Two Faces of One Suffix: Some Thoughts on Using Corpora in Usage-Based Studies of Word-Formation

This paper compares the semantic profile of a single multifunctional derivational suffix derived from data obtained in two general digital corpora of Croatian. The primary motivation is to explore whether our verdicts about the semantics of affixes may depend on the corpus selected as the source of empirical material. The issue is of vital importance, especially for those studying word formation from a usage-based perspective. If grammar is construed as the cognitive organization of our experience with language (Bybee 2006) and if we turn to large, general digital corpora for evidence of this experience, we must be aware that examining different corpora may lead to different hypotheses about users’ internalized grammar. The here-presented semantic analysis of the Croatian nominal suffix -ar(a) in the more controlled Croatian National Corpus v3.0 and the liberal web-based corpus hrWaC v2.2 yielded conspicuously different results about its dominant function. This does not mean that similar discrepancies would necessarily be observed with other affixes, and it most certainly does not negate the value of corpora in studying word formation. However, such results do caution us against generalizing corpus-relative findings into some general “truth” about the affixes studied.