The reliability and validity of three curriculum-based measures as indicators of learning English as a foreign language were examined. Participants were 260 Dutch students in Grades 8 and 9 who were receiving English-language instruction. Predictor measures were maze-selection, Dutch-to-English word translation, and English-to-Dutch word translation. Criterion variables were years of English instruction, school level, course grades, and scores on a standardized reading test. Different scoring procedures and time frames were compared. Alternate-form reliabilities ranged from 0.44 to 0.88. Significant differences in maze scores were found between school levels but not between years of English-language instruction. Correlations between predictor and criterion variables ranged from 0.19 to 0.79. A regression analysis revealed that a combination of maze and English-to-Dutch translation predicted English course grades better than a single measure alone. (Contains 10 tables.)