Reading the paper that Clay introduces below strengthens my sense that the people who pioneer group-forming practices are those who have a marked interest in something that is not generally shared by the rest of the population. Consider the top Meetup topics in the Music category to see what I mean.
Odd Associates with Odd. At least that's the conclusion in a paper about social clustering in Club Nexus, a service for Stnaford University's online population, written by Lada A. Adamic, Orkut Buyukkokten, and Eytan Adar.
Because the users of the service left such a rich metadata trail, they were able to test a number of assertions about social congres that had previously been made only as generalities. In addition to uncovering the expected gross patterns (power laws, clustering, small worlds networks, low hop-counts between people, etc), they were able to make refined observations about what sorts of affinities correlate with high clustering (the higher the listed ratio is above 1, the stronger the correlation with social clustering):
We found further that, in general, activities or interests that are shared by a smaller subset of people showed stronger association ratios than very generic activities or interests that could be enjoyed by many. For example, raving (1.64), ballroom dancing (1.61), and Latin dancing (1.49) showed stronger association in the social activity category than barbecuing (1.20), partying (1.18), or camping (1.11) [...]
In sports in particular, multi-player team or niche sports were better predictors of social contacts than sports that could be pursued individually or casually. Among water sports, synchronized swimming, diving, crew, and wake boarding were better predictors than boating, fishing, swimming or windsurfing. In the land sports category, team sports, in particular women's team sports such as lacrosse and field hockey were better predictors than soccer (often played casually as opposed to in a competitive college team), tennis, or racquetball. [...]
We observed that niche book, movie, and music genres were more predictive of friendship than generic ones. Gay and lesbian books, read by 63 users, had a ratio of 4.37, followed by professional and technical, teen, and computer books. In contrast, the general category of 'fiction & literature' had a ratio of 1.09.
