Chapter 1 the study of collocations

Table 38.  Nemenyi's multiple comparisons tests of mean rank differences for 
the translation data- Group 3 
Types 
1 5  11  14  16 27 13 
1. Noun Prep  
--- 
1.52* 
1.33* 
.056 
.8 
.789 
.972 
5. Adj Prep  
 
--- 
.189 
1.47* 
.728 
.739 
2.5* 
11. SV(O) Prep O  
   
--- 
1.28* 
.539 
.55 
2.31* 
14. SVV-ing  
 
 
 
--- 
.744 
.733 
1.028 
16. SVO Inf 
 
    
--- 
.011 
1.77* 
27. Verb Noun(creat) 
 
     
--- 
1.76* 
13. SV Inf 
 
      
--- 
*: Significant at the .05 level 
 
290

---

 
4.2.2.2.4  Implicational Scaling for the Translation Data by Groups 
 
As for the implicational scaling analysis of the translation data in the 
first hypothesis, each subject was coded as having translated accurately (1), or 
not (0), each of the types of collocations in the translation test according to the 
80% accuracy criterion.  The two axes of the matrix for the implicational scaling 
consisted of the collocation types ranked from most accurately translated by all 
subjects in each group to least accurately translated, and the  subjects in each 
group ranked in order of their accuracy of translation of all types of 
collocations, from  subjects translating accurately the most types to subjects 
translating accurately the fewest.  The matrix for each group is given in 
Appendix N and summarised in Figures 12, 13, and 14 below. 
 
For Group 1, the coefficient of reproducibility was .936, and the 
coefficient of scalability was .632.  The implicational scale for this set of data 
was found to be significant and the items on the scale are scalable (see 
Andersen, 1978; Hatch & Lazaraton 1991:212).  The implicational scale (Figure 
12) also shows that the first two items on the scale are the two types of 
collocation that were found to be tranlated significantly more accurately than 
all the other types in the post-hoc analyses (see Table 36). 
 
For Group 2, the coefficient of reproducibility was .97 and the coefficient 
of scalability was .78.  As for Group 1, the implicational scale for this set of data 
was found to be significant and the items on the scale scalable.  The 
implicational scale (Figure 13) also shows that the first item on the scale is Type 
 
291

---

11. SV(O) Prep O which was also found to be translated significantly more 
accurately than all the other types in the post-hoc analyses (see Table 37). 
 
For Group 3, the coefficient of reproducibility was .89, and the coefficient 
of scalability was .59.  Both coefficients are approaching significance and it can 
be conlcuded that the impicational scale for Group 3 is valid.  The implicational 
scale (Figure 14) shows that the first three items on the scale are the ones that 
were found to be translated significantly more accurately than all the other 
types according to the post-hoc analyses (see Table 38). 
 
0
.1
.2
.3
.4
.5
.6
.7
.8