Conversation is not an exchange of opinion. It is a sifting of opinion.
–Jacques Barzun
Welcome to the wonderful world of Teachout’s Cultural Concurrence Index! It’s a new game, everyone can play, and nearly everyone has. At least a dozen bloggers have offered their answers, in part or in full, to Terry’s list of 100 binary, mostly arts-related questions. This adds up to nearly 12,000 opinions, all told — a plethora of opinions, a cornucopia of opinions, more than a division’s worth of opinions, each as distinctive as a soldier in uniform. Terry himself makes modest claims for his Index:
So what does the TCCI do, accurately or otherwise? It measures the extent to which your taste resembles mine — but thats all. What’s more, you probably noticed in taking the test that my taste cant be “explained” by any one principle or theory. Had I scrambled the order of the alternatives and asked you to guess my answers based on your prior knowledge of my work, I doubt many of you would have scored much higher than, oh, 70%, unless you also knew me personally and very well indeed. Yes, Im a classicist, but I also prefer Schubert to Mozart, which tells you…what?
By that standard it is a resounding success, despite a few minor difficulties with selection bias. (In a recent bulletin from the Institute of Tautological Studies, 99.87% of bloggers who answer online blog quizzes report that they’d rather read blogs than magazines.) And I defer to no one in my admiration for Terry’s catholic taste, animated by no one principle or theory. Yet Terry has stumbled over a potentially useful tool. Let’s suppose what I increasingly doubt, that we actually want to learn something. How might we proceed?
The overall scores tell us little beyond, as Terry says, how far you agree with Terry. They range from 40% to 70% agreement, which scarcely differs from chance, once we account for the fact that the test-takers read Terry in the first place and can be expected to share his tastes at a higher than random rate. But what if we cross-correlated the questions? Suppose we could persuade 1,000 bloggers to take the test, which ought not to be too difficult; they seem to have little else to do. A hundred questions yield 4,950 possible cross-correlations (99 correlations for each answer, divided by 2, since they are symmetric). Now we calculate the agreement between the answers to each pair of questions, looking for the outliers, as far from 0.5 as possible. With close to 1,000 data points for each answer pair — some respondents will opt out of some questions — we put ourselves in a position to draw a few conclusions. Some answer pairs will likely match up quite closely; I would expect a few correlations of 0.9 or higher.
Then we take these outlying pairs and run them back against the data set a few times more, looking for clusters. Interdisciplinary clusters will be best of all; if we find, for example, that nearly everyone who prefers Astaire to Kelly also prefers Matisse to Picasso and Keaton to Chaplin, then we might be on to something. We examine the clusters, looking for commonalities. Looking for rules, in other words. Although Terry’s taste, or the taste of any educated person, cannot be explained by one principle or theory — this is a reasonable working definition of “cultivated” — I would wager that it can be explained pretty well by several.
Unfortunately this is work. I don’t do work around here; I just assign it.