Glossa Psycholinguistics

This study investigates whether crosslinguistic constraints on subject selection in physical causative constructions extend to causative psychological verbs (psych verbs, e.g., frighten, surprise), with a focus on subject volitionality. According to Wolff et al.’s (2009) initiator hypothesis, languages tend to restrict subjecthood in causative events to entities that can plausibly initiate a causal chain. While this has been established for physical causatives, it remains unclear whether similar constraints apply in psychological causation. To test this, we conducted an Acceptability Judgment Task in which native speakers of Korean, Mandarin Chinese, and English rated grammatical sentences varying in subject volitionality. The results showed that only Korean speakers consistently dispreferred non-volitional subjects, suggesting that their subject selection is more constrained by volitionality. These findings indicate that the initiator hypothesis extends beyond physical causatives to psych verbs and that crosslinguistic variation in subject selection persists across domains.

Recent work on sound symbolism has established a relation between roughness and the alveolar trill (e.g., Ćwiek et al., 2024). The available evidence is, however, ambiguous as to whether the phonetic properties of an alveolar trill or the phonological properties of rhotics are associated with roughness. This was tested in Maltese, which uses both alveolar approximants and alveolar trills for /r/. Participants were presented with a product that had either a rough or a smooth surface and were asked to choose a potential product name from two options, one containing an /r/ and the other an /l/. The /r/ was produced either as a trill or an approximant. The /r/-bearing word was chosen more often when the product was rough, but this was independent of the phonetic implementation of the /r/. This indicates that the phonological properties of rhotics, and not the phonetic properties of the trill, support the association of /r/ with roughness.

I examine morphological priming in spontaneous dialogue by tracking the alternation between the long (-jál/-jél) and the short (-j) variant of the 2SG indefinite subjunctive in Hungarian task-oriented conversations, to see whether members of dyads persist and converge in their choice of suffixed variants. I combine conversational data from the Akaka Maptask and Budapest Games corpora with distributional estimates from the Hungarian Webcorpus, resulting in a dataset of 334 subjunctive tokens. Hierarchical Bayesian generalised linear models show robust within-verb priming: following a long prime, a long target is about 3x more likely, with no comparable effect across different verbs. Speaker identity and prime–target distance did not improve model fit. Independent of priming, conversational choices track corpus baselines, indicating an additive contribution of lexical distributions. The results provide naturalistic evidence for the priming of inflected morphological variants in Hungarian, and clarify how lexical distributions scaffold variant choice in conversation.

Structure generation accounts of ellipsis propose that silent syntactic structure is generated but unpronounced in ellipsis. Reactivation accounts propose instead that the antecedent is reactivated at the ellipsis site, with no silent structure. Existing psycholinguistic findings have not distinguished these hypotheses (Phillips & Parker, 2014). We address this issue by investigating voice mismatches in VP ellipsis in English (e.g., A new language was supposed to be taught in the fall semester, but no professors could __ because the schedule was too tight). In Experiment 1, participants' performance at recognizing that a word did not appear in the sentence is poorer when a related word appeared (e.g., teach, when taught appeared), versus an unrelated word (e.g., worst). This effect is greater when the elided clause is in the active voice, as in the previous example, versus the passive (...but no new language was...). The contrast between active and passive conditions cannot be accounted for solely in terms of antecedent reactivation, since the antecedents are identical. Experiment 2, which compares sentences with VP ellipsis with sentences containing an overt pro-form (e.g., do that...), provides further evidence in support of this conclusion. The processing of VP ellipsis must include more than just antecedent reactivation. The findings of these two experiments are thus most consistent with the structure generation account of ellipsis.

This study aims to explore crosslinguistic variation in island sensitivity between Danish and English, specifically focusing on the extraction patterns of different types of adjunct clauses. Additionally, it investigates how context, dependency type, information structure and syntactic structure may impact the acceptability of extraction from adjunct clauses. The results indicate that Danish exhibits acceptability patterns similar to those found in Swedish and Norwegian regarding extraction from conditional and causal adjunct clauses. However, the raw scores of extraction from adjunct clauses are unexpectedly low in Danish. The study discusses potential explanations involving factors related to syntax, discourse function and processing but finds limitations in each approach and concludes that a fine-grained account is needed in order to capture the observed variability. This research contributes to a deeper understanding of island phenomena and highlights the need for investigation involving direct comparisons between languages.

The present study investigates whether the relationship between voice gender (female and male voices) and gender-associated meanings (e.g., flower [female] vs. football [male]) influences learning of an artificial lexicon after brief auditory exposure. Participants were trained on 16 novel words in either a gender-aligning condition (i.e., female voices produced novel words with female-associated meanings and male voices produced novel words with male-associated meanings) or a gender-mismatching condition (e.g., female voices produced words with male-associated meanings, etc.). Listeners either heard voices that contain stereotypical or non-stereotypical gross acoustic patterns. Listeners showed worse word learning in gender-mismatching conditions. And, learning performance was mediated by voice stereotypicality. This study demonstrates how gender associations implicitly modulate social attention when learning a new language. Also, since the experiment was under the guise of a language learning application, the results also have implications for gender biases in human-computer interaction and speech technology. More broadly, these findings demonstrate the impact of gender associations on speech perception and language learning. 

Typological data show a tendency for languages to exhibit harmonic (i.e., consistent) ordering between heads and dependents. Previous experimental work using artificial language learning experiments has shown that learners prefer harmonic patterns. This suggests that the typological trend for harmony may, at least in part, be driven by a cognitive bias. However, it is well-documented that specific categories sometimes contradict this tendency. Here, we investigate one such case in the domain of the noun phrase. While many nominal dependents exhibit harmony, adjectives and genitives do not: adjectives tend to follow the noun and genitives tend to precede. Previous experiments have identified the existence of cognitive biases that keep these dependents split across the head noun in contexts where there is no conventional language system in place. In this study, we use a silent gesture experiment to examine whether the specific biases that apply to these two dependent types compete with a general preference for harmony in an artificial language learning task. Specifically, we examine whether participants’ learning behaviour is consistent with a preference not just for harmony, but for a non-harmonic order where adjectives follow and genitives precede the noun. What we find, instead, is that participants’ preference for consistent language systems is not modulated by category-specific biases for prenominal genitives and postnominal adjectives. We discuss the implications of this finding for explanations of typological tendencies which link them to cognitive biases.

We investigated the processing of resumptive pronouns in English relative clauses. We asked whether resumptive pronouns in complex relative clauses formed from islands and non-islands establish dependencies with head nouns in incremental processing. Based on the findings from two reading time experiments using the Maze task with the gender-mismatch paradigm, as well as an off-line comprehension study, we conclude that (i) a dependency is formed with the head noun immediately upon encountering the resumptive pronoun; (ii) dependency formation is easier across islands than across non-islands; (iii) dependency formation privileges the head noun over intervening noun phrases; and (iv) these dependencies are maintained in off-line comprehension. The data suggest that although resumptive pronouns are judged to be unacceptable in English, comprehenders automatically and immediately integrate them using routines similar to filler-gap dependencies. We discuss the implications of our findings for the question of whether resumptive pronouns facilitate filler dependency formation in comparison to gaps, a possible reason why dependencies would be easier to establish across islands than across non-islands, and how the use of resumptive pronouns may benefit listeners as well as speakers.

This study investigates how English and Mandarin Chinese speakers linguistically encode transfer events (e.g., give, take). We test the hypothesis that transfer events, like motion events, involve Goal and Source paths and are therefore subject to a Goal bias, such that Goals are encoded more prominently than Sources. Across two experiments, speakers of both languages described transfer events. In both languages, descriptions showed a robust Goal bias: speakers mentioned Goals more often than Sources, even when controlling for potential Agent–Subject mapping effects, and they were more likely to use canonical Goal-encoding devices for Goals than canonical Source-encoding devices for Sources. These findings reveal a cross-linguistically reliable preference for Goals in transfer-event descriptions, consistent with the possibility that Goal prominence reflects general principles of event message planning. More broadly, the results provide support for accounts that extend Goal/Source path representations beyond spatial language.

The interpretation of round numerals inspired two competing hypotheses. The first, the preferred-approximation hypothesis, claims that round numerals are more likely to be interpreted as approximate, while sharp numerals are interpreted as precise by default (Krifka, 2002, 2007). This hypothesis is based on the idea that addressees recognize that speakers can use round numerals for approximation more readily than sharp ones, and, therefore, derive the approximate interpretation through rational inferencing. The second hypothesis, the precise-by-default hypothesis, holds that all numerals, round and sharp alike, are interpreted as precise by default (Ariel, 2021; Ariel & Levshina, 2024). The present study investigates the relationship between round counting numerals and approximation through four experiments that directly compare round and sharp numerals across different experimental settings. Our findings show that participants do recognize speakers’ license to use round numerals imprecisely more than sharp ones. However, this preference does not lead them to actually interpret round numerals imprecisely. We offer several explanations for this gap.

While bias in polar questions has generated an increasing body of cross-linguistic research, relatively little is known about their processing. Adapting the experimental design of Tian et al. (2016, 2021), we conducted a visual world eye-tracking study of assertions, polar questions with and without verb inversion, and polar questions containing particles razve and snad, respectively, in Russian and Czech. While participants heard small dialogues, their fixations on pictures showing either the positive outcome implied in the stimulus or its opposite were recorded. While, in assertions, participants focused their attention on the picture corresponding to the expressed polarity, fixations in questions became concentrated on the positive picture, largely independently of the polarity expressed in the stimulus. In questions with particles razve and snad, attention remained balanced between the positive and the negative pictures. We argue that the preference for the positive picture in questions is related to pragmatic processing, specifically, to the influence of a positive epistemic bias in polar questions.

Sentences are not just mere strings of words or signs, but manifest a complex internal structure. Linguistic research has demonstrated that sign languages and spoken languages both exhibit hierarchical constituent structure which determines how individual elements in a sentence relate to each other. Here, we report the first adaptation of the psycholinguistic click paradigm, which aims to demonstrate an impact of hierarchical constituent structure on cognitive processing load during auditory language processing, to the visuo-spatial modality of sign languages. We performed two independent online experiments: The main experiment with a group of 53 deaf signers using German Sign Language (DGS) as their primary means of communication and a control experiment with a group of 53 hearing non-signers. Both groups were shown videos of syntactically complex sentences in DGS. A white flash (mimicking the click in the auditory domain) to which participants had to respond could occur as an overlay to the video at different levels in the constituent structure. Our pre-registered inferential analyses yielded no effect for our syntactic manipulations, neither in the group of signers nor in the group of non-signers. Additional exploratory analyses suggest general effects of attention during the processing of communicative signals, as even the group of non-signers' behaviour was influenced by non-manual cues (i.e., eyebrow raising), despite their lack of knowledge of DGS. We conclude that the simultaneous and time-shifted presence of different syntax-relevant cues (i.e., hands, mouthings, and non-manuals) makes the processing of the sign stream robust against disruption by extraneous visual signals and argue that non-signers attend to some non-manual cues, due to their resemblance to communicative gestures independent of language.

(* Trettenbrein and Maran share first authorship.)

Speech perception and production are linked through shared neural representations, such that individual differences in auditory-perceptual acuity predict articulatory precision for segmental contrasts. Whether similar perception-production links (PP-links) extend to suprasegmental prosodic features, however, remains unclear. In the present study, we investigated whether individual differences in auditory-perceptual acuity for prosodic boundary cues (pitch range, pause duration, final lengthening) predict how those same cues are produced, and whether this relationship is modulated by cognitive load. Using Bayesian distributional models, we examined how auditory-perceptual acuity modulated both the distinctiveness (mean cue differences between bracket and no-bracket productions) and trial-to-trial variability of prosodic boundary cue production. The results revealed cue-specific PP-links. For pitch, higher auditory-perceptual acuity was associated with stronger modulation of the produced pitch range, and it further modulated trial-to-trial variability in pitch range under cognitive load: under high cognitive load, speakers with higher pitch acuity maintained variable pitch range patterns across all productions, whereas speakers with lower pitch acuity showed reduced pitch range variability. For pause duration, higher auditory-perceptual acuity was associated with greater trial-to-trial variability in pause duration and, exploratorily, with reduced reliance on pauses for boundary marking. Final lengthening showed no relationship between auditory-perceptual acuity and production. Across cues, cognitive load systematically affected both production distinctiveness and trial-to-trial variability, interacting with auditory-perceptual acuity in a cue-specific manner. These findings demonstrate a PP-link at the prosodic level that is not uniform across prosodic boundary cues and suggest that, particularly for pitch, higher auditory-perceptual acuity supports flexible and adaptive prosodic cue production.

(*Veríssimo and Wartenburger share senior authorship.)

A key question for psycholinguistic research is understanding how and when structure-sensitive constraints guide the real-time resolution of pronouns. In this study, we examined the processing of pronouns in crossover configurations, where a wh-operator moves across a pronoun to its left, rendering it structurally inaccessible to the pronoun. Our findings from self-paced reading experiments regarding strong crossover (SCO) and weak crossover (WCO) configurations suggest that in both cases, native English speakers only consider structurally accessible antecedents upon encountering the pronoun, adhering to structure-sensitive constraints. Nevertheless, the reading times and post-reading acceptability judgments indicate that the structurally inaccessible wh-operator can be considered during later processing stages, albeit exclusively in the WCO configuration. We discuss the implications of these findings for sentence processing models of pronoun resolution and for theories explaining the crossover phenomena. 

Results from previous corpus and untimed judgment studies indicate that the distribution and acceptability of relative clause extraposition (RCE) in German are modulated both by the extraposition distance and the length of the extraposed relative clause. The observed complexity effects have been argued to reflect processing efficiency constraints, such as a short-before-long ordering preference or a bias towards optimizing dependency lengths within a sentence. However, it is still unclear whether such processing-related biases affect the perception and production of RCE in the same way or differently, and how complexity measures interact with other factors that are known to modulate processing difficulty, such as relative clause expectancy. The present study used speeded acceptability judgment and timed sentence generation tasks to examine the processing and production of RCE in German. Our results show that complexity measures affect RCE perception and production in a similar way, with smaller VP-to-RC length ratios linked to higher RCE acceptability and increased production likelihood. Relative clause expectancy affected participants’ response times, but not RCE acceptability. Despite the high overall acceptability of RCE in German, even long relative clauses were preferably positioned adjacent to their head noun in our production task, however. This suggests that a bias towards minimizing the distance between an RC and its head noun, rather than a short-before-long ordering preference, is the primary constraint on German speakers’ production choices.

Many statistical models of properties of words in texts and in running speech use as their outcome variables properties of word tokens, rather than properties aggregated over word types. Here, we describe methodological and conceptual problems with token-level (or token-based) models, including assumption violations of the models typically used, model redundancy, and distortions of parameters estimating lexical effects – i.e. the very target of what such models seek to capture. We show that these problems persist – and in some ways, grow more extreme – when we follow widely-used practices for specifying random effects structure. We do so by comparing models of spoken word duration of homophones in the Switchboard corpus, a dataset that has been discussed and modeled extensively. We discuss possible remedies and recommendations.