Premium
The Cloze Test: Or Why Small Isn't Always Beautiful *
Author(s) -
Sciarone A. G.,
Schoorl J. J.
Publication year - 1989
Publication title -
language learning
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.882
H-Index - 103
eISSN - 1467-9922
pISSN - 0023-8333
DOI - 10.1111/j.1467-1770.1989.tb00599.x
Subject(s) - cloze test , parallelism (grammar) , psychology , point (geometry) , test (biology) , indonesian , word (group theory) , linguistics , arithmetic , statistics , mathematics , reading comprehension , reading (process) , paleontology , philosophy , geometry , biology
This article presents the findings of an experiment aimed at determining the number of blanks minimally required to ensure parallelism for cloze tests differing only in the point at which deletion starts. Two 200‐item cloze tests were constructed, both based on the same Dutch text and differing only in that, in the their second halves, deletion in one version lagged one word behind those in the other. The two versions were administered to two groups of 38 and 36 Indonesian learners of Dutch. Analysis of their scores on various subsets of 100, 75, and 50 items revealed that the required minimum depends upon the scoringmethod used. With the exact‐word method, tests should contain a minimum of about 100 blanks; with the acceptable‐word method, a minimum of about 75 blanks will suffice. With tests containing only 50 blanks–the number generally held to be sufficient–parallelism was found to be a matter of pure chance. In an additional experiment, the tests involved were shown to satisfy a major requirement for th e validity of any L2 proficiency tests: administration to two groups of 20 and 19 native speakers of Dutch resulted in high scores, with mean acceptable responses of 190 or more for a total of 200 items.