Research has shown that testing during learning can enhance the long-term retention of text material. In two experiments, we investigated the testing effect with a fill-in-the-blank test on the retention of text material. In Experiment 1, using a coherent text, we found no retention benefit of testing compared to a restudy (control) condition. In Experiment 2, text coherence was disrupted by scrambling the order of the sentences from the text. The material was subsequently presented as a list of facts as opposed to connected discourse. For the incoherent version of the text, testing slowed down the rate of forgetting compared to a restudy (control) condition. The results suggest that the connectedness of materials can play an important role in determining the magnitude of testing benefits for long-term retention. Testing with a completion test seems most beneficial for unconnected materials and less so for highly structured materials.