Research Dirk Pijpops

Abstract

Constructional contamination occurs when a morphosyntactic alternation becomes lexically skewed under the influence of another construction. A well known example is the Dutch partitive genitive, where an indefinite pronoun is followed by an attributive adjective that may appear with or without an -s ending, e.g. 'iets verkeerd(s)' 'something wrong' (Pijpops & Van de Velde 2016). Some adjectives frequently occur in superficially similar but structurally different contexts, such as 'iets verkeerd interpreteren' 'to interpret something the wrong way', where 'verkeerd' 'wrong' is an adverb and therefore never takes -s. The more often an adjective appears in such non genitival contexts, the more speakers tend to drop the -s in genuine partitive genitives. This effect has been documented across several constructions in Dutch and English (Pijpops, De Smet & Van de Velde 2018; Van de Velde & Pijpops 2018; Hilpert & Flach 2022; Bouso 2022; Delaby & Colleman 2026). Constructional contamination runs counter to the principle of ambiguity avoidance. If speakers aimed to avoid ambiguity, they would be expected to use the -s more often precisely with adjectives like 'verkeerd' 'wrong', where confusion with an adverbial reading is likely. Instead, speakers appear willing to tolerate additional ambiguity so they may reuse familiar lexical combinations stored in memory. So far, research has focused on cases where ambiguity is relatively harmless. In sentences like 'Ze zijn bang dat ze iets verkeerd zullen doen' 'they are scared to do something wrong', misinterpreting 'verkeerd' 'wrong' as an adverb does not seriously hinder communication. However, ambiguity becomes more problematic in domains such as argument structure, where distinguishing who did something to whom is crucial. Wasow (2015) shows that argument structure is one of the few areas where languages consistently avoid ambiguity. This project therefore examines whether constructional contamination also affects variation in argument structure, focusing on personal pronouns in Dutch and English. In both languages, traditional object forms may appear as subjects. Examples include the Dutch hun-subject, hem-subject and the English them-subject. If constructional contamination plays a role, the former object forms should occur more often as subjects with verbs that frequently take them as objects, such as 'zien' 'see'. Frequent combinations like 'zag hem' 'saw him' may become entrenched in memory, increasing the likelihood of hem appearing as a subject. Ambiguity avoidance predicts the opposite: verbs that often take 'hem' 'him' as object should discourage the use of these forms as subjects, because ambiguity would be greatest there.

Researcher(s)

• Promoter: Pijpops Dirk

Research team(s)

• Centre for Computational Linguistics, Psycholinguistics and Sociolinguistics (CLiPS)

Funding

• BOF

Project type(s)

• Research Project

Abstract

PASTA (Premodern Antwerp Spelling Traditions Analysed) examines how late medieval and early modern Antwerp scribes encoded speech in writing, which phonetic and other principles motivated those practices, and how the implementation of these principles evolved over several centuries. Spelling systems are governed by the interacting principles of pronunciation, uniformity, analogy, etymology and graphotactics. For Dutch, there is still no large-scale, diachronic, token-based study of a single urban writing centre that jointly analyses these principles and the hierarchy of phonetic features expressed in spelling. PASTA fills this gap with a four-century corpus of Antwerp 'schepenbrieven' (aldermen's charters, 13th–16th c.), combining existing materials with circa 200 newly transcribed charters from the Felixarchief, all uniformly lemmatised and PoS-tagged. A dedicated grapheme-phoneme correspondence layer, implemented in an open-source Python module, supports quantitative and qualitative analyses of feature consistency through time. By comparing Antwerp outcomes with later norms, the project reconstructs an "alternative history" of Dutch orthography.

Researcher(s)

• Promoter: De Wulf Chris
• Co-promoter: Pijpops Dirk

Research team(s)

• Centre for Computational Linguistics, Psycholinguistics and Sociolinguistics (CLiPS)

Funding

• BOF

Project type(s)

• Research Project

Abstract

Mechanisms of language variation and change often take as starting point lexical biases in grammatical variation, i.e. the finding that particular words engender speakers to prefer one construction over another while forming utterances. For example, a frequent word that is biased towards a construction may 'rub off' its meaning to the construction itself. What is unclear, however, is how such lexical biases develop in the first place. To understand this, the project introduces two mechanisms that can create such lexical biases, viz. exemplar-driven and index-driven lectal contamination. Both mechanisms start from language contact between two varieties of the same language, but differ in how such contact leads to lexical biases within the varieties. Exemplar-driven contamination relies on the cognitive storage of exemplars, while index-driven contamination assumes that the words and constructions act as social indices. A pilot study focusing on nominal morphological variation has already been completed, with positive results. The project will conduct three more corpus-based case studies that test the effect of both mechanisms among other types of variation. Next, I will build an agent-based simulation of each mechanism. This will allow us to validate both mechanisms in-silico and derive exact theoretical predictions for each mechanism. Finally, these predictions will be put to the test through corpus research and a forced-choice and receptive experiment.

Researcher(s)

• Promoter: Pijpops Dirk

Research team(s)

• Centre for Computational Linguistics, Psycholinguistics and Sociolinguistics (CLiPS)

Funding

• FWO

Project type(s)

• Research Project

Abstract

Dutch features both two-place verbs with direct and indirect objects where the use of the preposition aan 'to' is optional, e.g. respectively 'ik bouw (aan) een konijnenverblijf' ('I'm building (to) a rabbit shelter') en 'het contract ontglipte (aan) ons bedrijf' ('the contract slipped away from our company'). There are also a number of verbs whereby the status of the object is unclear, e.g. 'hij gehoorzaamt (aan) de heilige wet' ('he obeys the holy law'). This project aims to find out (i) when and why language users choose to employ or omit the preposition aan 'to' among these verbs, and (ii) which objects behave more like direct or indirect objects. The answer to the first question will also inform the second: for which verbs does the alternation behave more like the dative alternation (e.g. 'Sophia geeft (aan) hem een dikke knuffel' 'Sophia gives him a big hug') or more like the transitive-prepositional alternation (e.g. 'Frederik zoekt (naar) zijn vrachtwagen' 'Frederik is searching for his lorry'). These questions will be dealt with through in-depth corpus research and experiments.

Researcher(s)

• Promoter: Pijpops Dirk
• Fellow: Van Herpe Alexander

Research team(s)

• Centre for Computational Linguistics, Psycholinguistics and Sociolinguistics (CLiPS)

Funding

• BOF

Project type(s)

• Research Project