This study examines the on-line processing of sentences with movement using an auditory anomaly detection task (after Boland, Tanenhaus, Garnsey, & Carlson, 1995). Eight agrammatic aphasic participants (four of whom had undergone treatment focused on comprehension and production of filler-gap sentences) and 24 young normal participants listened to sentences and pressed a button when the sentences ''stopped making sense.'' Critical sentences contained an anomaly in object relative clauses or conjoined clauses. Results showed that both young normals and aphasic participants were able to reject anomalous sentences of both types. In addition, both groups showed evidence of filler-gap resolution on-line. Importantly, however, there was evidence of a treatment effect for the aphasic patients: those who received treatment showed better performance than those who had not. Treated patients were more successful than the untreated patients in detecting the anomaly in filler-gap conditions, rejecting the anomalous filler-gap sentences reliably more often than the non-anomalous ones, like the young normals. This effect was not noted for untreated participants, i.e., there was no statistical difference between their rejection of anomalous and non-anomalous filler gap sentences. Further, the reaction time data showed that the treated aphasic patients' rejections came before sentence's end (within 2000ms), while the majority of responses made by untreated patients did not. These results indicate that individuals with agrammatic aphasia appear to retain some gap-filling capacity and that treatment can improve their ability to make use of this capacity.