Gender and test item-response formats.

Abstract

The purpose of this study was to develop a better understanding of the patterns of science achievement for 154 ninth-grade girls and boys on multiple-choice and short-answer constructed-response items. The study was guided by a model, developed from an extensive review of the literature, incorporating the dimensions of generalised self-efficacy, item- specific self-efficacy and worry. These variables were operationalised through selected or specifically developed quantitative and/or qualitative research methods, and a series of equivalent multiple-choice and short-answer constructed-response achievement items was constructed for two different unit tests. The participants in the study rated their item-specific self-efficacies on 5-point Likert-type scales immediately before answering each of the achievement items, and they completed a series of worry items from Spielberger's Test Anxiety Inventory halfway through each test. Qualitative data were collected by surveying all the students and by interviewing selected students. The quasi-experimental analyses revealed the absence of any practically important gender-related differences in achievement for the multiple-choice and the constructed-response achievement items. However, the boys reported more item-specific self-efficacy and less worry than the girls for each of these item-response formats, and each of these gender-related differences was judged to be practically significant. The qualitative data provided additional evidence that the girls' self-perceptions of their efficacy for answering multiple-choice and short-answer constructed-response items was lower than that of the boys. It also provided support for the model underpinning the study. Overall, there was no evidence of any practically important interactions between gender and item-response formats, for either item-specific self-efficacy, worry or achievement, indicating that neither of the item-response formats used in the study, with this group of students, advantaged one sex over the other. Additionally, the findings from this study suggested that sufficient time should be allowed during testing so that all students can complete tests to the levels of their capabilities and that, during tests, the influence of students' self-efficacies is mediated through the quality of their engagement with test items.