Peter de Swart’s research while affiliated with Radboud University and other places

Op de basisschool moeten leerlingen enig inzicht opdoen in taal: hoe zit taal in elkaar, hoe verandert taal en hoe kun je reflecteren op taalgebruik? Om deze taalbeschouwingsvaardigheden te ontwikkelen, is grammaticaonderwijs onontbeerlijk, maar dan niet op de traditionele manier. Jimmy van Rijt en Peter de Swart laten zien hoe het grammaticaonderwijs op een verrijkte manier kan worden vormgegeven.

Neural language models (NLMs) have frequently been tested on their ability to model grammatical phenomena in a "human-like'' way, but their behavior is often not compared to actual human data, making it unclear whether their behavior is actually human-like. An example of this can be seen with long distance filler-gap dependencies and their island constraints: Wilcox et al. (2018) devised a method to directly test whether NLMs can model these phenomena in a human-like way, and this method is accompanied by specific predictions about how models should perform. These predictions are, however, not all supported by experimental research with human participants, making it difficult to conclude that the NLMs can model these phenomena comparably to humans. Consequently, in the current study, we investigated whether the predictions made by Wilcox et al. (2018) are supported by human data by testing both a Long Short-Term Memory language model and human participants in English with the method devised by Wilcox et al. (2018), and comparing their results. The results showed that the model might exhibit "human-like'' behavior according to the predictions of Wilcox et al. (2018), but this behavior was not fully comparable to that of humans. Therefore, this study not only shows that it is important to obtain human data to validate the predictions of this specific method, but also that predictions of "human-like'' behavior for NLMs on other grammatical phenomena should always be grounded in data from human experiments.

The current study investigates whether a Long Short-Term Memory (LSTM) network can learn the wh-island constraint in Dutch in a way comparable to human native speakers. After establishing with an acceptability judgement task that native speakers demonstrate a clear sensitivity to wh-island violations, the LSTM network was tested on the same sentences. Contrary to the results of the native speakers, the network was not able to recognize wh-islands and to block gap expectancies within them. This suggests that input and the network’s inductive biases alone might not be enough to learn about syntactic island constraints, and that built-in language knowledge or abilities might be necessary.

The current study investigates whether a Long Short-Term Memory (LSTM) network can learn the wh-island constraint in Dutch in a way comparable to human native speakers. After establishing with an acceptability judgement task that native speakers demonstrate a clear sensitivity to wh-island violations, the LSTM network was tested on the same sentences. Contrary to the results of the native speakers, the network was not able to recognize wh-islands and to block gap expectancies within them. This suggests that input and the network’s inductive biases alone might not be enough to learn about syntactic island constraints, and that built-in language knowledge or abilities might be necessary.

An eye-tracking experiment was conducted with speakers of Dutch (N = 84, 36 male), a language that falls between grammatical and natural-gender languages. We tested whether a masculine generic pronoun causes a male bias when used in generic statements—that is, in the absence of a specific referent. We tested two types of generic statements by varying conceptual number, hypothesizing that the pronoun zijn “his” was more likely to cause a male bias with a conceptually singular than a conceptually plural antecedent (e.g., Someone (conceptually singular)/Everyone (conceptually plural) with perfect pitch can tune his instrument quickly). We found male participants to exhibit a male bias but with the conceptually singular antecedent only. Female participants showed no signs of a male bias. The results show that the generically intended masculine pronoun zijn “his” leads to a male bias in conceptually singular generic contexts but that this further depends on participant gender.

An eye-tracking experiment was conducted with speakers of Dutch (N=84, 36 male), a language that falls in between grammatical and natural gender languages. We tested whether a masculine generic pronoun causes a male bias when used in generic statements, that is, in the absence of a specific referent. We tested two different types of generic statements by varying conceptual number, hypothesizing that the pronoun "zijn" ‘his’ was more likely to cause a male bias with a conceptually singular than a conceptually plural antecedent (e.g., "Someone (CONCEPTUALLY SINGULAR)/Everyone (CONCEPTUALLY PLURAL) with perfect pitch can tune his instrument quickly"). We found male participants to exhibit a male bias, but with the conceptually singular antecedent only. Female participants showed no signs of a male bias. The results show that the generically-intended masculine pronoun "zijn" ‘his’ leads to a male bias in conceptually singular generic contexts, but that this further depends on participant gender.

The results of a self-paced reading experiment show that reading times in Dutch increase when there is a gender mismatch between the subject and a subsequent possessive pronoun, signaling an increase in processing difficulty. We hypothesized that Dutch learners of French incorrectly apply the rules of their L1 in their L2 and should therefore also show an increase in reading times in French upon encountering a possessive pronoun for which grammatical gender differs from the biological gender of the subject (the possessor). At the same time, we expected that they would have no or less difficulties in processing ungrammatical French sentences in which the biological gender of the subject/pos-sessor matches the gender of the possessive pronoun. We did not find either of these effects in a second self-paced reading experiment. We assume that the Dutch learners of French parse the foreign language sentences in a shallow fashion .

Acceptability judgements of syntactic island violations are often claimed to improve by either increasing the complexity of the wh -filler phrase or integrating the violating sentence into a discourse. In two acceptability judgement tasks, we looked at wh -island violations in Dutch by varying the complexity of the filler phrase and by presenting the sentences either in isolation or with a preceding discourse. We found that neither variable had a significant effect in isolation, but that only in their combination a significant effect was observed. The same effect showed up in non-island conditions, however. This is in contrast to findings in the literature on English and French and suggests that the complexity effect in Dutch is not syntactic. We therefore conclude that wh -islands are strong islands in Dutch ( Broekhuis & Corver 2015 ) and show that the contrast with English and French can be made to follow from featural Relativized Minimality ( Rizzi 2017 ), taking into account the verb second property of Dutch.

Animacy, commonly defined as the distinction between living and non-living entities, is a useful notion in cognitive science and linguistics employed to describe and predict variation in psychological and linguistic behaviour. In the (psycho)linguistics literature we find linguistic animacy dichotomies which are (implicitly) assumed to correspond to biological dichotomies. We argue this is problematic, as it leaves us without a cognitively grounded, universal description for non-prototypical cases. We show that ‘animacy’ in language can be better understood as universally emerging from a gradual, cognitive property by collecting animacy ratings for a great range of nouns from Japanese and Persian. We used these cognitive ratings in turn to predict linguistic variation in these languages traditionally explained through dichotomous distinctions. We show that whilst (speakers of) languages may subtly differ in their conceptualisation of animacy, universality may be found in the process of mapping conceptual animacy to linguistic variation.