This study compared holistic and analytic marking methods for their effects on parameter estimation (of examinees, raters, and items) and rater cognition in assessing speech act production in L2 Chinese. Seventy American learners of Chinese completed an oral Discourse Completion Test assessing requests and refusals. Four native Chinese raters evaluated the examinees’ oral productions using two four-point rating scales. The holistic scale simultaneously included the following five dimensions: communicative function, prosody, fluency, appropriateness, and grammaticality; the analytic scale included sub-scales to examine each of the five dimensions. The raters scored the dataset twice with the two marking methods, respectively, and with counterbalanced order. They also verbalized their scoring rationale after performing each rating. Results revealed that both marking methods led to high reliability and produced scores with high correlation; however, analytic marking possessed better assessment quality in terms of higher reliability and measurement precision, higher percentages of Rasch model fit for examinees and items, and more balanced reference to rating criteria among raters during the scoring process.

No abstract