Constructional contamination in morphological and syntactic variation

Constructions that are structurally unrelated, occasionally give rise to strings that are superficially alike. As language users employ shallow parsing (Dąbrowska 2014), such structural differences are not consistently uncovered. As a result, these superficially resembling constructions may quantitatively affect the realization of one another. For instance, Dutch partitive genitives occur both in a variant with and without an -s ending, as in (1), and the use of variant without -s is boosted by the occurrences of superficially similar adverbial constructions that consistently occur without -s, as in (2). Concretely, partitive genitives containing lexemes that often occur in adverbial constructions, were shown more often to exhibit the variant without -s, even when only looking at clearly unambiguous partitive genitives (Pijpops and Van de Velde 2016). This effect is called constructional contamination. 1. Ik heb iets verkeerd(s) gegeten. ‘I have eaten something wrong.’ 2. Ik heb iets verkeerd geïnterpreteerd. ‘I have interpreted something in the wrong way.’ So far, constructional contamination has been shown to be operative in the Dutch partitive genitive (Pijpops & Van de Velde 2016), as well as for various other Dutch noun phrase phenomena (Van de Velde and Pijpops 2018), and in the placement of adverbs in the English passive (e.g. ‘The disease was sexually transmitted’ vs. ‘The disease was transmitted sexually’, Hilpert & Flach forthc.). We now investigate three more case studies from Dutch morphological and syntactic variation. The first case study pertains to variation in verbal clusters, as in (3), which shows an alternation between the PARTICIPLE-AUX order and the AUX-PARTICIPLE order. This alternation is multifactorially driven (De Sutter 2005; Bloem, Versloot and Weerman 2017). When the participle is an adjective in a copular construction, however, only the PARTICIPLE-AUX order is possible. We show that the more often a participle is used as an adjective – relative to its usage as a verb – the more often it prefers the PARTICIPLE-AUX order, even among its unambiguously verbal occurrences. As expected, this effect was stronger and clearly significant among the instances with the auxiliaries zijn ‘be’ and worden ‘become’, which exhibit direct formal resemblance to the adjectival participles in the copular construction, while it was weaker among the instances with the auxiliary hebben ‘have’, where the resemblance is more indirect. In fact, among the latter instances, the effect no longer reaches significance in the regression model. 3. …dat de situatie door hem verziekt wordt / wordt verziekt. ‘…that the situation is being screwed up by him.’ The second case study is the competition between weak and strong past tense inflection, e.g. graafde vs. groef ‘digged’ (cf. Vosters 2012: 244, Fn. 16, De Smet and Van de Velde 2019) and the third case study concerns the usage of long vs. bare infinitives with the posture auxiliaries staan ‘stand’, zitten ‘sit’ and liggen ‘lie’ (Lemmens 2005; space limitations preclude full discussion in this abstract). In both cases, we will offer quantitative evidence that superficially similar constructions exert contamination effects on the alternations.status: publishe