The aim of the present study was to compare two most widely employed versions of the reading span task in Serbian speaking population, with respect to the procedure and validity. First step was to compare psychometric properties of several measures of reading span, derived from each task version. Second step was to compare predictive validity of memory span and processing measures, for each task version. Two task versions proved to be parallel, with small discrepanciens, providing reliable and discriminative measures. The only exception was traditional reading span measure, which has not met any of psychometric standards. Therefore, we propose this measure to be abandoned. While processing measures showed better predictive validity with respect to criteria of verbal ability and reading comprehension, reading span measures independently contributed to prediction of the same criteria with only one exception. Results obtained suggest a small modifications of reading span tasks required in order to improve their psychometric properties. They also suggest combining processing and span measures might yield a better verbal working memory efficiency estimate.