5.3 Results and discussion by grammatical constructions
5.3.1 The translation task: Subject drop in complement clauses vs. in adverbial
5.3.1.1 The English IL of the French, Finnish and Arabic speakers compared within subgroups
In light of all the descriptive data presented in subsection 5.2.2, which demonstrated that all three groups of learners of all proficiency levels showed high success rates at overtly producing the referential embedded subject pronouns when translating the given sentences into English, it can be predicted that the different L2 syntactic constructions under investigation (complement clause vs. adverbial clauses) have no effect on the learners’ performance as far as embedded referential subject pronouns are concerned. The groups’ results presented above in subsection 5.2.2 are summarized below in Table 5.13 in terms of mean scores of subject drops.
Table 5-12. Summary table of mean scores of sentences with null subjects produced by French, Finnish, and Arabic L2 learners of English
L1 Lower-intermediate Upper-intermediate Advanced
French participants 0.11 0.10 0.00
Finnish participants 0.83 0.18 0.00
Arabic participants 1.75 0.64 0.18
131 Note the results on the previous subsections have shown that the learners did not manage to converge to native-like if the L1 and L2 differ in parameter values.
The subgroups’ low averages use of null referential subjects can be misleading in the sense that it is possible that subject drop might have occurred more frequently in a certain tested syntactic construction but not in another. Therefore, this prediction that the learners’ performances do not vary from grammatical construction to grammatical construction cannot be assumed until a comparison within each subgroup is made between the learners’ performance in complement clauses and in adverbial clauses with reference to referential subject drop.
After breaking down the results of the translation task’s experimental sentences by item types (complement vs. adverbial), the new descriptive statistics of the three lower-intermediate-level subgroups of learners with respect to their use of null referential subjects are presented below in Table 5.14. It shows that both the lower-intermediate French participants and lower-lower-intermediate Finnish participants as subgroups drop subjects at a slightly higher rate on sentences with adverbial embedded clauses (1.1% and 7.5%, respectively) than on those with complement embedded clauses (0.0% and 3.8%, respectively). However, the inferential tests conducted to compare the performance within each subgroup indicated that there were no significant effects for item type. The results arrived at via the Wilcoxon tests were P-value = 1.000>0.05 for the French lower-intermediate learners and P-value = 0.371>0.05 for the Finnish lower-intermediate learners, which suggests that those subgroups of learners treated missing subjects equally across the different investigated contexts.
Table 5-13. Descriptive statistics by items types (complement vs. adverbial) for sentences with null subjects produced by lower-intermediate subgroups of learners
L1 Items type Included
Unlike their French- and Finnish-speaking counterparts, the Arabic-speaking lower-intermediate participants as a subgroup noticeably dropped subjects at a much
higher rate on sentences with adverbial clauses than on those with complement clauses. As Table 5.14 shows, subject omission occurs in 17.4% of the total included sentences with adverbial clauses, whereas it occurs only in 2.7% for sentences with complement clauses. The result of the Wilcoxon test confirmed that Arabic-speaking learners as a subgroup performed significantly differently in the two types of experimental sentences (P-value = 0.005<0.05). The finding that lower-intermediate Arabic learners drop fewer subject pronouns in sentences with complement clauses can be explained by the fact that, in Arabic, a pronominal bound form, which functions as the subject of the complement clause, must be cliticised onto the complementiser ʔanna (‘that’).132
Interestingly, when these lower-intermediate-level results of the Arabic-speaking learners were compared with the results of the English native controls via the Wilcoxon tests, the inferential results show that their performance differed significantly from that of the native English speakers only on sentences with subjectless adverbial clauses (P-value = 0.010<0.05).133 The fact that the difference between the lower-intermediate subgroup and the native group did not achieve significance on sentences with subjectless complement clauses (P-value = 0.571>0.05) indicates that they performed within the native-like ranges at the lower-intermediate-level on this type of structure as far as null subjects are concerned.
As the results in subsection 5.2.2 indicate, both the French and the Finnish lower-intermediate learners performed native-like with regard to the referential embedded overt subject; it is expected that their performance would still be within native-like ranges even after their results were broken down further by sentence
132 For more detail about resumptive pronouns in Arabic, see subsection 2.3.2.
133 Statistically speaking, the results of the native control presented in Tables 5.2 above need not be broken down further by the item types because they did not accept null subjects and therefore did drop them. Thus, where it is required to see how their results diverge from the natives’ and when their performance converged to native-like, the learners’ results for each item type in this section will be compared directly against the control group’s results presented in Table 5.2 for the GJ task to avoid repetition and for space reasons.
structural types. This expectation was confirmed when the inferential tests were applied; all the results obtained by the Wilcoxon tests indicate there were not significant differences between these lower-intermediate participants’ performances and the English control with respect to subject dropping in the different structures tested. Consider the results in Table 5.15.
Table 5-14. Inferential comparisons between the lower-intermediate French and Finnish participants and the native controls: subject drop in complement clauses vs.
in adverbial clauses
Inter-subgroup comparisons Clause type Inferential test P-value LI French participants vs. native
English controls
Complement Wilcoxon test N/A*
Adverbial Wilcoxon test 0.450>0.05 LI Finnish participants vs. native
English controls
Complement Wilcoxon test 0.355>0.05 Adverbial Wilcoxon test 0.140>0.05
*N/A= not available (the results of both groups were identical: zero number of subjects dropped).
It should be mentioned before leaving this subsection that the translation task results of both the upper-intermediate and advanced subgroups were intentionally not presented here because these two proficiency subgroups of learners, regardless of their linguistic backgrounds, performed either completely native-like or very close to native-like with respect to overt referential subject pronouns in English, as shown above in subsection 5.2.2. This implies that the results of these more proficient learners would not reveal any information to confirm or disconfirm whether L2 syntactic structures would bring about different performance with regard to pronominal subject production in L2A.
5.3.1.2 Discussion of the translation task’s results by grammatical structures
The results and discussion in the previous subsection show that the L2 syntactic constructions under investigation have no effect in the performance of both the lower-intermediate French and Finnish L2 learners of English as far as embedded subjects are concerned. Conversely, the lower-intermediate Arabic participants were found to drop subject pronouns differently in the two types of experimental sentences,
dropping subjects at a much higher rate in sentences with adverbial clauses than in those with complement clauses.134
Therefore, as both the French and Finnish lower-intermediate learners performed native-like in the two types of experimental sentences and since the performance of the lower-intermediate Arabic participants was attributed to transfer from Arabic, these findings confirm the prediction of hypothesis H3, which stated that different L2 syntactic structures might not bring about different performance on the overt and/or null realisation of the embedded subject pronoun. However, this hypothesis cannot be confirmed on the basis of the findings that emerged from the translation task alone because
i. the translation task appeared to be quite easy compared to the GJ task, which has proven to be much more difficult for participants regarding embedded subjects (refer to the results and discussion in section 5.2)
ii. Even though no significant effect for item type (complement vs. adverbial) was found in either the French the Finnish lower-intermediate learners’ data, there seems to be a tendency in these groups to drop more subjects in sentences with adverbial clauses than in those with complement clauses. For the sake of illustration, the distribution of the data presented above in Table 5.14 is presented graphically in Figure 5.10 below with respect to the percentage of null referential subject pronoun use.
134 Some possible explanations to account for why the L1 Arabic-speaking L2 learners of English dropped more subjects in adverbial clauses than in complement clauses are proposed below, on pages 185-187.
Figure 5-10. Percentages of subject drop in complement clauses and in adverbial clauses by the French, Finnish, and Arabic lower-intermediate learners of English.
Therefore, we need to examine the groups’ results obtained via the GJ task in both syntactic contexts to be able to see an even clearer picture before a conclusion can be drawn about the possible effects of L2 sentences structures on learners’ performance. This will be the focus of the following subsection.
5.3.2 The Grammaticality Judgment Task: Null Subject Acceptance in