N-gram measures and L2 writing proficiency

Productive knowledge of English phraseology can assist second language writers in composing more effective and target-like texts. As such, researchers have taken a great interest in L2 phraseological competence by comparing the use of multi-word sequences by L2 writers with that of native speakers as well as comparing the use of multi-word sequences of L2 writers from different proficiency levels. This study expands on this research by investigating how multi-word sequence use can predict human judgments of writing proficiency using a multi-faceted approach. The study specifically examines multiple indices of bigrams and trigram use (frequency, range, association strength) in a corpus of English placement test essays written by L1 Korean learners of English. Results of the multiple regression revealed that n-gram proportion and association strength measures were predictive of human judgments of writing proficiency, indicating that higher rated essays include more strongly associated academic bigrams, a greater proportion of frequent academic trigrams, and more strongly associated spoken trigrams. These results highlight the importance of including instruction on multi-word sequences in the second language writing classroom.