mPFC-lesioned rats spent more trials using spatial response patterns for bowl choice than control rats, and interestingly, the number of remaining ED trials, after excluding spatial trials, was similar for the lesion and control groups. In the ED stage, rats need to disengage from the old attentional set (i.e., the previously relevant but currently irrelevant dimension) and then learn a new attentional set. During the disengagement process, rats would probably focus on the two stimuli of the previously rewarded perceptual dimension. After disengagement, rats may try various response patterns,
including spatial response patterns, in order to learn the association between food reward and the new dimension. Therefore, the result we found here may indicate that the lesion of mPFC may mainly affect rats’ ability to attend to or associate with the new reward-relevant dimension (e.g., mPFC-lesioned rats might react to uncertainty differently by requiring more evidence before they commit to reward hypothesis), rather than affect disengagement from the old dimensional set.
This is inconsistent with the previous study’s result showing that mPFC-inactivation mainly caused perseveration errors in the shift stage (Floresco et al., 2008). Such inconsistency may come from multiple factors. The previous study is based on a lever-pressing task which requires rats to shift from visual cue discrimination to position discrimination, whereas our data is from the 7-stage task which requires rats to shift from a discriminating medium to discriminating odours (or from odours to medium). In the lever-pressing task, the same visual stimuli were still there and spatial response is relevant with reward in the shift stage, while the medium and odour stimuli in the ED stage of the 7-stage task are novel and have not appeared in previous stages, and spatial response patterns are never usable to solve the discrimination in the 7-stage task. In the lever-pressing task, it is expected that rats would always get perseverative errors in the shift stage because the previously reward-relevant stimulus (not just dimension) was still present in the shift stage. The differences in types of discrimination learning and in novelty of stimuli in the ED stage between the two tasks may somehow interact with the effect of mPFC lesion on ED shift, causing different difficulties during the rat’s learning. Another factor may come from the data collection procedure. In the first four trials of each learning stage in the 7-stage task, each rat was allowed to dig the other bowl if it dug in the first, unrewarded bowl (Tait et al., 2014). This special step may make rats more quickly disengage from the previously relevant but currently irrelevant dimension, thus suppressing the perseverative errors that rats would have made. From this perspective, the lack of mPFC lesion effect on perseverative errors from the Bayesian analysis may be potentially from the special data collection step during the first four trials in
the ED stage. Further study without this step in data collection with the 7-stage task may help clarify whether mPFC lesion really affects perseverative errors or not.
Interestingly, besides the ED stage, we also found that lesioned rats spent more trials using spatial response patterns than control rats, particularly in the latter two reversal stages in the 7-stage task. Together with the above finding for the ED stage, this finding suggests that lesions of mPFC may affect rats’ ability to not only attend to the previously irrelevant perceptual dimension in the ED shift, but also attend to the previous unrewarded but currently rewarded stimulus of the same perceptual dimension in reversal learning. This again is not consistent with previous findings that mPFC lesion only affects rat’s or mice’s ED shift performance (Birrell & Brown, 2000; Bissonette et al., 2008; Brown & Tait, 2016). However, initial analysis of our large set of rats’ data (46 lesioned and 47 controls) did show that lesioned rats spent significantly more trials completing not only the ED stage but also the latter reversal stages than control rats. Therefore, that previous studies did not find the effects of mPFC lesions on reversal learning might have been due to insufficient statistical power given the modest effect size. Alternatively, there might be other noise factors which caused such inconsistency between our findings and previous studies which also used the 7-stage task for data collection, such as the location and size of the lesions. Therefore, further study particularly with a large dataset is necessary to clarify how mPFC lesion can really potentially affect reversal learning in compound discrimination.
Following the investigation of spatial response patterns, we also found that there are strong correlations between the number of spatial trials and the total number of trials within each stage for the lesion group. This is reasonable, because when rats took longer to find the correct hypothesis, they would likely use various spatial response patterns in more learning trials. Unexpectedly, the same rationale does not seem valid for control rats, because we found the correlations between the number of spatial trials
that normal rats may more flexibly change hypotheses during learning, such that spatial response patterns used by normal rats are more difficult to estimate than those used by lesioned rats.