Abstract
The discovery of general patterns and their subsequent explanation is
a familiar method in linguistics and other cross-cultural research.
This paper addresses the computerized discovery of significant
cultural and linguistic patterns, specifically implicational
universals. We dispute published suggestions that the mechanical
generation of universals is inadvisable, by arguing that such claims
confuse the discovery of a pattern with its explanation. We propose a
principled method that uses permutation tests to estimate whether the
discovered universals are significant. Finally, these methods are
used to mine the rich kinship dataset contributed by G.P. Murdock in
1970, and to report eleven universals found therein.
full paper (rtf)