Chapter 1 the study of collocations

 
Table 26.  Collocational use distinguishing proficiency levels 
 
260

---

Group 1 
Group 2 
Group 3 
Collocation Types 
Collocation Types 
Collocation Types 
26. SV c** 4. 
Prep Noun** 1. 
Noun Prep** 
29. Adjective Noun** 12. 
SV to Inf** 13. 
SV Inf** 
11. SV(O) Prep O* 36. 
Prep Det Noun** 19. 
SV(O) that** 
 37. 
Phrasal Verb** 5. 
Adjective Prep* 
 5. 
Adjective Prep*  
 11. SV(O) Prep O*  
**: Significantly more occurences than the other two groups 
*: Significantly more occurences than one other group 
 
4.2.1.1.3 Implicational Scaling for the Essay Data (All Groups) 
 
For the implicational scaling analysis the Guttman procedure was used.  
When the Guttman analysis reveals that a particular scale is consistently 
interpretable, that is if one item on the scale is statistically consistently more 
difficult than another, which is in turn harder than another, then the scale 
attains a certain predictive power (Davidson 1987).  The coefficient of 
reproducibility, which shows how accurately a subject's performance can be 
predicted from that person's position in the matrix, and the coefficient of 
scalability, which is a single statistic detailing the strength of the items as an 
ordered scale and indicating whether a given set of features is truly scalable 
and unidimensional, were calculated.  The higher the value of the coefficient of 
scalability, the more "implicational" the scale (Davidson 1987). 
 
261

---

 
Each subject was coded as having used (1), or not used (0), each of the 37 
types of collocation in their essays.  The two axes of the matrix for the 
implicational scaling consisted of the 37 items ranked from most commonly 
used by all subjects to least commonly used, and the 275 subjects ranked in 
order of their frequency of use of all types of collocations, from  subjects using 
the most types to subjects using the fewest types.  This matrix is summarised 
for the first 17 types, mean >.1, in Figure 4 below.  The coefficient of 
reproducibility for this analysis was .90.  The coefficient of scalability was .33.  
While the coefficient of reproducibility is at the level necessary for this 
implicational scale to be considered valid (see Andersen 1978), the coefficient of 
scalability is below the recommended level of .6 (Hatch & Lazaraton 1991:212).  
This suggests that while the implicational scale for the essay data is valid, the 
variance in terms of numbers of errors, and the fact that most subjects did not 
use the majority of the scaled collocations, resulted in the low coefficient of 
scalability. 
 
 
262

---

0
.1
.2
.3
.4
.5
.6
.7
.8
.9
1
 
 
Note.  Only those types with a mean > .1 are included in this figure.