In behavioral psychology, the "Incongruity Theory" is currently the most well-accepted explanation as the underlying mechanism behind humor. This theory postulates that humor is caused due to the playful violation of the subjective expectations of an agent; caused by the presence of incongruous stimuli in the agent's observations. But whether humor is caused merely due to the presence of incongruous stimuli or whether it is caused by the resolution of an apparent incongruity by the sudden realization of another competitive explanation still remains an open question. Our work attempts to address this question in a data-driven fashion. We develop an information-theoretic algorithm to investigate this question in the semantic space and report results on three different data-sets. We observe that simpler stimuli like slapstick jokes, puns, one-liners etc. are better explained by the mere presence of incongruity, whereas more complex stimuli; like high-quality jokes, jokes with hidden meanings etc. contain greater amount of conceptual dissonance and hence, are better characterized using the idea of resolution of incongruity. Our methodology opens up exciting new avenues of research in the areas of humor and engagement and our results could find potential applications in areas like marketing and advertising.