Abstract
Studies in cross language speech perception sometimes use self‐reported values of language proficiency as an independent variable. In relation to other measurement methods, proficiency self‐reports have been found susceptible to several sources of bias including demographic variables and proficiency level itself [S. Yamashita, Six Measures of JSL Pragmatics, U. Hawaii Press (1996)]. An inaccurate scaling of an independent variable can limit the precision of performance models that are based on that variable’s relation to the dependent variable. This paper presents a convenient method (Versant tests) for measuring spoken language proficiency that locates subjects on a 60 point scale with a standard error measurement of less than 3 points, thus reliably locating subjects into 20 levels along a linear continuum of proficiency. The testing procedure takes about 15 minutes to complete. The test‐retest reliability of Versant scores for spoken English and Spanish are 0.97 and 0.96, respectively, both with correlations greater than 0.85 with independent human rated tests. The development, operation, and psychometric properties of the Versant tests are presented briefly, along with examples of their use in evaluating language instruction programs and standard listening comprehension tests.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call