Representativeness

One should not compromise on the essential standard of representativeness of the corpus as a whole. A corpus is only good if it can be reasonably trusted to represent the way language is used by those people whose usage one is interested in describing.

Our examination of the representativeness of WEIRD subjects is necessarily restricted to the rather limited database currently available.

Drawing on texts recommended in curricula and controlling for two countries with benchmarked curricula improves the external representativeness of the corpus.