Primary objective: The purpose of this study was to explore and extend previous findings that training with variant items improves generalization performance on novel semantic sentences in an individual with amnesia.

Research design: A case study of an individual with severe amnesia, Patient T.E., who participated in an extended training and multiple testing paradigm.

Methods and procedures: Patient T.E. participated in 16 training sessions and eight test sessions on his recognition and recall performance for novel 3-word sentences. Three conditions were compared: No Variance (one version of each sentence studied), Early Variance (three versions of each sentence studied from the onset) and Late Variance (each of three versions of each sentence gradually introduced throughout training). Performance on studied items and semantically related items was evaluated.

Main outcomes and results: Patient T.E. demonstrated better learning for Variance items than No Variance items and better generalization to semantically related items for the Late Variance condition. However, he showed an advantage for the No Variance condition on the recall task.

Conclusions: Gradually introducing the variant items into training may be the optimal strategy for training an individual with severe amnesia to learn and generalize new semantic information.