Careful systematic tests of hearing ability may miss the cognitive consequences of sub-optimal hearing when listening in the real world. In Experiment 1, sub-optimal hearing is simulated by presenting an audiobook at a quiet but discriminable level over 50 min. Recall of facts, words and inferences are assessed and performance compared to another group at a comfortable listening volume. At the quiet intensity, participants are able to detect, discriminate and identify spoken words but do so at a cost to sequential accuracy and fact recall when attention must be sustained over time. To exclude other interpretations, the effects are studied in Experiment 2 by comparing recall to the same sentences presented in isolation. Here, the differences disappear. The results demonstrate that the cognitive consequences of listening at low volume arise when sustained attention is demanded over time.