Chicago under revision
Download 1.36 Mb.
|
Morphological Typology and SLA Inflectio
|
Procedure
Participants performed the experimental tasks in groups numbering from one to five participants, with one to two participants being typical. As with the corresponding Russian experiment, the Kazak experiment was carried out under field conditions, rather than in controlled laboratory conditions. This meant that there were differences in background noise, temperature levels, possibilities of interurpions, etc. Every effort was made to minimize these factors, and our subjective evaluation was that such variables were reasonably comparable across the two experiments, but may have added a small amount of noise to the data. Task 1: Dual Task (picture selection and error detection) Participants listened to the texts, and after hearing each numbered segment they chose the appropriate picture, a, b, or c, and wrote the letter on the answer sheet. In addition, they occasionally marked an "X" next to the letter, indicating the presence of a "mistake". After participants listened to the first segment of Text 1, and attempted to choose the appropriate picture, there was free discussion, during which participants were asked if they had heard a mistake. If they had not, the mistake was pointed out to them. Following discussion of the first item, the texts were played continuously, with only a small break between texts one and two. In order to maintain the five-second pause following the final segment of Text 1, the words "End of Story 1" were included on the tape at the appropriate point. Participants were instructed not to worry if they had to leave picture-selection items blank, and urged not to mark an X, indicating a mistake, unless the mistake was very clear to them. In fact, the anomalies were of such a nature that participants would likely either miss them or be certain about. That is, they were crashing, elementary mistakes from a native speaker viewpoint, and would also be clearly elementary from an L2 pedagogical viewpoint. Task 2: Listening Only Apart from the printed instructions, the response forms were the same as for the Dual Task, with a number corresponding to each segment of the text. The pictures were removed, and the participants were instructed to leave the numbered items in the answer sheets blank unless they felt confident they had heard a mistake, in which case they were to mark an "X" next to the number of the text segment. Task 3: Printed Form The participants were given the numbered items containing the inflectional anomalies in printed form, and informed that there was one and only one mistake in each of the segments now before them. They were to find and correct the mistakes. There was no time limit in this task. Total time for the experiment, including explanations, was consistently around thirty minutes. Results The fifteen native controls only performed the Dual Task. They detected 91.67% of the anomalies. Only one anomaly was missed, and it was missed by five participants. This was in segment 5, the first oblique-for-oblique case subsitution (ablative for instrumental). For the L2 Kazak users, the mean anomalies detected by the high group in the Dual Task was 1.88 out of 8 possible (cf. the average of 7.33 for the native controls), or 23.5%. The low group mean anomaly detections was .72 anomalies out of a possible eight, or 9%. The high group participants detected an average of 4.53 out of 8 possible anomalies in the Listening Only task, still well below the performance of the native controls' average of 7.33 in the Dual Task. In the Listening Only task, the number of detections by high group participants ranged from 0 to 6 (of the possible eight) , with 94.44% of them detecting from 0 to 3 anomalies. The low group participants detected an average of 2.61 anomalies out of 8 possible in the Listening Only task, ranging from 0 to 3 with 83.33% of them detecting 0 or 1. Finally, the high group detected an average of 6.53 of the 8 possible anomalies in the Printed Form task (approaching the average number of detections by native controls in the Dual Task), while the low group detected an average of 4.67 out of 8 possible anomalies in that task. These results are summarized in Table 1.
Table 1: Summary of mean anomaly detections. The high group began learning Kazak an average of seven years, eight months earlier, while the low group began an average of one year, nine months earlier. A two-way ANOVA was performed (task group). The results are displayed in Table 2, and Figure 3. Looking at the two groups and three tasks as a whole, the ANOVA revealed significant main effects of both task (F[2,33] = 71.455, MSe = 2.212, p < .001) and level (low vs. high), (F[1,33] = 10.814, MSe = 5.007, p < .01). The interaction of task by level was marginally significant (F[2,66] = 3.079, MSe = 2.212, p = .0527). Table 2: ANOVA source table for Kazak experiment, three tasks by two levels. Figure 3. Interaction graph for Kazak Experiment. Maximum possible anomaly detections was eight. Turning to the examination of the pattern of reactions to items, the 2 text allows a conclusion that the distribution of responses across different items was nonrandom, suggesting that some anomalies are more readily detectable that others. The distribution of responses to the different anomalies is shown in Table 3.
Download 1.36 Mb. Do'stlaringiz bilan baham: 
|