This paper presents experimental evidence for the role of lexically specific representations in the processing of English be and get passive constructions. Using a self-paced reading task, processing of full and truncated be and get passives was compared on sentences containing verbs strongly associated with these constructions, as determined by supercluster analysis (see e.g., Myachykov et al. 2012). This study complements these corpus-based studies by providing evidence from an on-line processing task that tests whether native speakers are sensitive to the observed distributions. The results support a usage-based functional account of processing and interpreting English be and get passive constructions. Participants' performance was influenced by frequency and lexical specificity. The study also provides evidence of education-related differences in language attainment – the higher educated participants were significantly better at interpreting be and get full passive constructions than the lower educated participants.