Rasch analysis of a congruent and incongruent collocations test

Article appearing in Shiken 20.2 (November, 2016) pp. 25-43.

Authors: Christopher Nicklin & Garrett DeOrio
Nippon Medical School

In order to investigate the hypothesis that collocations might be easier to acquire productively through the use of illustrations due to the pictorial superiority effect (Nelson, Reed, & Walling, 1976), the Congruent and Incongruent Collocations (CIC) test was specifically designed to assess the knowledge of a group of students regarding a group of 15 congruent and 15 incongruent collocational phrases. The CIC test was developed to be administered as a pretest, posttest, and delayed posttest to a group of second year Japanese medical students (N = 109).

The results of the pretest were analysed using the Rasch dichotomous model (Rasch, 1960), which revealed that the CIC test was of an appropriate difficulty level for the students with the majority of the items being well targeted. However, not all of the items fit the expectations of the Rasch model, and a test for invariance showed that the CIC test was not a reliably invariant instrument for assessing a samples knowledge regarding the group of collocations being tested.

Keywords: Rasch analysis, congruent and incongruent collocations, collocations test, pictorial superiority effect, vocabulary

Download full article (PDF)