This book is about how to measure the relative stabilities of linguistic
typological features. It discusses three alternative methods and tests
their performance by applying them to simulated datasets having preset
stabilities. The best metric is then applied to the data of The World Atlas
of Language Stuctures (Haspelmath et al. 2005) to produce empirical
estimates of stability for 134 features and 445 feature values. The
numerical results concur with many specific categorical statements in the
typological literature, and also substantiate the general suggestion that
stable features are more strongly interrelated than are unstable features.
The results also show that features have approximately the same relative
rates of change in different languages, even in widely separated
geographical areas. Surprisingly, however, tendencies for features to
diffuse vary among areas with no consistent differences among features. It
follows that stability and diffusibility are separate dimensions rather
than opposite ends of the same dimension.