Different kinds of complexity: Plural variation in Dutch and German

To date, the investigation of morphological complexity has focused on formal aspects of complexity. Specifically, it is often referred to in terms of allomorphy; quantitatively as in the number of allomorphs involved in the marking of a category or qualitatively in terms of the conditioning factors according to which they are distributed across the lexicon (cf. Dammel & Kürschner 2008). In this context, the Dutch plural system, with its two productive native allomorphs -s and -(e)n and its largely phonologically based conditioning of plural marking (the trochaic output constraint, cf. Booij 2019: 20), is often analysed as relatively “simple” in comparison to other Germanic languages such as German (cf. Dammel & Kürschner 2008; Dammel, Kürschner & Nübling 2010). However, in German, there is a relatively transparent distribution of allomorphs to lexemes. In this sense, we find a striking difference between the two languages: Dutch has a large cohort of nouns which allow for both native allomorphs, making it less transparent than German. Often cited examples are nouns ending in schwa (e.g., gewoontes vs. gewoonten ‘habits’), in -aar, -eur, -ier, and -or (e.g., leraars vs. leraren ‘teachers’), and -el, -en, or -er (e.g., appels vs. appelen ‘apples’) (cf. Booij 2019: 17–29; van Haeringen 1947). What is missing to date are empirical studies that substantiate claims made in the literature about the status of -en as the stylistically “higher” variant (for instance, Dammel, Kürschner & Nübling 2010: 608; Taeldeman 1980) and the variants’ regional distributions, as well as a usage-based quantification of their relative occurrence in various domains.

The purpose of this study is to empirically investigate the degree to which Dutch language use reflects the concept of complexity outlined above. The analysis is based on two sets of data. The Corpus of Spoken Southern Dutch Dialects (GCND) was queried for plural doublets (i.e., lemmas varying between -s and -(e)n in the plural). Since phonology is usually presented as the most relevant steering factor in Dutch plural marking, the doublets were subsequently grouped based on formal characteristics:

1. trochaic output with both -s and -(e)n, e.g., gewoonten versus gewoontes ‘habits’; professoren versus professors ‘professors’
2. trochaic output with -s only, e.g., appelen versus appels ‘apples’; leraren versus leraars ‘teachers’
3. trochaic output with -en only, e.g., trammen versus trams ‘trams’; redacteuren versus redacteurs ‘editors’

These groups were first enriched with additional lemmas that the Van Dale dictionary (den Boon & Hendrickx 2022) lists as doublets for Standard Dutch. Then, the token frequency of each word form was ascertained on the basis of the OpenSoNaR corpus. Since this corpus includes both Netherlandic and Belgian Dutch varieties in addition to metadata on text types, both country and register (i.e., text types such as legal texts vs. chats) could be included in the analysis as relevant variables. 

The findings indicate that there is not one clear factor that conditions the use of a variant. Rather, their occurrence depends not only on (an interplay of) the factors register and region, but also on animacy. On top of these factors come apparent (intra)individual and lexical idiosyncrasies in addition to the degree of phonographic integration of a loan word. Hence, wherever the Dutch plural system does not allow for variation, it can certainly be regarded as less complex than the German one. However, the plural doublets show that the conditioning factors are much less straightforward than in German: in this sense, Dutch is highly complex.