A Case of the Validity Investigation of Concordance-based Cloze Testing: Construct Relevance Revisited

Abstract

Investigating construct validity is a process that is essential especially for devising a new item type. However, the distinction between construct-relevant and construct-irrelevant variances may not always be sufficient. In this article, the construct of the concordance-based cloze item type (henceforth ConCloze) will be defined through two research projects. The first one is the doctoral dissertation (Kongsuwannakul, 2017 Investigating the Construct Validity of a Concordance-based Cloze Test: A Mixed-methods Study), in which the construct validity of the item type is defined with an iterative research design. Item components and a variety of changes to them are used insofar as information about the language processes and domains is obtained out of the data. The other research project is a follow-up study (Kongsuwannakul, 2019 Suranaree University of Technology Students’ Language Domains in Engaging with a Concordance-based Cloze Test: A Contrastive Approach), in which a contrastive approach is used for identifying distinguishing language domains in ConCloze. It is found that language domains such as knowledge of lexical semantics are not sufficiently distinguishing the construct of the item type. As for theoretical implications, it will be argued that construct-relevant variance could be divided into construct variance and construct-peripheral variance.