Common Errors in Teacher-Made Test Design

Common Errors in Teacher-Made Test Design

Michael Fields (University of Delaware, USA)
DOI: 10.4018/978-1-5225-6986-2.ch016

Abstract

Teaching and test writing require different skill sets. While teachers are often required to create language tests, they often have inadequate training. This may lead to tests that lack validity and reliability, making tests unfair for test-takers. Tests may contain more general errors in their development, leading to construct underrepresentation or construct-irrelevant variance, decreasing reliability and validity of the entire test and rendering results meaningless. Well-designed tests may also contain items that are not well constructed, which may again lower overall validity and reliability. These item-related errors include word matches, testing single words, issues with phrasing, developing good sets and distractors, and testing outside the text (such as language in the item or math skills). Increasing awareness of these issues and improving teacher skills in test writing will ensure more fairness for students in decisions based on test results.
Chapter Preview
Top

Reliability In Language Testing

Reliability refers to the degree to which a test measures consistently (Weir, 1990). In the absence of any intervening learning, a test should provide similar results no matter when the test is taken, where the test is taken, or who gives the test or scores the results. One measure of reliability is to give a test twice, with little or no intervening instruction. This is known as test-retest reliability (Weir, 1990). A test taken a second time should yield roughly the same results as when given the first time. We would expect a multiple choice reading test to be highly reliable, while a writing test, marked my different raters, may be less so. However, for obvious practical reasons, tests are usually not given to the same population twice.

Complete Chapter List

Search this Book:
Reset