Item Types And Reading Abilities

Reading is one of the most important abilities for EFL learners. It constitutes a relatively large part of almost all the large-scale tests such as CET 4 and CET6, TOEFL and GRE. Most reading tests, however, exclusively adopt the multiple-choice items proved to be somewhat invalid. Some researchers (Frederiksen 1984, Samson 1983, Bracht & Hopkins 1970, etc.) believe that item types make no or a small difference when they are used to test the same abilities; on the other hand, some other researchers (Wood 1993, Cooper 1984, etc.) argue that different types of tests are appropriate for measuring different abilities or different components of an ability. This is an open question meriting our discussion and research since this research is bound to have implications for language teaching and testing. Little empirical research at home and abroad, however, has been performed up to now. It is high time that due attention was paid to the question. Therefore, the thesis, an empirical study, aims at investigating the relationship between item types and reading abilities.Alderson's validation procedure was employed in the study to assess the overall Validity of the three types of reading test (multiple-choice, completion and short-answer). Four research questions were addressed to guide the study: "Which type of reading test is more valid?" "Is students' reading test performance influenced by the item types?" "Are students with higher reading ability and lower reading ability both influenced by the item types?" and "What reading abilities is each type of reading test appropriate for measuring?" The following data were collected to answer the four research questions above: data for validity, reading test scores, and reading ability scores. In order to collect the data, a reading test with a reading validity questionnaire was administered to 237 Chinese college students, and their reading scores for English course final examination, their scores for another three reading tests and their teachers' ratings of their reading performance were also available for their reading ability scores. Descriptive statistics, one-way ANOVA, two-way ANOVA and Pearson correlation coefficient, were employed to analyze the data.The result of this research indicated that the short-answer type is the most face valid and response valid whereas the completion type is the mostconcurrent valid and predictive valid. The students' reading test performance is actually influenced by the reading test item types, especially the lower-reading-ability students. Completion and short-answer types, especially the latter, are more valid for measuring the ability to comprehend words and phrases. The completion type is the most valid one for measuring the ability to comprehend facts and details. It is not clear, however, what item types are more valid for measuring the ability to comprehend main ideas and gist.Finally, a few suggestions were made for language testing and teaching and future research. Different item types should be adopted on reading tests with different purposes and for testing different reading abilities. Teachers should attempt to adopt item types as many as possible in small-scale tests such as classroom reading tests for measuring reading abilities. Another suggestion is that the reading abilities be further or otherwise classified in future research. The last suggestion is that item type research be further extended to other tests, such as writing, speaking, and listening.