Does the Exam Score Predict Later Performance? Examination Performance and Delayed Parallel-Form Performance in a Grade 5 Science Classroom in Nepal

Preprint, not peer reviewed. Summative school examinations are widely used to certify learning, but whether routine examination scores predict performance on a comparable, delayed parallel-form assessment is rarely tested directly, particularly outside Western, Educated, Industrialized, Rich, and Democratic (WEIRD) educational contexts. This study compared Grade 5 students’ scores on their school’s regular first-term written Science examination with scores on a parallel-form Science assessment, whose subject and content were not disclosed in advance, covering the same curriculum objectives, administered seven days later in the same format and duration. Forty-one students (ages 10 to 12; 25 males, 16 females) from one Nepali classroom completed both measures. Scores were strongly related, r(39) = .84, 95% CI [.72, .91], p < .001, and examination score significantly predicted delayed assessment score, R² = .71, F(1, 39) = 93.98, p < .001. Scores were nevertheless lower on the delayed form, t(40) = −9.66, p < .001, mean difference = −9.39 points, 95% CI [−11.35, −7.43], dz = −1.51 (dav = −0.86), with 38 of 41 students (92.7%) scoring lower. Absolute agreement was moderate (Lin’s concordance coefficient ≈ .61) despite the high correlation. Because the two forms differed in content and were not empirically equated, this decline cannot be attributed to forgetting alone. Examination scores strongly preserved students’ relative standing, but scores were generally lower a week later. Implications for classroom assessment practice in examination-centered education systems are discussed.