomalies behave so differently here, this is a clearly suspect move, and with the small number of categories, the rank order correlation can not be significant, at any rate.
Table 5. Theoretically predicted order of sensitivity to anomalies, and observed orders, with anomalies of the same category (two of each) combined. Discussion The marginal interaction detected in the ANOVA appears to reside in a bigger gain on the part of the high group in going from the Dual Task to the Listening Only task (see Figure 3). It would not be surprising if the high group were better able than the low group to take advantage of the improved processing conditions of the Listening Only task, vis-a-vis the Dual Task. In fact, however, the difference in mean gain for the two groups (.76 error detections) is not signicant (t[34] = 1.3914, p = .1734). Predictions (1) - (3) of the introduction were supported. With regard to the first hypothesis, some level of learning that affects sensitivity to inflectional anomalies does appear to occur over the time frame represented by the differences between the two groups in time since onset of learning. With regard to the second hypothesis, improved processing conditions of the Listening Only task (and possible increased scope for applying L2 metalinguistic knowledge) led to a clear improvement over the heavily meaning-focused conditions of the Dual Task. This is compatible with an account according to which L2 processing mechanisms are highly sensitive to processing conditions, unlike the robust processing mechanisms characteristic of native listeners. Also related to hypothesis 2, in comparing the Printed Form task with the listening tasks, we find evidence that, given time to refect and analyze, formally instructed L2 users will show an extent of awareness of grammatical form that in some respects exceeds the extent to which they show sensitivity to grammatical form-function relationships in normal processing. The difference between groups in this task, given the elementary nature of the pedagogical rules related to the inflectional anomalies, may indicate that the level of mastery of basic L2 metalinguistic rules is limited by the level of development of genuine L2 language ability. Finally, in relation to hypothesis 3, there appears to be a nonrandom pattern in the tendency of different anomalies to trigger reactions, although it is not completely clear to what extent this was due to the sort of general developmental principles argued for in Thomson (2000), and to what extent it was due to other factors more specific to Kazak. In short, most nonnative Kazak participants displayed little sensitivity to anomalies in the very circumstances where the native controls displayed a high level of sensitivity (the Dual Task condition). Participants differ in their tendency to detect anomalies, although only two of the seventeen high group participants detected all of the anomalies even in the Listening Only task. In fact, fourteen of the seventeen reacted to half or less of the anomalies in that task. Recall that in that task the stories were already familiar to the listeners, the exact sentences having been heard shortly before, and there were ten second pauses to allow extra time for processing. In the Dual Task, sensitivity was lower still. The Dual Task was intended to be the most revealing of level of sensitivity to form that would characterize normal listening, since if forced listeners to focus heavily on meaning. These findings suggest that inflectional acquisition by adult learners of Kazak is not particularly rapid, nor necessarily thorough. Although learning occurs, if we were able to project how much time would be required for full acquisition of inflectional processing, we suspect that it would reflect the need for an enormous amount of experience processing accessible input. The acquisition of robust inflectional processing in Kazak appears to be a long term prospect. Under the assumption that the development of normal linguistic production mechanisms depends on the prior development of corresponding comprehension mechanisms, we might wonder whether enough sensitivity to inflectional form-function relationships would develop in the early years of Kazak learning to strongly influence the production system. Beyond the overall levels of sensitivity to inflectional anomalies, the pattern of sensitivity to the various anomalies (see Figure 4 below) is of interest because of the light it could shed on the nature of the acquisition process. It was noted that the distribution of anomaly detections, i.e., the tendency of participants as a group to be sensitive to certain anomalies more than others, does not appear to be random (2 (7) = 34.16, p < .001 in the Dual Task, and 2 (7) = 28,2195, p < .001 in the Listening Only task). In the same regard, the inability to reject the null hypothesis in the Printed Form task (2 (7) = 110,2887, p = .1728) may suggest that when L2 metalinguistic knowledge and strategies can be brought to bear on the task, the results will not reflect systematic developmental processes. That, of course, should not be surprising, but should serve as another warning to researchers who would employ acceptability judgments to written stimuli as a dependent variable. Figure 4: Number of participants detecting specific anomalies in the three tasks. (Maximum Possible 35) Various factors might have contributed to the nonrandomness of the distribution of reactions across items. With only two examples of each general substitution type (nominal-for-oblique, oblique-for-oblique and oblique-for-nominal) it is easy for unintended factors to distort the picture. Some possible factors are trivial, such as gaps in lexical knowledge. Also, it was expected, based on the Russian experiment, that in the Dual Task the first oblique-for-oblique substitution (ablative-for-instrumental) would not trigger many reactions, due to its falling within a refractory period following the relatively highly detectable anomaly of the preceding segment.15 Note that even a third of the control participants appear to have been similarly affected.16 There are other more interesting uncontrolled variables, such as various syntactic and discourse properties of the contexts in which the anomalies occurred.17 However, such factors did not appear to play a major enough role in the Russian experiment to entirely obscure the principled systematicity that was being sought. The theoretical considerations discussed above, would lead us to expect that the number of case-marking anomaly detections in each of the two listening tasks (but not in the Printed Form task), as they are presented in Figure 4, should decrease monotonically going from left to right through the pairs of case-marking substitutions (oblique-for-nominative > oblique-for-oblique > nominative-for-oblique). In the Dual Task of the Russian experiment, the instrumental-for-locative substitution did not fit the predicted pattern, and this was discussed at some length in Thomson (2000). Although the rank order of the reactions to the items in the Kazak experiment did not correlate significantly with the theoretically predicted rank order, if we allow for the divergences that parallel those in the Russian results, the distribution of reactions as shown in Figure 4 is not wildly out of line with the theoretically predicted pattern, with one exception. The most problematic exception to the predicted pattern is the locative-for-nominative substitution, which should have been one of the two anomalies with the highest number of reactions. Instead, it faired worse than the nominative-for-locative substitution. In the Kazak experiment, as in the Russian experiment, this anomaly occurs in a syntactically complex context: the anomalously case-marked noun is part of a conjoined subject, 'shirt and pants', the anomaly being in the word for pants. In Russian, this was also the only case-marking anomaly in which the noun was plural. Yet, in spite of these complexities, this was the second most frequently detected anomaly in the Listening Only task of the Russian experiment. The corresponding anomaly in the Kazak experiment did not involve a plural noun, since the word for 'pants' is singular in Kazak. However, it did involve an extra inflectional category, since the words for 'shirt' and 'pants' must be marked as possessed nouns (discussed above). That meant that the Kazak anomaly involved the second conjunct of a conjoined NP subject, with the first conjunct in the contextually expected possessed-nominative form, and the anomalous one in the possessed-locative form (see the Appendix). In the Dual task there was only a single participant detected this anomaly, making it clearly one of the anomalies with low detecatability for these participants. Even in the Listening Only task, where its counterpart fared so well in the Russian experiment, there were only 10 detections in the Kazak experiment, again setting it in clear contrast with the other oblique-for-nominative substitution (instrumental-for-nominative), which was detected by 27 participants (out of the possible 35). This low detectability could point to a processing problem with possessed case-marked forms. In fact, before the experiment was conducted, one experienced instructor asserted that possessed case-marked forms are an area of special difficulty in L2 Kazak. It may be that our listeners, confronted with such a form, were prone not to process the inflection at all, which meant, of course, that they were unable to detect the anomaly. This particular difficulty promises to bear on the broader issue of what makes inflectional forms relatively more versus less readily acquirable, and so this area of L2 Kazak merits special further research. For reasons discussed earlier, the nominative-for-instrumental and nominative-for-locative substitutions were expected to be relatively low in detectability. It turned out that this was the case for the former, but not for the latter. One participant indicated responding to an unfamiliar but nonanomalous feature during the two listening tasks, and only discovering the real anomaly in the Printed Form task.18 Another possible explanation is that the locative form for 'in the chair', as with the form for 'in the bed,' has a special status, and is less marked (in Tiersma's 1982 sense of local markedness) than the nominative form. That is, chair as location could be less marked than chair as subject. Obviously, it will require much more careful research with more examples of each anomaly type to solve such problems. A highly detected anomaly in the Dual Task, in line with expectations, was the person agreement anomaly, where the subject was marked as polite second person, in place of the contextually appropriate third person form. A comparable anomaly in Experiment 2 of Thomson (2000) was the most highly detected anomaly in either Russian experiment (detected by 91% of the high group participants in the Listening Only task). It was reasoned that the highly concrete referential semantics of the verbal person endings (despite their frequent redundancy) leads to their relatively ready acquisition, since concrete referents are stable in working memory, allowing favorable conditions for the formation of form-meaning associations. It is perhaps of some interest that in the Kazak experiment the instrumental-for-nominative substitution was noticed as often or more often than the person agreement anomaly in all three tasks. This pattern also holds across the high and low groups. The place-harmony anomaly was an unknown prior the experiment, although, as noted, place-harmony is a pervasive (though not totally general) feature of the language. This anomaly slipped past 31 of the 35 participants unnoticed in the Dual Task, suggesting that place-harmony is not readily exploited by nonnative listeners. (We assume it has important functions for native listeners, for example, a role in word demarcation). In the Listening Only task, there was a three-fold increase in detections, raising the possibility that place-harmony is weakly active role for some L2 listeners. Interestingly, twenty-two participants, including five of the seventeen high group participants, failed to find it in the Printed Form task, again, suggesting that it is not a feature that is consistently robustly acquired by language learners. The Combined Russian and Kazak Experiments Casual inspection of the results of the L2 Russian Experiment 1 of Thomson (2000) and the L2 Kazak experiment reported here will provide a strong impression of similarity in the results. There are comparable main effects, a comparable interaction (though in the Russian experiment, it was nonsignificant, p = .139). Even many details of the relative detectability of specific anomalies appear similar. In order to formally test predictions 3) and 4) of the introduction, we performed an ANOVA directly comparing the two experiments, and investigated the possible correlation between the percentage of participants reacting to specific case anomalies in the two experiments. Results A three-way ANOVA (Target Language Level Task) was performed, and the results are summarized in Table 6 and Figure 5. Table 6: ANOVA source table for comparison of Russian and Kazak experiments (Level = high and low; Task = Dual Task, Listening Only, and Printed Form). The main effects for level and task were expected based on the separate analyses of the Russian experiment (Thomson, 2000) and the Kazak experiment (above). The variance introduced by combining data from the separate languages did not have a negative impact on these effects. What was predicted as well, (prediction 4 in the introduction) was that there would be a main effect for Target Language, with Kazak faring better than Russian, due to the predicted advantage of agglutinative inflection over fusional inflection in second language acquisition. However, no such effect was found (F[1,68] = .050, p = .8229). Figure 5. Interaction graph for comparsion of Kazak and Russian experiments. A significant Target Language Task interaction was found: (F[2,136] = 4.412, MSe= 1.012, p = .0135), apparently due in part to the fact that the difference between the listening tasks and the Printed Form task was greater for L2 Russian users than for L2 Kazak users (see Figure 6). Figure 6. Interaction graph of Target Language by Task in the combined experiments. Turning to the relationship between specific items in the Kazak experiment (considering only the six case-marking anomalies) and the corresponding items in the Russian experiment, we predicted a significant positive correlation, and that is what we appear to have found. For the Dual Task, r = .81, which is significant (p < .05, one-tailed). For the Listening Only task, we find that r = .77, which is also significant (p < .05, one-tailed). For the Printed Form task, r = .53, which is nonsignificant (p > .05). The pattern of sensitivity to the various case-marking anomalies in two listening tasks of the combined experiments is displayed graphically in Figure 7. Figure 7. Comparision of sensitivity to specific anomalies in L2 Kazak and L2 Russian. Download 1.36 Mb. Do'stlaringiz bilan baham: 
|