Neural changes associated with relational learning in schizophrenia

Abstract

Relational learning, which is learning the relationship among items, is impaired in schizophrenia but can be improved with training. This study investigated neural changes with functional magnetic resonance imaging before and after training on a relational learning task in schizophrenia and healthy control subjects. Despite their acquiring similar relational learning performance, the groups exhibited different neural activation patterns before and following training. Controls engaged regions within the relational learning network that included frontal, parietal, and medial temporal lobe, before and following training. Controls also exhibited activation reductions in region and spatial extent with relational learning proficiency, a commonly observed phenomenon in successful learning. In contrast, subjects with schizophrenia displayed no positive activations compared with the control condition before training. After training, subjects with schizophrenia displayed bilateral inferior parietal region activation as predicted. Contrary to hypothesis, hippocampal activation was not observed following training in schizophrenia. These findings suggest that the parietal lobe may be receptive to cognitive training interventions and that successful relational learning may be achieved in schizophrenia through the use of alternative extrahippocampal brain regions.

Fig. 1.

Illustration of Task Conditions and Training. For a trial, one pair of stimuli is presented. The subject’s goal is to learn the correct item in each pair. A “+” sign indicates the correct choice and a “−” sign indicates an incorrect choice. The transverse patterning (TP) condition requires relational learning, whereby an item is correct or incorrect depending on the item it is paired with. The SD condition does not require relational learning. Training consisted of 6 phases, 1–3 for the SD condition and 4–6 for the TP condition. Trials were presented in blocks that contained 16 trials per pair. Advancement to the next phase required the subject to score 13 out of 16 correct responses per pair within a block. Phases 3 and 6 depict the full SD and TP task that were administered during the functional magnetic resonance imaging sessions.

Fig. 2.

Different Hippocampal Responses Before and After Relational Learning Training in Control and Schizophrenia Subjects. A. Before training: Left posterior hippocampal activation difference between control and schizophrenia groups (top) and percent signal change per group (bottom). Statistical threshold refers to voxel-based significance set at P < .005. B. After training: Right anterior hippocampal activation difference between control and schizophrenia groups (top) and percent signal change per group (bottom). Statistical threshold refers to voxel-based significance set at P < .005.