Chapter 1 the study of collocations

5.2.1.1  Between-Group Differences 
 
 
The students in Group 3 performed with the greatest accuracy in the 
translation test.  The results showed that Type 1. Noun Prep, 5. Adjective Prep, 14. 
SVV-ing, and 27. Verb Noun (creation) collocations were translated significantly 
more accurately by higher proficiency subjects. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18 

---

 
The results for Type 1. Noun Prep reflect to some extent the treatment of this 
collocation type in the subjects' textbooks: 76 tokens in TWE1, 80 tokens in TWE2 
and 145 tokens in TWE3.  The use of Type 27. Verb Noun (creation) collocations also 
increases in the textbooks in a pattern similar to the one found in the results of the 
translation test, i.e. as level increases the number of collocations found in the 
textbooks increases.  The results for Type 5. Adjective Prep collocations show a 
significant difference in accuracy only between Group 1 and Group 3, with the 
highest proficiency subjects, Group 3, were more accurate than the others.  A 
similar pattern is also found in the students' textbooks.  Although Type 14. SVV-
ing collocations appear more in TWE1, students were able to translate them with 
significantly more accuracy after their level increased.  Finally, Type 11. SV(O) Prep 
O collocations were translated most accurately by subjects in Group 2.  
Collocations of this type were also found more in TWE2 than in the other two 
textbooks. 
 
Implicational scaling analysis was also used for the between-group 
differences.  The coefficient of reproducibility was found to be significant, which 
confirmed the predictive power of the scale.  The coefficient of scalability for the 
translation data approached significance (C
scalability
 = .57).  The smaller number of 
collocation types included in the translation test could have contributed to the high 
coefficient of scalability.  This result also indicates that a small number of items 
and a translation test are more likely to yield strong enough differences in 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19 

---

performance for a set of collocation types to be truly scalable, i.e. implicationally 
ordered. 
 
Considering the implicational scale for the translation data, the following 
order of accuracy was found after the 80% criterion of acquisition was applied to 
the data (types at the top of the order were more accurately translated than types at 
the bottom):