This experiment tests between two competing hypotheses about the source of failures in comprehension by Broca-type aphasics with agrammatic production. These are characterised as (1) the hypothesis that these aphasic individuals have sustained a partial loss in syntactic knowledge, and (2) the hypothesis that, despite intact structural knowledge, they suffer from an inability to put that knowledge to use in comprehension tasks such as object manipulation and sentence-picture matching. To decide between the hypotheses, this study compared the speed and accuracy of Broca-type aphasics with a control group of normal subjects using an on-line grammatically judgement task in which the anomaly involved closed-class vocabulary items. The results are in accord with the view that the source of agrammatic performance is not a loss of syntactic knowledge, as the responses of the aphasic group closely mirror those of the control group (e.g. word position effects were found for both groups). The results are interpreted, instead, as support for the alternative view that agrammatic aphasics have difficulties in processing syntactic knowledge.