A Test of Frequency-based Processing: Evidence from Korean Numeral Classifier Constructions

Abstract

The Korean numeral classifier constructions are a good choice to test frequency-based processing because one of the two types of construction (prenominal and postnominal) occurs more often than the other in production, and distributional frequency matters in comprehension. The study conducted an experiment with a self-paced reading task. Findings showed that (i) comprehension accuracy is significantly higher for the more-frequent postnominal constructions than for the less-frequent pre-nominal constructions and (ii) reading times at the critical region of pre-nominal constructions are faster in the condition with a general classifier than in the condition with a specific classifier, although reading times in postnominal constructions did not differ in the two conditions. These findings indicate that the familiarity (i.e., experience) of frequent structures facilitates faster and more accurate comprehension. The study thus contributes crosslinguistic evidence to add to the evidence supporting the frequency-based account based on distribution.