Theoretical Review: Understanding Test Reliability in Learning

Abstract

Evaluation is a crucial instrument in the Independent Curriculum instructional cycle for measuring the achievement of learning objectives and providing pedagogical feedback. However, in practice, the provision of valid and consistent evaluation tools still faces challenges, characterized by fluctuations in test scores due to weak control over measurement error. This article aims to examine in depth the basic concept of reliability, visualize its operation, and reconstruct techniques for estimating test instrument reliability coefficients. The research method used is library research, analyzing classic and contemporary literature in the fields of educational evaluation and psychometrics using a descriptive-qualitative approach. The results of the discussion indicate that reliability reflects the level of consistency, dependability, and stability of measurement results, which can be estimated through three main approaches: stability (test-retest), equivalence (parallel forms), and internal consistency (such as Split-Half, KR-20/21, and Cronbach's Alpha). Through simulation of the Spearman-Brown (Split-Half) formula calculation on a multiple-choice test, a reliability coefficient of 0.79 was obtained. Based on Guilford's criteria, this value is included in the high category and exceeds the minimum limit of feasibility for learning evaluation instruments (0.70), so the instrument is highly reliable.