Abstract Previous research on syntactic complexity is primarily focused on the synchronic distribution of clausal and phrasal features and the diachronic shift from clausal elaboration to phrasal compression. However, the interrelationship between clause complexity and phrase complexity remains unexplored. This study investigated syntactic complexity at different linguistic levels across three disciplinary groups (Social Sciences, Humanities and Natural Sciences) using a corpus of research article abstracts. Sentence complexity was measured by the number of clauses per sentence, clause complexity by the number of clausal constituents per clause, and nominal group (NG) complexity by the number of words per NG. The results show that: (1) sentences are the least complex in Natural Science (NS) texts; (2) clauses are also the least complex in NS, despite having the highest average number of clausal constituents; (3) NGs are the most complex in NS texts. Furthermore, the study found that NG complexity could be more accurately measured by the number of premodifiers of the head noun (HN) of the NG. These findings have important implications for instructing English as a Foreign Language (EFL) learners in discipline-specific academic writing.