Mankoff addressed some of these findings based on his own experience as cartoon editor. Captions with fewer punctuation marks fared better than others, as did captions that were harder to visualize. Editors also liked captions that were on average one word shorter than others.

The researchers then took the database they created, as well as additional information they collected, and conducted statistical and textual analysis.Ĭaptions with uncommon words were more likely to make the shortlist.From that analysis, the team determined that captions with uncommon words were more likely to make the shortlist. Since the available databases didn't contain values for abstractness or imaginability for all the words, the researchers took the 5,291 captions submitted for the contest and narrowed the pool to 86 - the 43 captions the New Yorker editors had shortlisted and a randomly selected additional 43 captions.Ī research assistant then judged that sample set on imagery (from 'very hard' to 'very easy' to imagine) and abstractness (from 'very concrete' to 'very abstract'), on a scale of one to seven. Basically, this answered questions like 'Is this word imaginable?' or 'Is this word abstract?' He looked at a psycholinguistic database and compared the words in the captions to what they mean psychologically. 'So the question is (A) 'What can you get out of that?' and (B) 'If you think that you can't get that much out of it, you have to go and collect some additional data.''įernbach considered using machine-learning techniques to find some meaning in the words, but because the captions were so short, the algorithms, which typically analyze pages of text, didn't offer much information. 'Basically, we have these 5,000 captions, information about whether they made the shortlist or not and whether or not they were finalists, but that's about all we had,' Fernbach said.

So how did McGraw et al come to their conclusions? First, they had to figure out what data they could glean from the captions. 'That would be a grave mistake which would only be justified if we could make a lot of money out of it.' 'I can say with certainty that we shouldn't compromise our research ethics by generalizing the results beyond this particular caption contest to all the other 283 contests, or, for that matter, to non-caption-contest New Yorker cartoons, or to humor in general,' Mankoff said in his response on McGraw's blog post about the caption contest findings. The results of the study weren't shocking, said Robert Mankoff, The New Yorker's cartoon editor.

New yorker caption contest winners