5.2 Results and discussion by task
5.2.1 The grammaticality judgment task
5.2.1.2 Comparisons across all language subgroups: ILs vs. ILs of the
Although the comparisons conducted above among the different subgroups of proficiency levels (lower-intermediate, upper-intermediate and advanced) within each L1 group of learners show that the higher the proficiency level, the better at rejecting ungrammatical sentences with null subjects, we cannot assume that starting off by accepting ungrammatical sentences with a null subject is a developmental phenomenon affecting all L2 learners regardless of their L1 background until we “compare the developmental paths of L2 speakers of typologically different L1s. If there is divergent development, then this [may]
constitute evidence for transfer” (Hawkins, 2001b, p. 354). In contrast, if a similar pattern of development is noticed among learners from different linguistic backgrounds, then this constitutes evidence for UG. For better comparative visualization, the results in Tables 1 and 3 are summarised below in Figure 5.1 in terms of means of ungrammatical acceptance.113
113 One of the examiners suggested that the use of line graphs may have been a better way to show group performance. I have not followed this recommendation because I think that such R boxplots manage to present the complex data in a comparatively simple and meaningful way. Given to the large number of sub-groups being compared (nine sub-groups in Figure 5.1 above), the use of such boxplots makes figures easier to read and interpret compared to the use of line graphs. This is true especially in cases when the performance of the nine sub-groups of participants included in the present study are
Figure 5-1. Summary and comparison of the subgroups’ results in terms of means of null subject acceptance.
Therefore, inferential tests were conducted for comparisons between the L1 French, L1 Finnish, and L1 Arabic learners of the same proficiency level to determine whether the source of their performance was the product of their L1 backgrounds or UG. A one-way ANOVA revealed that there is a statistically significant difference with respect to accepting sentences involving null referential subjects between the lower-intermediate subgroups of the L1 French, Finnish, and Arabic learners (P-value = 0.0032<0.05). Similarly, a Kruskal Wallis test showed a significant difference between upper-intermediate subgroups for the same variable (P-value = 0.0021<0.05); however, there was no significant difference between the advanced
compared for two variables as in Figure 5.17 on page 202. Note, such boxplots are only used for better comparative visualisation; the groups’ results are reported in tables including information about the mean, minimum, maximum, standard deviation, and percentage of acceptance or omission for each variable under investigation.
subgroups (P-value = 0.1305>0.05) – a finding which suggests that interlanguage (IL) restructuring occurs with time and increased proficiency.
It should be mentioned for the sake of illustration that there is a noticeable difference between the non-target performance of L1 Arabic learners and their French and Finnish counterparts. It is clear from Figure 5.1 above that L1 Arabic learners generally accepted more sentences with null embedded subjects than did their French and Finnish counterparts of the same proficiency level. It is, however, surprising that the performance of French learners as subgroups was comparable to that of lower-intermediate, UP, and advanced Finnish learners though their L1s are different in terms of parameter setting; they accepted about the same number of ungrammatical sentences at mean scores of 12.44, 3.00, and 0.52 for the French lower-intermediate, upper-intermediate, and advanced subgroups, respectively, and 11.67, 3.12, and 0.52 for lower-intermediate, upper-intermediate, and advanced Finnish learners, respectively. These observations are supported statistically by the results of the inferential tests performed as shown in Table 5.5; the results of the French vs. Finnish comparisons are shaded.
Table 5-5. The results of the statistical tests comparing the different subgroups of the participants on acceptance of null subjects
Proficiency level Subgroups compared Inferential test P-value
ADV Arabic learners and L1 French learners and between the L1 Arabic learners and L1 Finnish learners are all significant at both lower-intermediate and upper-intermediate proficiency levels with respect to their performance on the variable under discussion.
In contrast, the statistical tests showed that the differences between the L1 French
learners and the L1 Finnish learners of the same proficiency level were not in fact significant among all the proficiency subgroups. In other words, the data reveal that the Arabic-speaking learners accepted null subject sentences at a much higher rate than their French and Finnish counterparts, who unexpectedly accepted null subjects in a somewhat similar fashion.
Given how the three languages differ from each other in terms of the possibility of allowing null subjects in tensed clauses, the noticeable differences between the non-target-like performance of Arabic learners whose L1 allows subject-drop freely and their Finnish counterparts whose L1 allows subject-drop optionally in some contexts and excludes them in other contexts and between the non-target-like performance of Arabic learners and their French counterparts whose L1 does not allow subject-drop can be attributed to transfer from the L1. However, the L1-based knowledge influence cannot explain the similarities in performance between the L1 Finnish learners and the L1 French learners. Yet it can be argued that the Finnish participants accepted null subjects in a somewhat similar fashion as their French counterparts because Finnish allows null subjects in a restricted manner. This would lead to their preference for overt pronouns over null forms. Nevertheless, L1 transfer cannot explain why all of the lower-intermediate subgroups of learners, including the French, accept ungrammatical sentences with null embedded referential subjects despite the fact that their L1s are different in parameter setting (full non-pro-drop French, partial pro-drop Finnish, full pro-drop Arabic). This will be discussed in detail in the following section.