本研究分析2001至2005年度「國民中學學生基本學力測驗」自然科之性別差異和差別試題功能(DIF)。自然科整體表現的性別差異分析結果顯示:全體受試並無顯著的性別差異出現,低成就組(後10%)女生的表現較佳,高成就組(前10%)男生的表現較佳。分科表現的性別差異分析結果顯示,全體受試的女生在健康教育表現較佳,男生在物理和地球科學表現較佳,男女生在生物和化學則無顯著的差異;低成就組女生在健康教育的表現較佳,其餘科目則無顯著的性別差異出現;高成就組女生在健康教育的表現較佳,男生在物理的表現較佳,其餘科目則無顯著的性別差異出現。在性別DIF分析上,本研究採Mantel-Haenszel法評估DIF,分析結果顯示,在分析的575題中計52題出現DIF,DIF 出現率約為9%,在探討之試題特徵中以科目對DIF量數影響最明顯,其中健康教育、化學和生物DIF題有利女生居多;地球科學和物理DIF題則有利男生居多。研究者對DIF試題的內容進行進一步審視,初步並未發現和測驗目標無關的因素,也就是說,雖出現DIF現象但尚不構成試題偏誤。篇末根據研究發現提出建議供相關人員參考。
This research investigated gender differences and differential item functioning (DIF) on Science parts of the Basic Competence Test for Junior High School Students (BCTEST) from 2001 to 2005 administrations. Results based on overall science performances indicated that there were no visible gender differences among total groups. However, females performed slightly better than males among low-achieving groups (bottom 10%); males performed slightly better than females among high-achieving groups (top 10%). Results based on performance of subjects indicated that females outperformed males on Health Education, males outperformed females on Physics and Earth Science, and there were no visible gender differences on Biology and Chemistry among total groups. In addition, there were no visible gender differences among low-achieving and high-achieving groups with only a few exceptions. The exceptions were that females outperformed males on Health Education among both low-achieving and high-achieving groups, and males outperformed females on Physics among high-achieving groups. In the study of gender DIF, Mantel-Haenszel procedure was used. The results of DIF analyses showed that total 52 (about 9%) of the 575 items across test administrations were categorized as DIF. Among the factors explored in the study, including subject, cognitive demands and types of attachment materials, the subject was the most salient factor that affected the performance of the DIF measures. The DIF items from Health Education, Chemistry and Biology subjects tended to favor females; however, the DIF items from Earth Science and Physics subjects tended to favor males. The follow up review and judgment were conducted for these DIF items, and no construct irrelevant factors were found. That is, though these items displaying DIF, they were not biased items. Implications based on the findings of this study were proposed for educators, test developers and researchers.