Univerbation of N+V constructions in German is probabilistic and driven by morphosyntactic prototypes (Schäfer & Sayatz)

This publication is in the INCUBATOR section.

Roland Schäfer & Ulrike Sayatz (in preparation) Univerbation of N+V constructions in German is probabilistic and driven by morphosyntactic prototypes. To be submitted in Q1/2018.

This paper deals with the formation of German verbs through univerbation of a verbal root with a bleached noun, such as teetrinken ‘drink tea’ or mutmachen ‘encourage’. The status of these verbs (one word or two words) is often disputed, and the problem plagues normative attempts to systemise their spelling. On the theoretical side, we argue that on top of individual differences between single items (higher or lower degree of univerbation as a result of individual diachronic development), the syntactic context plays a crucial role in how likely speakers are to conceive of the item as one or two words. Prototypical nominal contexts trigger univerbation much more strongly than prototypical verbal contexts. Empirically, we use the large DECOW16A web corpus, which has the advantage of containing a huge amount of non-standard spontaneous writing. In addition, we conducted a large-scale dictation experiment, where participants were exposed to items with a high degree of univerbation (according to the corpus analysis) and other items with a lower degree – both across all relevant morphosyntactic contexts. The corpus findings as well as the experimental results corroborate our theoretical predictions strongly. Finally, we make a methodological point by arguing that spontaneously produced written data as found in appropriate corpora as well as controlled writing experiments are instrumental in inferring speaker-writers’ conceptualisation of grammatical categories.