4.2 Underused and Overused V-N and A-N Collocations in Taiwanese EFL
4.2.2 T-score and MI-value Analysis of the Underused/Overused
behaviors of V-N and A-N collocations were related to the association strength of the combinations, the t-scores and MI-values of individual underused and overused collocations were obtained. Because this study considered that the published authors’
writing serves as a model for the learners to emulate, t-scores and MI-values of these underused and overused collocations in RAC were calculated for analysis. These collocations were then further categorized into different bands of t-score and MI-value, as follows:
t<2; t=2-3.99; t=4-5.99; t=6-7.99; t=8-9.99; t=10-14.99; t=15-19.99; t=20-
MI<3; MI=3-3.99; MI=4-4.99; MI=5-5.99; MI=6-6.99; MI=7-7.99; MI=8-8.99; MI=9-
The researcher then grouped the underused and overused V-N and A-N collocations into different t-score and MI-value groups, and calculated their number of types/tokens and corresponding percentage of all the V-N/A-N collocations in MTC to examine whether t-score/MI-value were the factors influencing the learners’ underuse/overuse behaviors.
4.2.2.1 T-score and MI-value Analysis of the Underused Collocations
The collocations underused by the Taiwanese EFL leaners were first investigated.
Results regarding the number of types, the number of tokens, and the percentage of tokens of underused V-N and A-N collocations in different t-score bands were presented in Table 4.10.
57
Table 4.10. Overview of Underused V-N and A-N Collocations Found at Different Levels of t-score.
t-score
Underused V-N Underused A-N
Type Token Percentage Type Token Percentage
t<2 0 0 0.0% 0 0 0.0%
t=2-3.99 0 0 0.0% 0 0 0.0%
t=4-5.99 2 43 0.1% 2 40 0.1%
t=6-7.99 12 322 0.7% 20 441 0.7%
t=8-9.99 19 775 1.6% 48 1,414 2.1%
t=10-14.99 21 1,851 3.9% 29 2,008 3.0%
t=15-19.99 9 1,672 3.5% 12 1,835 2.8%
t=20- 4 1,606 3.4% 10 3,954 6.0%
Total 67 6,269 13.2% 121 9,692 14.7%
As shown in Table 4.10, most of the underused V-N collocations were found in the band of t=10-14.99, and no underused V-N collocations were found in the bands of t<2 and t=2-3.99. A similar situation was also observed regarding the underused A-N collocations, that is, no underused A-N collocations fell into the first two bands of t-score (i.e. t<2 and t=2-3.99). The largest group of underused A-N collocations, however, was in the band of t=20-, rather than in the band of t=10-14.99. A difference like this might suggest that A-N collocations with high t-score were more likely to be underused than V-N collocations in the learners’ writing. Nevertheless, the proportional difference between the underused V-N and A-N collocations would be smaller if different t-score bands were regrouped into bigger units, namely bands of low score (t<2), medium t-score (t=2-9.99), and high t-t-score (t>10-). Figure 4.2 illustrates the regrouped percentage of underused V-N and A-N collocations in different t-score bands in a bar chart.
58
Figure 4.2. Regrouped Percentage of Underused V-N and A-N collocations Found at Different Levels of t-score.
Results of the regrouping revealed that both underused V-N and A-N collocations were more likely to be those with high t-scores (10.8% and 11.8% respectively), such as complete task score=24.03), ask participant score=23.23), previous research (t-score=29.38), and further research (t-score=24.64), and V-N and A-N collocations with low t-scores were almost never underused by the learners (0% for both).
0.0% 2.4% 10.8%
0.0% 2.9% 11.8%
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
12.0%
14.0%
T < 2 T = 2 - 9 . 9 9 T > 1 0 . 0 0
% OF TOTAL V-N/A-N TOKENS IN MTC
UVN UAN
59
In addition to t-score, the MI-value of each underused V-N and A-N collocation was also examined. Results of both underused V-N and A-N collocations found at different MI-value levels were given in Table 4.11.
Table 4.11. Overview of Underused V-N and A-N Collocations Found at Different Levels of MI-value. formed the largest group among all the other MI-values, and the second and the third largest groups were MI=3-3.99 and MI=4-4.99, respectively. As for underused A-N collocations, most of the underused A-N collocations were in the band of MI=6-6.99, and the second and third largest groups were MI=5-5.99 and MI=7-7.99. Regrouping these different bands of MI-value into larger unit, namely bands of low MI-value (MI<3), medium MI-value (MI=3-6.99), and high MI-value (MI>7-), revealed that both underused V-N and A-N collocations were mostly in the medium MI-value groups, accounting for 11.4% and 10.6% respectively, and that very few underused V-N and A-N collocations were in the low MI-value group (0.7% of V-A-N and 0.6% of A-A-N). A proportional difference between underused V-N and A-N collocations, however, was obtained in the percentage of V-N collocations and that of A-N collocations in the band
60
of high MI-value (MI=7-). For the high MI-value group, the underused A-N collocations, such as obligatory context (MI=7.83) and tertiary level (MI=7.30), accounted for 3.5%, whereas the underused V-N collocation (e.g. corroborate finding (MI=8.05) and guide study (MI=8.45)) accounted for only 0.9%. This difference might suggest that A-N collocations with higher MI-values were more likely to be underused than the V-N collocations. Figure 4.3 below illustrates the regrouped percentage in a bar chart.
Figure 4.3. Regrouped Percentage of Underused V-N and A-N Collocations Found at Different Levels of MI-value.
0.7% 11.4% 0.9%
0.6% 10.6% 3.5%
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
12.0%
M I < 3 M I = 3 - 6 . 9 9 M I > 7
% OF TOTAL V-N/A-N TOKENS IN MTC
UVN UAN
61
4.2.2.2 T-score and MI-value Analysis of the Overused Collocations
For overused collocations, analysis of combinations in different bands of t-score and MI-value was also conducted. Table 4.12 below presents the number of types, the number of tokens, and the percentage of V-N and A-N collocations in different t-score levels.
Table 4.12. Overview of Overused V-N and A-N Collocations Found at Different Levels of t-score.
t-score
Overused V-N Overused A-N
Type Token Percentage Type Token Percentage
t<2 6 1,145 2.4% 5 465 0.7%
t=2-3.99 12 1,237 2.6% 16 2,171 3.3%
t=4-5.99 33 3,878 8.1% 28 3,318 5.0%
t=6-7.99 27 4,352 9.1% 32 4,528 6.8%
t=8-9.99 17 2,908 6.1% 21 3,339 5.0%
t=10-14.99 16 6,920 14.4% 13 3,721 5.6%
t=15-19.99 3 2,030 4.2% 6 4,038 6.1%
t=20- 7 10,580 22.1% 10 25,910 39.2%
Total 121 33,050 69.0% 131 47,490 71.8%
Results of the t-score analysis demonstrated that, similar to what had been reported on the underused collocations, both overused V-N and A-N collocations in the band of t=20- formed the largest groups among all the other bands, accounting for 22.1% and 39.2% respectively. Regrouping of all the bands into bigger units (i.e. low t-score, medium t-score, and high t-score), as illustrated in Figure 4.4, revealed that both overused V-N and A-N collocations in the high t-score group (t=10-) ranked the largest, taking up 40.7% and 50.9% respectively. Collocations in the low t-score group (t<2), however, were the smallest ones among all the other groups. In other words, overused V-N and A-N collocations were more likely to be those with higher t-scores, such as
62
answer question (t-score=29.30) and current study (t-score=38.67), whereas those with low t-scores were less likely to be overused.
Figure 4.4. Regrouped Percentage of Overused V-N and A-N Collocations Found at Different Levels of t-score.
2.4% 25.9% 40.7%
0.7% 20.1% 50.9%
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
T < 2 T = 3 - 9 . 9 9 T > 1 0
% OF TOTAL V-N/A-N TOKENS IN MTC
OVN OAN
63
T-score analysis of the overused V-N and A-N collocations indicates that the learners exhibited a similar pattern of overusing both V-N and A-N collocations with higher t-scores. Results of the MI-value analysis, however, suggests a somewhat different usage pattern of overusing V-N and A-N collocations. Table 4.13 presents the number of types, the number of tokens, and the percentage of overused V-N and A-N collocations at different levels of MI-value.
Table 4.13. Overview of Overused V-N and A-N Collocations Found at Different Levels of MI-value.
MI-value
Overused V-N Overused A-N
Type Token Percentage Type Token Percentage
MI<3 35 9,718 20.3% 26 3,338 5.0%
MI=3-3.99 32 5,395 11.3% 24 2,877 4.4%
MI=4-4.99 25 5,786 12.1% 26 8,188 12.4%
MI=5-5.99 20 5,978 12.5% 24 4,728 7.2%
MI=6-6.99 6 1,944 4.1% 20 13,027 19.7%
MI=7-7.99 1 307 0.6% 10 14,625 22.1%
MI=8-8.99 1 1,855 3.9% 1 707 1.1%
MI=9- 1 2,067 4.3% 0 0 0.0%
Total 121 33,050 69.0% 131 47,490 71.8%
As can be seen in Table 4.13, many of the overused A-N collocations were found in the band of MI=7-7.99, and the second largest group was in the band of MI=6-6.99 (19.7%). In contrast, the largest group of overused V-N collocations was in the band of MI<3, taking up 20.3%. However, regrouping these bands into bigger units (i.e. low MI-value, medium MI-value, and high MI-value) revealed a different situation, as illustrated in Figure 4.5.
64
Figure 4.5. Regrouped Percentage of Overused V-N and A-N Collocations Found at Different Levels of MI-value.
Overused A-N collocations were mostly found in the medium MI-value group, accounting for 43.7%. Overused A-N collocations in the high MI-value group took up 23.3%, ranking the second largest group, while those in the low MI-value group only accounted for 5.0%. As for overused V-N collocations, those in the medium MI-value group also formed the largest group, accounting for 40%. However, while overused A-N collocations in the low MI-value band formed the smallest group, overused V-A-N collocations in the low MI-value group was the second largest group, taking up 20.3%, and those with high MI-value only accounted for 8.8%. These findings suggest that, while both V-N and A-N collocations in the medium MI-value bands were tended to be overused, V-N collocations with low MI-values (e.g. make conclusion (MI=2.18) and require student (MI=2.25)) were more likely to be overused as compared to those with high MI-values.
65
4.2.2.3 T-score and MI-value Analysis on Synonymous Pairs of Underused and Overused Collocations
Exploration of the V-N and A-N collocations underused/overused by the Taiwanese EFL learners identified several synonymous pairs, in which one (or more) collocation(s) was underused while other(s) was overused. These synonymous pairs of underused/overused collocations were then discussed in details in the current section.
To uncover the possible cause(s) of underusing one but overusing another, the t-score and MI-value of these synonymous collocation pairs were closely examined. In addition, detailed analysis on the concordance lines of each collocations was also undertaken to examine whether there were other factors in the learners’ underuse/overuse behaviors.
Important Difference verses Significant Difference
In this pair, the underuse of the collocation important difference could be attributed to its less robustness features (i.e. frequent and highly-associated feature). As shown in Table 4.14, the t-score as well as the MI-value of the underused collocation important difference was observably lower than the t-score and MI-value of the overused collocation significant difference.
Table 4.14. t-scores and MI-values of ‘Important Difference’ and ‘Significant difference’.
Underused Overused
important difference significant difference
t-score of collocation 9.39 48.98
MI-value of collocation 3.74 7.79
In addition, detailed examination on these two collocations also revealed a functional difference between their usage patterns, which might be another cause of the underuse of important difference. In the two corpora, both of the two collocations
66
could be used to highlight the high degree of differences between findings, as illustrated in (1) to (4):
(1) We also find an important difference in the frequency of the sub-step that describes the work done (SS2A) in both corpora. (ESP_3001004017, RAC) (2) Thus, the transcriptions revealed important differences between the two groups
in the depth of word-processing and the strength of inferencing. (NCUE_06003, MTC)
(3) As can be seen in Table 14, no significant differences were found on any of the tests, and all three effect sizes were negligible, apparently indicating that lexicalization had no lasting effect. (AL_3105693713, RAC)
(4) An ANOVA did not yield significant differences between the experimental group and the control group. (NCCUC_07001, MTC)
In addition to describing findings, the collocation important difference were also employed to describe the big difference(s) between research methods in the published authors’ writing, as illustrated in (5) and (6):
(5) However, there is an important methodological difference between Silverberg and Samuel's study and ours, as they used a semantic priming paradigm and a lexical decision task, whereas we employed a translation recognition task.
(LL_5604571608, RAC)
(6) This discrepancy between findings is probably attributable to several important methodological differences between the studies. (TMLJ_9303399417, RAC)
On the contrary, the same usage pattern of important difference in describing methodological differences was not found in the Taiwanese EFL learners’ writing. In other words, the learners only used this collocation to highlight differences in research findings but not in other aspects. It is possible that the learners were not aware of using important difference to highlight the differences between research methods and they thus underused this construction in their writing.
67
As for the overuse of significant difference, in addition to the robustness features of this construction, the genre differences (i.e. differences between master’s thesis and research articles) might also play a part in the learners’ overuse. In theses, learners would usually report their findings in details, which increased the possibility of using significant difference to describe the statistical differences in their writing. On the contrary, authors of published journal articles were often required to condense their writing because of word limits, and these authors thus used significant difference less frequently as compared to the learners.
Early Study verses Previous Study
Both early study and previous study were mainly employed to denote past research, as illustrated in (7) to (10).
(7) Earlier studies have found that the phonological paragraph appears to share similarities with the orthographic paragraph (Yule, 1980), in that both represent a larger scale chunking of information into sequential-hierarchical segments.
(JEAP_0201005020, RAC)
(8) Earlier studies show that Chinese regard giving direct or hedged advice as appropriate ways to build relationships and group belongings with others (Hall
& Ames, 1987; Hu & Grove, 1991; Jin & Cortazzi, 1993; Hinkel, 1994).
(NCCUC_03004, MTC)
(9) Previous SLA studies of recasts tend to view recasts as a kind of 'corrective feedback' and investigated their effectiveness for the learners' interlanguage development. (AL_2702195219, RAC)
(10) Previous studies have applied ZPD to develop instruction and learning activities. (NCCUC_07015, MTC)
In terms of function, no differences in usage pattern were found in these two collocations. The learners, however, still overused previous study but underused early study in their writing. A possible cause of this phenomenon was the more robustness of one of the collocations. In this pair, the overused collocation previous study was
68
considered more robust than the underused collocation early study, in which the t-score and MI-value of the overused combination were strikingly higher than its underused counterpart (see Table 4.15).
Table 4.15. t-scores and MI-values of the Synonymous A-N Collocation Pair: ‘Early Study’ and ‘Previous Study’.
Underused Overused
early study previous study
t-score of collocation 16.64 34.79
MI-value of collocation 4.74 6.30
In this case, the overused collocation was considered more frequent (i.e. higher t-scores) and more strongly-associated (i.e. higher MI-values) than the underused collocation. It is arguably to conclude that the Taiwanese EFL learners overused previous collocations because of their robustness as frequent A-N collocation in the published authors’ writing, which corroborates with the findings reported in Li and Schmitt (2010).
Confirm/Corroborate Findings verses Support Findings
In addition to the robustness of a collocation, the higher level of t-scores was also found to be an influential factor causing the overuse of one collocation but not the other.
In the synonymous pair of ‘confirm/corroborate findings’ and ‘support findings’, all of these three constructions were employed to show the findings reported in one research is (or is not) consistent with those in another, as illustrated in (11) to (16):
(11) With respect to native speakers, the present study further confirms previous findings that idioms are read faster than novel language. (SLR_2702251272, RAC)
(12) The result also confirmed findings of Ho's and Huang‘s (2002) studies.
(NCCUC_05005, MTC)
69
(13) This result also strongly corroborates the findings in previous empirical studies (Cheung, 2002; Lasagabaster & Sierra, 2005; Lipovsky & Mahboob, 2010; Ma, 2009; Mahboob, 2004; Medgyes, 1992; Pacek, 2005). (TESOLQ_4602280305, RAC)
(14) The results corroborate the findings of previous studies attempting to explore learners' attributional beliefs regarding success and failure in learning a second or foreign language (Graham, 2003; Tse, 2000; Williams & Burden, 1999;
Williams, Burden, & Al-Baharna, 2001). (NCUE_04004, MTC)
(15) On the other hand, the current results support the findings of earlier studies that the diglot reader would reduce the affective barrier in studying a second language. (CALL_2001067077, RAC)
(16) This result supports the findings of F. J. Kuo's (2004) and W. Y. Lee's (2006) studies which controlled for stem words and derived words. (NCUE_09010, MTC)
The construction support findings, however, was still overused by the learners.
One possible explanation for this overuse was due to the high t-score of this construction. As shown in Table 4.16, the t-score of the overused combination ‘support findings’ was higher than the two underused counterparts (i.e. ‘confirm findings’ and
‘corroborate findings’), whereas the MI-values of the two underused collocations were both slightly higher than the MI-value of the overused one. It is thus possible that the Taiwanese EFL learners overused support findings because of its higher t-score as compared to the other two synonymous constructions.
Table 4.16. t-scores and MI-values of the Synonymous V-N Collocation Pair:
‘Confirm/Corroborate Findings’ and ‘Support Findings’.
Underused Overused
confirm findings corroborate findings support findings
t-score of collocation 9.14 7.65 10.24
MI-value of collocation 6.08 8.05 5.42
70
In addition to the different levels of t-score, further analysis on the usage pattern of support findings in the two corpora also revealed some differences in the use of this construction between the published authors and the Taiwanese EFL learners. The learners in the current study were found frequently using the structure of to support findings in their writing to show the intention of applying one evidence/method to test the validity of their previously reported findings, as illustrated in (18) and (19):
(17) To support these findings, the TM and the FBM responses recorded from the interview are translated from Chinese into English and extracted as follows.
(NKNU_10006, MTC)
(18) Finally, some subjects' answers were extracted to support the study findings.
(NPUE_06006, MTC)
This usage pattern was found occurring 31 times in the learners’ writing, whereas the same usage pattern only 11 times in the published authors’ writing. The research thus conducted a further comparison between the raw frequency of support findings in RAC and that in MTC after excluding the occurrences of to support findings in the two corpora, and the log-likelihood test on the two modified frequencies revealed no significant difference between two (RAC:99 times, MTC: 112 times, LL:0.96, Sig.:0.328). Based on these findings, it is also plausible that the learners’ overuse of support findings was attributed to their preference of using the structure to support findings in their writing.
Undertake Study verses Conduct Study
Both of the two V-N collocations were used to denote the action of do research in the two corpora, as illustrated in (29) to (32):
71
(19) With the intention of shedding further light on the formulation process, the present study was undertaken as an attempt to deepen our knowledge of the way advanced EFL writers search for lexical units when generating their L2 texts, to analyze the role their L1 plays in those searches, and to ascertain whether these processes vary as a function of task difficulty. (JSLW_1902061081, RAC) (20) Trahey and White (1993) undertook a study to explore the effects of input flood
on the learning of English adverb placement by young French-speaking ESL learners aged between 10 and 12. (NCUE_11001, MTC)
(21) The study was conducted in an 18-week EFL writing course offered for the second-year English-major students at a comprehensive Chinese university in Beijing in 2007. (LTR_1404464484, RAC)
(22) Few studies have been conducted to examine the transferability to oral proficiency development. (NTHU_09003, MTC)
In the learners’ writing, however, the construction undertake study was underused while conduct study was overused. Analysis on the t-scores as well as MI-values of the two V-N collocations suggest that the levels of t-score might be one of the main factors influencing the learners’ overuse and underused behaviors.
Table 4.17. t-scores and MI-values of the Synonymous V-N Collocation Pair:
‘Undertake Study’ and ‘Conduct Study’.
Underused Overused
undertake study conduct study
t-score of collocation 8.02 24.84
MI-value of collocation 5.21 5.52
As shown in Table 4.17, the t-score of the overused collocation (i.e. conduct study) was three times higher than the t-score of the underused collocation (i.e. undertake study). As for the MI-values, the difference between the overused and the underused collocations was not obvious. Based on the findings of the two association measures, it is possible that the learners overused conduct study because of the construction’s strikingly high t-score.
72
In addition, the degree of semantic transparency of the two verb collocates might also play a part in the learners’ overuse and underuse behaviors. The research again consulted the online LDCE of the two collocates, and results of the consultation revealed that the verb collocate conduct might be more semantically transparent as compared to another verb collocate undertake. In LDCE, the meaning of conduct is to carry out a particular activity or process, especially in order to get information or prove facts, which was easy for the learners to comprehend. In contrast, in LDCE the meaning of undertake is stated as to accept that you are responsible for a piece of work, and start to do it, which might be more complex for the learners to comprehend and utilize. The lower semantic transparency of undertake thus caused the learners to avoid collocating this verb with study and led to its underuse in the learners’ writing.
Further Research verses Future Research
Both of the published authors and Taiwanese EFL learners employed further research and future research to denote research done in near future, as illustrated in (33)
Both of the published authors and Taiwanese EFL learners employed further research and future research to denote research done in near future, as illustrated in (33)