L2 Retention and Attrition of Korean Numeral Classifiers

Abstract

This study examined L2 maintenance and loss of Korean numeral classifiers to see if the regression hypothesis (Jacobson, 1941; 1968) holds. The regression hypothesis states that language components learned first will be lost last. Taking the regression hypothesis as a theoretical starting-point, we used both comprehension and production tasks to measure 82 L2 attriters’ knowledge of Korean numeral classifiers. The results showed that while the recall of classifiers declined rather quickly and precipitously after L2 disuse, the recognition of classifiers remained stable for a considerable period of time. The results also demonstrated that the attrition of numeral classifiers tended to occur in an reverse order in which they had been learned. That is, the classifiers retained longest were those learned earliest. Such classifiers were the human classifier myeng and neutral classifier kay. These findings appear to be consistent with the predictions made by the regression hypothesis.