1993Language TestingRequires access

What are the characteristics of natural cloze tests?

James Dean Brown

Open publisher page 33 citations

Abstract

This study investigates the characteristics of natural cloze tests. Natural cloze tests are defined here as cloze procedures developed without intercession based on the test developer's knowledge and intuitions about passage difficulty, suitable topics, etc. (i.e., the criteria which are often used to select a cloze passage appropriate for a particular group of students). Fifty reading passages were randomly selected from an American public library. Each passage was made into a 30-item cloze test (every twelfth word deletion). The subjects were 2298 EFL students from 18 colleges and universities in Japan. Each student completed one of the 30-item cloze tests. The 50 cloze tests were randomly administered across all of the subjects so that any variations in statistical characteristics could be assumed to be due to other than sampling differences. The students also took a 10-item cloze test that was common to all students. The 50 cloze tests were compared in terms of descriptive, reliability and validity testing characteristics. The results indicate that natural cloze tests are not necessarily well-centred, reliable and valid. A typical natural cloze is described, but considerable variations were also found in the characteristics of these cloze tests (with many of them having skewed distributions and/or poor reliability). The implications for cloze test construction and use are discussed.

About this research paper

What this paper is about

This study investigates the characteristics of natural cloze tests. Natural cloze tests are defined here as cloze procedures developed without intercession based on the test developer's knowledge and intuitions about passage difficulty, suitable topics, etc. (i.e., the criteria which are often used to select a cloze passage appropriate for a particular group of students). Fifty reading passages were randomly selected from an American public library. Each passage was made into a 30-item cloze test (every twelfth word deletion). The subjects were 2298 EFL students from 18 colleges and universities in Japan. Each student completed one of the 30-item cloze tests. The 50 cloze tests were randomly administered across all of the subjects so that any variations in statistical characteristics could be assumed to be due to other than sampling differences. The students also took a 10-item cloze test that was common to all students. The 50 cloze tests were compared in terms of descriptive, reliability and validity testing characteristics. The results indicate that natural cloze tests are not necessarily well-centred, reliable and valid. A typical natural cloze is described, but considerable variations were also found in the characteristics of these cloze tests (with many of them having skewed distributions and/or poor reliability). The implications for cloze test construction and use are discussed.

Why it matters

OpenAlex reports 33 citations for this work. Citation counts describe recorded attention and do not establish research quality.

Key contribution

A contribution statement is not available in the OpenAlex record.

Method / approach

Method details are not available in the OpenAlex metadata.

Main findings

Findings are not separately available in the OpenAlex metadata.

Limitations

Limitations are not available in the OpenAlex metadata.

Applications

Application details are not available in the OpenAlex metadata.

Available abstract

This study investigates the characteristics of natural cloze tests. Natural cloze tests are defined here as cloze procedures developed without intercession based on the test developer's knowledge and intuitions about passage difficulty, suitable topics, etc. (i.e., the criteria which are often used to select a cloze passage appropriate for a particular group of students). Fifty reading passages were randomly selected from an American public library. Each passage was made into a 30-item cloze test (every twelfth word deletion). The subjects were 2298 EFL students from 18 colleges and universities in Japan. Each student completed one of the 30-item cloze tests. The 50 cloze tests were randomly administered across all of the subjects so that any variations in statistical characteristics could be assumed to be due to other than sampling differences. The students also took a 10-item cloze test that was common to all students. The 50 cloze tests were compared in terms of descriptive, reliability and validity testing characteristics. The results indicate that natural cloze tests are not necessarily well-centred, reliable and valid. A typical natural cloze is described, but considerable variations were also found in the characteristics of these cloze tests (with many of them having skewed distributions and/or poor reliability). The implications for cloze test construction and use are discussed.

Key concepts: Cloze test, Psychology, Test (biology), Reading (process), Reliability (semiconductor), Mathematics education, Significant difference, Natural (archaeology)

Related papers

Back to paper searchBrowse research topicsOriginal source
What are the characteristics of natural cloze tests? — Research Paper | ScholarLens