Essay-type and proficiency effects on lexical sophistication in L2 writing: A corpus-based TAALES study


Kavanoz S., Yılmaz S.

17th International Teaching and Language Corpora Conference, Caceres, İspanya, 16 - 18 Temmuz 2026, ss.33, (Özet Bildiri)

  • Yayın Türü: Bildiri / Özet Bildiri
  • Basıldığı Şehir: Caceres
  • Basıldığı Ülke: İspanya
  • Sayfa Sayıları: ss.33
  • TED Üniversitesi Adresli: Hayır

Özet

Lexical sophistication is widely conceptualized as a multidimensional construct encompassing frequency, phraseological patterning, psycholinguistic properties, and register-sensitive vocabulary (Kim et al., 2018; Lu, 2012). However, corpus-based research on how these dimensions vary simultaneously across essay types and proficiency levels is somewhat limited. This study investigates lexical variation across three essay types (argument, cause, effect) and three proficiency levels (B1, B1+, B2) using a learner corpus of 1,139 timed essays produced by EFL students at a major state university in Türkiye. Lexical features were extracted using TAALES 2.8.1 (Kyle & Crossley, 2015) and organized into interpretable dimensions reflecting core lexical, phraseological, and semantic properties. Results revealed remarkable and consistent prompt effects across most dimensions, pointing to the task-specificity of the learners’ lexical profile. Proficiency effects were more selective and dimension-specific, with clear progression identified in indices related to semantic richness, abstraction, frequency-association links. In addition, a limited number of interaction effects were found in dimensions pertaining to collocational strength, academic phraseology and lexical abstraction, which suggests context-dependent rather than developmental differences. Finally, regression analyses showed that a subset of lexical dimensions consistently predicted writing quality, with frequency, collocational strength, and semantic abstraction emerging as stable predictors across tasks, alongside task-specific extensions. These findings highlight the value of multidimensional lexical profiling for uncovering genre- and level-sensitive lexical patterns in learner corpora and support prompt-aware and dimension-specific interpretations in writing assessment.