Exploring new measures of accuracy in second language task-based spoken performance

Abstract

This article focuses upon the measurement of accuracy in second-language task-based spoken performance, with measures related to unit of analysis, error gravity, and accuracy linked to clause length. Building on established indices, three new measures of accuracy, all linked to clause length, are described. In addition, variations in an error gravity formula are also used. The measures, both established and new, are trialed with an existing dataset from a study exploring planning effects on second language narrative performance. A range of established measures is shown to generate appreciable effect sizes with this dataset. The newly developed measures also work well, as do alternative error gravity formulae. The new length-accuracy measures also seem sufficiently distinct from measures of language complexity. We discuss the implications of these findings for measuring accuracy in task-based spoken language performance, as well as the relevance of the new measures for theoretical accounts of second language performance.