One of the key unresolved issues in affective science is understanding how the subjective experience of emotion is structured. Semantic space theory has shed new light on this debate by applying computational methods to high-dimensional data sets containing self-report ratings of emotional responses to visual and auditory stimuli. We extend this approach here to the emotional experience induced by imagined scenarios. Participants chose at least one emotion category label among 34 options or provided ratings on 14 affective dimensions while imagining two-sentence hypothetical scenarios. A total of 883 scenarios were rated by at least 11 different raters on categorical or dimensional qualities, with a total of 796 participants contributing to the final normed stimulus set. Principal component analysis reduced the categorical data to 24 distinct varieties of reported experience, while cluster visualization indicated a blended, rather than discrete, distribution of the corresponding emotion space. Canonical correlation analysis between the categorical and dimensional data further indicated that category endorsement accounted for more variance in dimensional ratings than vice versa, with 10 canonical variates unifying change in category loadings with affective dimensions such as valence, arousal, safety, and commitment. These findings indicate that self-reported emotional responses to imaginative experiences exhibit a clustered structure, although clusters are separated by fuzzy boundaries, and variable dimensional properties associate with smooth gradients of change in categorical judgments. The resultant structure supports the tenets of semantic space theory and demonstrates some consistency with prior work using different emotional stimuli.