Abstract

After a brief discussion of what The world atlas of language structures can and cannot (easily) be used for, the critical emphasis is on value assignments for the 140 features mapped. In particular, in an attempt to replicate the value assignments for one language, well-described and much-analysed German, a non-negligible proportion of the values assigned is found to be problematic, in the sense of being arbitrary or uncertain in view of analytic alternatives, unappreciative of dialectal variation, unclear as to what has been coded, or factually erroneous. To give an impression of progress in areal linguistics over the last century, WALS is then compared with an earlier world atlas of language structures, that of Wilhelm Schmidt (1926): for the features mapped in both atlases, despite substantially improved factual knowledge about languages, there is considerable congruity of the areal distributions shown.