Chapter 1 the study of collocations

Table 32.  Blank filling accuracy distinguishing proficiency levels 
Group 1 
Group 2 
Group 3 
Collocation Types 
Collocation Types 
Collocation Types 
34. Noun Noun** 23. 
SV(O) Adverbial* 1. 
Noun Prep** 
5. Adjective Prep*  
11. 
SV(O) Prep O** 
  
24. 
SV(O) wh-word** 
  
27. 
Verb Noun (creat) ** 
  
33. 
Verb Adverb** 
  
36. 
Prep Det Noun** 
  
37. Phrasal Verb** 
 
276

---

  
23. 
SV(O) Adverbial* 
 
 
5. Adjective Prep* 
**: Significantly more accurate than the other two groups 
*: Significantly more accurate than one other group 
 
4.2.1.3.3  Implicational Scaling for the Blank Filling Data (All Groups) 
 
For the implicational scaling analysis the Guttman procedure was used.  
Each subject was coded as having answered accurately (1), or not (0), each of 
the 11 types of collocations repeated across groups in the blank filling test.  As 
with the translation data, an 80% accuracy criterion was used for the coding of 
the data, i.e. accuracy less than 80% was coded as 0, accuracy 80 to 100% was 
coded as 1.  The two axes of the matrix for the implicational scaling consisted of 
the eleven types ranked from most accurately answered by all subjects to the 
least accurately answered, and the 275 subjects ranked in order of their 
accuracy of response to all types of collocations, from types most accurately 
answered to types least accurately answered.  This matrix is summarised in 
Figure 8 below.  The coefficient of reproducibility for this analysis was .91.  The 
coefficient of scalability was .4.  As with the essay data, even though the 
coefficient of reproducility for the blank filling data is at the level necessary for 
this implicational scale to be considered valid, the coefficient of scalability is 
below the recommended level of .6 (see Andersen 1978; Hatch & Lazaraton 
1991:212). 
 
277

---

0
.0 5
.1
.1 5
.2
.2 5
.3
.3 5
3 4
3 3
2 4
2 3
4
3 6
5
3 7
2 7
1
1 1
 
Col l oca t i on  Ty p es
 
Accura cy     -  Bl a nk   Fi l l i ng  Da t a   -  Al l   Group s