In recent years, three indicators of language proficiency accuracy, fluency and complexity have been receiving great attention in analyzing speech production. However, there are a number of different statistical procedures used in examining each indicator. What methods should be used is an important issue especially when we have cross-sectional research. This paper addresses several problems of the measures for accuracy, fluency and complexity used in previous studies on planning and task type and in a past study held in 2002. In addition, it offers possible measures that we could employ in future research and in classroom language observation.
