This dissertation attempts symbolic music sentiment transfer across Russell’s four quadrants through bar-level control of musical attributes associated with changes in perceived sentiment. Based on a survey carried out by Panda et al., musical attributes associated with changes in perceived sentiment were identified and methods to extract them were created. These were added to the existing MuseMorphose model, a program designed for style transfer through attribute control. The model was retrained on these additional attributes and sentiment-transferred output was generated to be fed back into EMOPIA, an emotion classification tool. The results from this research found that transferring high valence pieces to low valence were the most often correctly identified as their target sentiments. In addition, modifications to control of attributes related to changes in perceived arousal yielded higher accuracy in classification in later experiments, although no cases where accuracy rose above 25%. This suggests that through further changes to how attributes are controlled in sentiment transfer, symbolic music sentiment transfer across Russell’s four quadrants through control of musical attributes will be achievable. Suggestions for further work are included in the conclusion of the dissertation.