Fifteen adults with traumatic brain injury (TBI) and 15 adults without brain injury listened to narratives, made delayed predictions of recall, and took a delayed recall test. Narrative questions differed by salience (main ideas, details) and explicitness (implied, stated) (R. H. Brookshire & L. E. Nicholas, 1993). TBI survivors recalled less than control participants regardless of question type. All participants recalled main ideas and implied information with greater accuracy than details and stated information. Predictive accuracy for recalling stated information was strong regardless of group. Participants were unable to predict recall for implied information. The materials-appropriate-processing (MAP) hypothesis proposes that predictive accuracy is biased by text type (i.e., predictive accuracy for recalling main ideas should be higher than for details when learning narratives). However, there were no differences in predictive accuracy for recalling main ideas and details, with both groups predicting recall modestly well. Controlling for explicitness appears to be an important variable for future metamemory text studies.