Berndt and Caramazza (1999) claim that Grodzinsky, Pinango, Zurif, and Drai (1999) were able to show a canonical-noncanonical difference (e.g., actives vs. passives) in the comprehension of Broca's aphasic patients only because of a patient selection bias. We show that the canonical-noncanonical comprehension pattern exists apart from any such bias, and that this pattern bears importantly on our understanding of the neuroanatomical organization of comprehension capacity at the sentence level.