Data from 96 subjects were obtained on either location (indicating presence or absence of probe words) or retrieval (responding to probe words with the next word) for both active and passive sentence examples. The subjects were timed to respond in both tasks. The resulting latency profiles were compared and the relative contribution to these of surface structure and semantic features was assessed. In the location profiles, fast, medium and slow positions were isolated and found to be associated with semantic function in both the active and passive sentences. In the retrieval profiles structural factors to a large extent predicted the latencies obtained. Differences between active and passive forms were more evident in location than in retrieval.