The present study builds upon research in the CAF (complexity, accuracy, fluency) framework for examining learner performance to compare the lexical and syntactic complexity of learner output in spoken discourse and synchronous computer-mediated communication (SCMC) during completion of narrative tasks. Data were generated from transcripts and video-enhanced chatscripts for (N = 21) university learners of English from linguistically, educationally and digitally diverse backgrounds. Results revealed no significant difference in either the lexical or syntactic complexity of the narratives generated in these two modalities and instead found evidence that different types of learners in this heterogenous population were able to generate more complex language predominantly in one context over the other.