Investigating the Predictive Power of Grammatical Metaphor Nominalization as Lexical Complexity Benchmarks for EFL Writing Quality

Abstract

Grammatical metaphor (GM) has been empirically demonstrated to have a significant positive correlation with the quality of English as a Foreign Language (EFL) writing, with higher‐scoring writings tending to utilize a greater number of GMs. Among the various types of GM, grammatical metaphor nominalization (GMN) is a key focus when examining EFL writing. The GMN use is commonly measured by its normalized frequency within a text or GMN density, an index that falls within the dimension of lexical complexity and considers the overall presence of GMN. However, it is essential to separately analyze the frequencies of GMN subtypes, such as Process as Thing realized by deverbal nouns and Quality as Thing realized by deadjectival nouns. Furthermore, when evaluating the EFL writing quality, it is crucial to explore additional dimensions of GMN complexity, including the GMN sophistication and GMN diversity. In light of this context, the present study employs correlation and regression analyses to investigate the predictability of the GMN complexity for the assessment of EFL writing quality. The findings are summarized as follows: (a) neither the overall nor deverbal GMN, when related to GMN density, could effectively predict EFL writing quality; (b) in contrast, deadjectival GMN in relation to GMN density, low‐frequency GMN concerning GMN sophistication, and the entropy and relative entropy values of GMN concerning GMN diversity, were found to be statistically significant predictors; (c) when indices were combined, deadjectival GMN, low‐frequency GMN, and the relative entropy of GMN together accounted for the largest variance of EFL writing scores with an extremely large effect size. This research not only theoretically advances the understanding of GMN in assessing the EFL writing quality but also provides pedagogical insights for instructors teaching EFL writing.