Objective: A recent study indicated that amnesic patients have difficulties not only in describing past and imagined scenarios, but also in describing pictures that are in full view. This finding suggests that impaired memory hampers descriptions of scenarios more broadly. However, no such impairment in picture description in amnesic patients was observed in a related study. One key methodological difference between these studies was the complexity of the pictures to be described, hinting that group differences might be marked only if pictures are sufficiently complex to tax aspects of memory. Method: To test this complexity hypothesis, we examined whether differences in picture description between amnesic patients and controls increase with increasing picture complexity. As in previous studies, we also assessed our participants’ ability to describe imagined scenarios. Results: Amnesic patients reported significantly fewer elements than did controls when describing pictures and imagined scenarios. The group difference in picture description was significantly larger for complex than for simple pictures. Conclusion: Although variations in lesion sites might have accounted for the aforementioned cross-study differences in picture description in amnesic patients, our results suggest that, at least in amnesic patients with extramedial temporal lobe lesions, the complexity of pictures can determine whether or not a (substantial) picture-description deficit is observed. We interpret these findings in terms of a narrative-construction deficit. We hypothesize that, whereas brief narrative can be constructed via nonmemory cognitive processes, the construction of more detailed narrative depends upon intact functioning of a temporary memory system, such as the episodic buffer.