Prior applications of the lexical decision task in second language research have either examined performance accuracy (Meara and Buxton 1987) or speed of response to familiar items (Segalowitz and Segalowitz 1993). This paper examines how well the two measures together serve to discriminate among between-group levels of proficiency and within-group levels of difficulty on an English lexical decision task. Performance was compared across three levels of English proficiency (Intermediate L2, Advanced L2 and English L1 control) and four levels of item difficulty, as defined by frequency of occurrence (words from the 2000, 3000, 5000 and 10,000 most frequently occurring words). Accuracy and reaction time measures systematically decreased as a function of increasing proficiency and frequency level. Response variability, as measured by the coefficient of variance, also decreased as performance improved. The implications of the findings for the use of lexical decision tasks in second language research are discussed.