Premium
Cloze Tests Revisited: Exploring Item Characteristics with Special Attention to Scoring Methods
Author(s) -
Kobayashi Miyoko
Publication year - 2002
Publication title -
the modern language journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.486
H-Index - 83
eISSN - 1540-4781
pISSN - 0026-7902
DOI - 10.1111/1540-4781.00162
Subject(s) - cloze test , psychology , context (archaeology) , test (biology) , natural language processing , item response theory , artificial intelligence , linguistics , computer science , psychometrics , developmental psychology , reading comprehension , reading (process) , paleontology , philosophy , biology
This study examined the effects of item characteristics on cloze test performance. A particular focus of the study was to investigate the relationship between a number of cloze item characteristics and scoring methods. The item characteristics examined were: (a) content/function words, (b) parts of speech, (c) word frequency, (d) the number of occurrences of a word in the text, (e) alternative answers, (f) syntactic variation, (g) the amount of context, and (h) knowledge base. A total of 255 Japanese university students participated and were randomly assigned 1 of 4 sets of cloze tests. The data revealed variations in the results of cloze tests attributable to different scoring methods. The study showed the complex effects of item characteristics and scoring methods on test performance, both in terms of item difficulty and discrimination. These findings have important implications for the use of cloze tests.