Investigating the Quality of Language Tests; Calculating the Reliability Coefficient of Placement Test as a Case Study

Abstract

Language measurement tools should be valid and reliable by which the test scores obtained help drawing meaningful inferences and, fairly enough, significant decisions are made. This study attended to the reliability of an English language placement test. The aim of the study is to find out the reliability estimate of that type of test. Based on the Classical Test Theory, the reliability coefficient of the test should be computed to find out the range of the measurement error and achieve an estimate of the candidates’ true scores. To this end, the data obtained from the administration of University of Raparin’s Language and Development Center Placement Test[i], which was administered to 889 freshmen students, was analyzed. The method is descriptive and exploratory in which, by using the split half method, the three formulas – Kuder-Richardson 20 and Spearmen-Brown Prophecy – were used to calculate the reliability estimate of the test. The results showed that the reliability estimate of the test is 0.81. They also showed that the number of the items of the test should be doubled in order to achieve the desired reliability of 0.9 and above. Thus, based on the results obtained, it is recommended that the test structure should be improved to achieve a higher rate of reliability estimate. It is also recommended that the test should be modified; test items need to be increased and the content amended to cover all the language skills. Additionally, it is recommended that other sources of unreliability, namely, test administration and individual test items need to be studied to find the source of measurement error.