Using lexical indices to predict produced and not produced words in second language learners

The paper explores how linguistic indices related to lexical networks and psycholinguistic models of lexical knowledge can be used to predict produced and not produced words in second language (L2) speakers. Two hypotheses are tested in this study. The first addresses how lexical properties thought to be important in word knowledge interrelate with word production. The second addresses which lexical properties are most predictive of word production. To test these hypotheses, a set of 45 frequent nouns and verbs produced by L2 learners were collected. A comparison word list of 45 frequent nouns and verbs produced by native speakers, but not found in the L2 data set were also collected. Polysemy and hypernymy values from the WordNet database along with word meaningfulness, concreteness, familiarity, and imagability values from the MRC Psycholinguistic Database and frequency values from SUBTLEXus were collected for each word. ANOVA analyses of variance and discriminant function analyses were conducted for each data set to examine which lexical indices discriminated between produced and not produced words and how these indices interrelated. The results of the noun analysis indicate that produced nouns are more frequent, more meaningful, and more familiar than not produced nouns. Results from the verb analysis show that produced verbs are more frequent, more meaningful, less specific, and more familiar than not produced verbs. These findings provide evidence for the importance of word properties in lexical production.