测试项目的相对难度假设

Hypothesis of Relative Item Difficulty

语言测试是一个从受试者的观察分推导出其能力真分的过程。目前流行的能力评估模式都是建立在单维能力假设的基础上,即:影响受试者测试行为的因素只是受试者某一特定的单一语言能力。但是,在实际的语言测试中,影响受试者测试行为的因素除了受试者的语言能力外,还有许多其他因素。如何在能力估算中考虑这些因素,是本文讨论的中心问题。笔者提出了相对难度假设:任何一个测试项目,其难度值只是相对于影响测试过程的某(些)个性特征,如:语言能力、测试策略或其它认知心理特征。对于不同的受试者,由于作用于其项目反应的因素不同,不论其能力水平是否相同,答对的概率都不完全相同,能力估算的结果也就不相同。

Language testing is a process making a general inference about an examinees unobservable true score from his/her observed score. The present commonly used ability estimation is made on the basis of the unidimensionality hypothesis that the examinee抯 test performance is accounted for by his/her single trait or ability. But as we know, there are many other factors that may affect the test performance. Then the problem arises as to how to include these factors in the ability estimation. In the present thesis, the author propose the hypothesis of relative item difficulty, which claims that the item difficulty value is relative to the individual factor(s) that may affect the test performance, including ability level, test strategies and other cognitive factors. Different examinees, even though their ability levels are the same, have different probabilities to answer a given item correctly. The ability estimation with relative item difficulties will become more accurate and reliable.