Comparisons between Frequent V-/A-N Collocations in the Two

4.1.3.1 Comparison between All Collocations Identified in the Two Corpora

In Section 4.1.1, the published authors were found employing more types of A-N collocations than V-N collocations, and in Section 4.1.2 the learners were also found exhibiting a similar behaviors of producing more types of A-N collocations than V-N collocations. Based on these findings, it seems that the learners’ usage pattern of V-/A-N collocations resembled to the published authors’. However, a closer examination on the numbers of types, ANFs, and TTRs of the two writer groups’ use of the identified V-/A-N collocations yielded some discrepancies. Table 4.4 presents the number of types, ANFs, and TTRs of V-N and A-N collocations in RAC and MTC.

Table 4.4. Comparisons between Average Normalized Frequency and Token-Type Ratio of V-/A-N Collocations in RAC and MTC.

RAC MTC

Type ANF TTR Type ANF TTR

V-N 181 15.3 0.55% 203 19.4 0.44%

A-N 248 16.4 0.51% 231 23.5 0.33%

As shown in Table 4.4, while both of the two writer groups produced more types of A-N collocations in their writing, the learners were found using more types of V-N collocations compared to the published authors. On the contrary, the published authors

5 VMTC= V-N collocations in MTC.

6 AMTC= A-N collocations in MTC.

46

employed more types of A-N collocations than their counterpart. These findings appear to suggest that the published authors’ repertoire of A-N collocations was larger the learners’, whereas the learners possessed larger repertoire of V-N collocations.

Nevertheless, the TTRs of these identified V-N and A-N collocations in the two corpora offered findings partially against the claims.

First, comparison of the TTRs of V-N collocations in the two corpora revealed that the learners’ lexical diversity of V-N collocations was more limited than the published authors’. As presented in Table 4.3, TTRVRAC was 0.11% higher than TTRVMTC, which clearly proved that the published authors had a wider repertoire of V-N collocations.

Even though more types of V-N collocations were identified in the learners’ writing, the learners’ vocabulary size of academic V-N collocations was comparatively limited as compared to the published authors’ when frequency of occurrences were considered. In contrast, the higher TTRARAC as compared to TTRAMTC revealed that the published authors’ lexical diversity of A-N collocations was indeed wider than the learners’.

In addition to the differences between the TTRs, some differences were also observed between the ANFs of V-N and A-N collocations in the two corpora.

Comparisons between the four TTRs yielded the following findings. First, the learners averagely employed the V-N and A-N collocations more frequently as compared to the published authors, which was proven by the higher ANFVMTC (19.4 times PMWs) and ANFAMTC (23.5 times PMWs). Secondly, the learners’ average use of A-N collocations was also more frequent as compared to their use of V-N collocations, and this within-group difference in the learner writer within-group was also distinctly larger than the difference in the published author group. This discrepancy indicated that the learners tended to rely on the identified A-N collocations more in their writing.

The above findings suggested that the published authors demonstrated more

47

balanced knowledge of V-N and A-N collocations in their writing, as revealed by the approximate ANFs and TTRs of collocations. Compared to the published authors, the leaners varied their use of both V-N and A-N collocations to a lesser degree, and their use of A-N collocations was even less diverse than their use of V-N collocations. These findings suggested that the published authors’ lexical repertoire of both V-N and A-N collocations was richer as compared to the Taiwanese EFL learners’.

These findings, however, were obtained based on the analysis of the frequency data of all the identified V-N and A-N collocations in the two corpora, which might be greatly influenced by the extreme values of some high-frequency items. In fact, the researcher indeed identified some V-N and A-N collocations appearing more than 100 times PMWs in the corpora. To examine whether these high-frequency items could influence the previously reported results, the researchers carried out a further analysis by excluding items occurring more than 100 times PMWs in the two corpora and recalculated frequency data of the remaining V-N and A-N collocations. Table 4.5 below presents the numbers of types/tokens, ANFs, and TTRs of these collocations in RAC and MTC.

Table 4.5. Comparisons between Average Normalized Frequency and Token-Type Ratio of V-/A-N Collocations Occurring Less than 100 Times PMWs in RAC and

MTC.

RAC MTC

Type Token ANF TTR Type Token ANF TTR V-N 180 31,282 14.7 0.58% 198 37,570 16.2 0.53%

A-N 243 38,202 13.3 0.64% 222 38,107 14.6 0.58%

Analysis of the ANFs as well as TTRs of the V-N and A-N collocations occurring less than 100 times PMWs yielded findings quite different from the previously reported

48

ones. While the TTR of V-N collocations in RAC was still higher than the TTR in MTC, the difference between the two TTRs dropped to 0.05% as compared to the difference when high-frequency V-N collocations were included (i.e. 0.11%). A decrease of difference was also found between the TTR of A-N collocations in RAC and the TTR in MTC, which was only 0.06% and three times lower than the original difference (i.e.

0.18%). In addition to TTRs, the four ANFs also underwent a dramatic decrease in their values as the high-frequency items were excluded. As shown in Table 4.4, the ANF of V-N collocations in RAC was 1.5 times PMWs lower than the ANF in MTC, while the original ANF of A-N collocations in RAC (i.e. 15.3 PMWs) was 4.1 times lower than the original ANF in MTC (i.e. 19.4 PMWs). As for the ANF of A-N collocations in RAC, it was 1.3 times lower than the ANF in MTC, whereas as the original ANF of A-N collocations in RAC (i.e. 16.4 PMWs) was 7.1 times lower than the original AA-NF in MTC (i.e. 23.5 PMWs). In summary, the exclusion of high-frequency V-N and A-N collocations from frequency analysis generally caused decreases in both the differences between the four TTRs and the four ANFs. These findings in fact indicated that the Taiwanese EFL learners’ over-reliance of some high-frequency V-N and A-N collocations in their writing, which contributed to greater differences in the collocation usage pattern between the learners and the published authors.

4.1.3.2 Comparison between Collocations Overlapped in the Two Corpora

In the previous section, the frequent V-N and A-N collocations in the published authors’ and in the Taiwanese EFL learners’ writing were identified and compared.

Comparison of the two writer groups’ collocation use revealed that some items were mutually shared in the two corpora (i.e. overlapped). It is assumed that comparison of these overlapped collocations in the two corpora might offer more information regarding the two writer groups’ usage pattern of collocations in academic writing,

49

which could hence help the researcher to have a better understanding of the discrepancy between the two groups. The researcher thus extracted these overlapped collocations, and further explored the frequency of occurrences of these collocations in the two corpora respectively (see Appendix C for the list of frequent as well as overlapped V-N and A-N collocations). Results of the exploration were provided in Table 4.6.

Table 4.6. Overview of the Frequency of Overlapped V-/A-N Collocations in RAC and MTC.

Type RAC MTC

RF NF ANF RF NF ANF

V-N 136 28,870 2,434.6 17.9 38,710 3,302.1 24.3

A-N 159 40,321 3,400.3 21.4 54,798 4,674.4 29.4

Note. RF=Raw Frequency, NF=Normalized Frequency, ANF=Average Normalized Frequency (per collocation type), TTR=Type-Token Ration.

Cross-comparison of the V-N and A-N collocations in RAC and in MTC identified 136 types of V-N and 159 types of A-N overlapped in the two corpora. Observation on ANFs of these collocations in the two corpora revealed several deviations between the published authors’ and the Taiwanese EFL learners’ use of these overlapped collocations. First, the ANFs of overlapped V-N and A-N collocations in MTC were both larger than those in RAC. The ANF of overlapped V-N collocations in MTC was 6.4 times PMWs higher than that in RAC, and the ANF of overlapped A-N collocations in MTC was 8 times PMWs higher than that in RAC. These findings showed that the learners employed the overlapped V-N collocations and A-N collocations averagely more often than the published authors, which might be attributed to the leaners’ over-reliance of these overlapped collocations in their writing. In addition to the between-group differences, a within-between-group difference of the learners ANFs of the overlapped

50

N collocations and A-N collocations was also observed. ANFOAMTC7 was 5.1 times PMWs higher than ANFOVMTC8, whereas the difference between ANFOARAC9 and ANFOVRAC10 was only 3.5 times PMWs. The greater difference between the ANFs of overlapped V-N and A-N collocations in MTC might indicate that, compared to the use of overlapped V-N collocations, the learners relied on these overlapped A-N collocations more often in their academic writing.

Again, the above findings were obtained based on the analysis of all the overlapped V-N and A-N collocations; in other words, high-frequency items were also included and might influence the research outcomes in some way. To avoid the possible influence of high-frequency items, the researcher also conducted another analysis by excluding high-frequent items from the overlapped V-N and A-N collocations. Table 4.7 gives the frequency data of the remaining overlapped collocations in the two corpora.

Table 4.7. Overview of the Frequency of Overlapped V-/A-N Collocations Occurring Less than 100 Times PMWs in RAC and MTC.

Type RAC MTC

RF NF ANF RF NF ANF

V-N 131 24,438 2,060.9 15.7 30,043 2,562.7 19.6

A-N 150 26,997 2,276.1 15.2 29,262 2,496.1 16.6

Note. RF=Raw Frequency, NF=Normalized Frequency, ANF=Average Normalized Frequency (per collocation type), TTR=Type-Token Ration.

As revealed in Table 4.7, excluding high-frequency items from the analysis yielded observable decreases of every figure, and led to a smaller difference not only between the ANFs of overlapped V-N collocations in RAC and in MTC but also the ANFs of

7 ANFOAMTC=averaged normalized frequency of overlapped A-N collocations in MTC.

8 ANF_OVMTC=averaged normalized frequency of overlapped V-N collocations in MTC.

9 ANFOARAC=averaged normalized frequency of overlapped A-N collocations in RAC.

10 ANFOVRAC=averaged normalized frequency of overlapped V-N collocations in RAC.

51

overlapped A-N collocations. The difference between original ANF of V-N collocations in RAC and in MTC was 6.4 times PMWs, but the difference between the two modified ANFs dropped to 3.6 times PMWs. A more dramatic decrease was observed in the difference between the modified ANFs of A-N collocations. The ANF of A-N collocations in RAC was only 1.4 time PMWs lower than that in MTC, yet the original difference between the two ANFs was 8.0 times PMWs. Results reported here again demonstrated the powerful influence of the high-frequency collocations, especially high-frequency A-N collocations, on enlarging the differences between the collocation use of the published authors and the Taiwanese EFL learners.

Although excluding high-frequency items from the analysis contributed to smaller differences in the use of overlapped collocations between the two writer groups, comparisons of the original ANFs as well as the modified ANFs still suggested the possibility of the learners’ over-reliance on the overlapped V-N and A-N collocations.

This provisional conclusion, however, could not be substantiated without additional proof. To justify the learners’ over-reliance on these overlapped collocations, log-likelihood tests were applied to determine to what extent the frequency of the overlapped V-N and A-N collocations in MTC was different from that in RAC. Results of the log-likelihood tests, as presented in Table 4.8, showed significant differences between the frequency of not only V-N collocations but also A-N collocations across the two corpora. Findings obtained from the log-likelihood tests thus substantiated the learners’ over-reliance of these collocations in their writing. In other words, they overused these overlapped V-N and A-N collocations as compared to the published authors.

52

Table 4.8. Log-likelihood Test on the Frequency of Overlapped V-/A-N Collocations in RAC and MTC.

Type RAC MTC

LL-value p-value

RF RF

All Overlapped Items V-N 136 28,870 38,710 -1,552.72 0.000 A-N 159 40,321 54,798 -2380.82 0.000 Without High-Frequency Items V-N 131 24,438 30,043 -643.61 0.000 A-N 150 26,997 29,262 -118.99 0.000 Note. RF= Raw Frequency, LL=Log-likelihood Value, Sig.=Significance level (p<.05).

4.1.4 Summary of the Findings on Published Authors’ and Taiwanese EFL