Discrimination difficulty, cognitive burden, and reversal impairments in a maternal immune activation model of schizophrenia risk.

Reversal-learning impairments in the maternal immune activation (MIA) model of schizophrenia risk have been interpreted as being indicative of deficits in reinforcement learning. Here we sought to assess the specific role of cognitive burden in discrimination learning and reversal performance in this model. Control and MIA rats were trained on a visual discrimination task in which responses on either a left or right lever were rewarded depending on the location of a cue light at the beginning of the trial. Groups of MIA and control rats differed in the difficulty of the discrimination rule they initially learned (pressing the lever on the same side vs the opposite side of the cue light). Once the discrimination was learned, performance was tested across four reversals. Across all phases of the experiment, rats in Group Same performed better than rats in Group Opposite. There was no difference in performance of control and MIA rats during acquisition or baseline, but MIA rats displayed impaired performance across reversals, with performance decrements manifesting later in reversals after the new discrimination rule had been learnt. Across reversals, MIA rats also made more perseverative errors than control rats. These results are consistent with others that have shown reversal learning impairments in MIA rats. The results further suggest that impaired behavioural flexibility in the MIA model is not due to a deficit in reinforcement learning, but due to an impaired ability to organize information gained from experience into an accurate and stable representation of the current task requirements.