Validity and the automated scoring of performance tests

In recent decades, the trend toward more communicative approaches to language teaching and a growing demand for more authentic measures for assessing language abilities have given rise to the revival of performance-based language tests (McNamara, 1996). Performance-based tasks elicit authentic performance samples that are evaluated based on pre-established scoring criteria. They are often referred to as “constructed response” tasks (Bennett, 1993; Kane et al., 1999), which require responses other than selecting from fixed options.