Telling the world’s least funny jokes: On the quantification of humor as entropy

• Informally observation has shown us that some non-words (NWs) reliably make people laugh.
• A survey study confirms better-than-chance agreement in human judgments of which NWs are funny.
• Extending Schopenhauer’s (1818) theory, we conceptualize NW humor in terms of Shannon entropy.
• Two experiments demonstrate clearly that Shannon entropy does predict human judgments of NW humor.
• We conclude that perceived humor in NWs is a quantifiable function of how far NWs are from being words.

In assessing aphasics or conducting experiments using a lexical decision task, we have observed informally that some non-words (NWs) reliably make people laugh. In this paper, we describe a set of studies aimed at illuminating what underlies this effect, performing the first quantitative test of a 200 year old theory of humor proposed by Schopenhauer (1818). We begin with a brief overview of the history of humor theories. Schopenhauer’s theory is formulated in terms of detection/violation of patterns of co-occurrence and thereby suggests a method to quantify NW humor using Shannon entropy. A survey study demonstrates that there is much more consistency than could be expected by chance in human judgments of which NWs are funny. Analysis of that survey data and two experiments all demonstrate that Shannon entropy does indeed correctly predict human judgments of NW funniness, demonstrating as well that the perceived humor is a quantifiable function of how far the NWs are from being words.