Glossa Psycholinguistics

The illusory licensing of negative polarity items has been an insightful phenomenon for accounts of human sentence processing, as its extreme selectivity has proven problematic to explain in terms of parsing principles that underlie the establishment of other item-to-item dependencies. Using speeded acceptability judgments, I provide novel experimental evidence that the NPI illusion may be restricted to a particular type of NPI—illusory licensing was replicated for German jemals 'ever', but was not confirmed for the attenuating NPI so recht 'really'. I argue that this finding challenges all current accounts of the NPI illusion, and propose an explanation that purports an interaction between a scalar NPI licensing mechanism and scalar properties of the illusory licensing context as the source of the NPI illusion.

Truth-value judgments are one of the most common measures in experimental semantics and pragmatics, yet there is no standardized way to elicit such judgments. Despite anecdotal remarks on how proper choice of prompts or response options could help disentangle pragmatic from semantic effects, little is known regarding the relation between parameters of the task and what it actually measures. We tested a range of prompts and two response options for their sensitivity to truth of the target sentence, prior evidence, and the interaction between these two factors. We found that participants attribute high value to true statements, even when they are not backed by evidence. Moreover, our results confirm that prompts vary wildly in their sensitivity to pragmatic factors, and should allow researchers to make an informed choice depending on what they want to test. There was no difference between the results generated by the response options, although the Likert scale required fewer participants and may therefore be preferable. In addition, we discuss some theoretical consequences of our results for pragmatics, philosophy of language, and social psychology. 

This article has been retracted by mutual agreement of the author and editors. After some investigation, the authors determined that unintentional error led to the presentation of theoretical conclusions that contradict important findings reported in the article, and it was determined that this could not be resolved with a minor correction.

The present study tested whether listeners hearing one form of a morpheme activate other forms of the same morpheme. Listeners performed lexical decisions while hearing Mandarin monosyllables; crucially, critical targets could be primed by related syllables that occurred 18–52 trials earlier (long-lag priming). The use of long-lag priming ensures that any facilitation effects are due to morphological relatedness and not to semantic or form relationships, which do not prime lexical decisions at long lags. Across three experiments (total N = 458), we consistently found that lexical decisions were primed when the same pronunciation of a morpheme occurred as prime and target (e.g., shiL – shiL) but were not primed when two different variants of the same morpheme occurred as prime and target (e.g., shiR – shiL, where both of these syllables are potential pronunciations of the same morpheme). In other words, we observed identity priming but not morphological priming, unlike other long-lag priming experiments, which almost invariably observe intramodal morphological priming if they test it. This surprising finding suggests that there are boundary conditions on the elicitation of long-lag morphological priming effects.

When comprehenders predict a specific lexical noun in a highly constraining context, they also activate the grammatical features, such as gender, of that noun. Evidence for such lexically mediated prediction comes from ERP studies that show that comprehenders are surprised by adjectives and determiners that mismatch the features of a highly predictable noun. In this study, we investigated whether comprehenders can (i) predict an abstract noun phrase in an upcoming argument position (without pre-activating a specific lexical item) and (ii) assign morphosyntactic features to the head noun of that phrase. To do so we used the processing of Dutch cataphors as a test case. We tested whether seeing a cataphor in a preposed clause triggered a prediction of a feature-matching antecedent NP in main subject position. If comprehenders predicted a feature-matching subject, we reasoned that they should also expect an agreeing main verb, which comes before the subject because Dutch is a V2 language. A single-word prediction experiment showed that comprehenders expect a main verb matching the number of the cataphor. In a follow-up self-paced reading experiment, we found a number-mismatch effect if the V2 main verb did not agree with the cataphor. We take the results as evidence that comprehenders predicted a matching antecedent in subject position. We argue that the results are better explained as involving prediction of an abstract noun phrase marked for morphological features, rather than a specific lexical item.

Prior research has shown that sentences with noncanonical argument order (e.g., patient-before- agent instead of agent-before-patient order) are associated with additional online processing difficulty, but that this difficulty can be alleviated if the discourse context licenses noncanonical order. Other studies demonstrated that noncanonical sentences are prone to misinterpretation effects: comprehenders sometimes seem to form interpretations with incorrect assignments of semantic roles to argument NPs. However, those studies tested noncanonical sentences in isolation. To further clarify the source of misinterpretation effects, we designed three experiments that investigated how discourse properties licensing noncanonical order affect online processing and final interpretation. All experiments tested unambiguous active declarative sentences in German with agentive verbs and two arguments, probing both online processing difficulty (using selfpaced reading) and accuracy of interpretation (using wh-comprehension questions). Besides word order (subject-before-object, SO vs. object-before-subject, OS), we varied the context preceding the target sentence (neutral context vs. context licensing OS, Experiment 1), the type of NP serving as object (definite vs. demonstrative NP, Experiment 2) and the type of question probing comprehension (two-argument vs. one-argument wh-questions, Experiment 3). Consistent with earlier findings, we observed that discourse properties licensing OS order facilitated online processing in early sentence regions. However, they did barely affect accuracy on comprehension questions, with accuracy instead being a function of word order and question type. Our results support models that explain misinterpretation effects in terms of task-specific retrieval processes. A retrieval mechanism capturing the effects of question type is proposed.

Studies on the reading of appositive relative clauses (ARCs) have found that ARCs seem to exhibit less influence in later parsing and decision-making than similar constructions (Dillon et al. 2014, 2017), a pattern we call discounting. Existing accounts often link discounting to the status of ARCs as independent segments in systems of pragmatic representation. This would predict discounting for other constructions as well. In this study, we test that prediction by investigating the reading of direct discourse speech reports and causal adjuncts in English. Diagnostics supplied by the theoretical literature show that these constructions contribute the same independent segments as ARCs in two different systems of pragmatic organization: direct discourse reports contribute an independent speech act, and causal adjuncts contribute their own discourse units. Nevertheless, in a series of five experiments, we find no evidence of ARC-like discounting for either. We conclude that discounting should not be linked to either of these pragmatic representations, and discuss the outlook for other approaches to the phenomenon.

This work tests the relative role of perception- and production-based predictors, and the relationship between them, in imitation of artificial accents varying in voice onset time (VOT), using a paradigm designed to target distinct sub-processes of imitation. We examined how explicit imitation of sentences differing systematically in voice onset time (VOT) was influenced by the type of VOT manipulation (lengthened vs. shortened) and by the presence vs. absence of voice-related variability in exposure. In contrast to previous work, participants imitated shortened as well as lengthened VOT, albeit with both qualitative and quantitative differences across the two manipulation types. The presence of voice-related variability inhibited imitation, but this inhibition was mitigated by a preceding session with no voice-related variability (i.e., sentences were acoustically identical except for VOT). We then tested the extent to which individual performance on the accent imitation task was related to performance on three other tasks: 1) discrimination of the target accents, 2) imitation of words in isolation drawn from a VOT continuum, and 3) discrimination of these same words. Performance on the accent discrimination task and the word-level imitation task, but not the word-level discrimination task, were independently predictive of accent imitation. Results are consistent with a conceptualization of explicit imitation as the sum of automatic phonetic convergence processes overlaid with distinct, controlled perceptual and articulatory factors that pattern differently across individuals. Phonetic imitation should not be considered as a monolithic skill, and models predicting variation in imitative ability must consider not only the potential sources of individual variability, but also at what level these sources of variability exert their influence. 

This study investigates the real-time processing of Chinese reflexives ziji and ta-ziji in discourse when multiple constraints are involved. Our primary goal is to examine the time course of syntactic and non-syntactic constraints in reflexive resolution. The Syntactic Filter Hypothesis argues that syntactic cues are prioritized at the early stages of processing, in contrast to the Multiple Constraints Hypothesis which posits that at this stage all streams of information can be recruited. The results of two self-paced reading experiments show that in neutral contexts where no antecedent is discourse-prominent, syntactic locality and verb semantics immediately impact real-time processing of ziji and ta-ziji. Crucially, participants’ processing patterns are also influenced at an early stage by the discourse topical status of the non-local antecedents. Overall, these findings suggest that syntax, verb semantics, and discourse prominence all play important roles at an early stage.

Replications are an integral part of cumulative experimental science. Yet many scientific disciplines do not replicate much because novel confirmatory findings are valued over direct replications. To provide a systematic assessment of the replication landscape in experimental linguistics, the present study estimated replication rates for over 50,000 articles across 98 journals. We used automatic string matching using the Web of Science combined with in-depth manual inspections of 274 papers. The median rate of mentioning the search string “replicat*” was as low as 1.7%. Subsequent manual analyses of articles containing the search string revealed that only 4% of these contained a direct replication, i.e., a study that aims to arrive at the same scientific conclusions as an initial study by using exactly the same methodology. Less than half of these direct replications were performed by independent researchers. Thus our data suggest that only 1 in 1250 experimental linguistic articles contains an independent direct replication. We conclude that, similar to neighboring disciplines, experimental linguistics replicates very little, a state of affairs that should be reflected upon.

Behavioral measures of word-by-word reading time provide experimental evidence to test theories of language processing. A-maze is a recent method for measuring incremental sentence processing that can localize slowdowns related to syntactic ambiguities in individual sentences. We adapted A-maze for use on longer passages and tested it on the Natural Stories corpus. Participants were able to comprehend these longer text passages that they read via the Maze task. Moreover, the Maze task yielded useable reaction time data with word predictability effects that were linearly related to surprisal, the same pattern found with other incremental methods. Crucially, Maze reaction times show a tight relationship with properties of the current word, with little spillover of effects from previous words. This superior localization is an advantage of Maze compared with other methods. Overall, we expanded the scope of experimental materials, and thus theoretical questions, that can be studied with the Maze task.

Listeners systematically extract two types of information from linguistic utterances: information about the world, and information about the speaker – i.e., their social background and personality. While both varieties of content have been widely investigated across different approaches to the study of language, research in pragmatics has mostly focused on the former kind. Here we ask how listeners reason about a speaker’s conversational choices to form an impression about their personality. In three experiments, we show that a speaker’s adherence to, or violation of, the pragmatic principles of Relevance and Informativeness, as well as the reasons underlying these violations, affect the evaluation of the speaker’s personality along the core social dimensions of Warmth and Competence. These findings highlight the value of enriching work in pragmatics with insights from sociolinguistics and social psychology about how people reason about human speech to draw inferences about the identity and personality of their interlocutors.

A long-standing debate in the sentence processing literature concerns the time course of syntactic and semantic information processing in online sentence comprehension. The default assumption in cue-based models of parsing is that syntactic and semantic retrieval cues simultaneously guide dependency resolution. When retrieval cues match multiple items in memory, this leads to similarity-based interference. Both semantic and syntactic interference have been shown to occur in English. However, the relative timing of syntactic vs. semantic interference remains unclear. In this cross-linguistic investigation of the time course of syntactic vs. semantic interference, the data from two eye-tracking during reading experiments (English and German) suggest that the two types of interference can in principle arise simultaneously during retrieval. However, the data also indicate that semantic cues are evaluated with a small timing lag in German compared to English. This cross-linguistic difference between English and German may be due to German having richer morphosyntactic marking than English, resulting in syntactic cues dominating over semantic cues during dependency resolution. More broadly, our cross-linguistic results pose a challenge for the cue-based retrieval model’s default assumption that syntactic and semantic cues are used simultaneously during long-distance dependency formation. Our work also highlights the importance of collecting cross-linguistic data on psycholinguistic phenomena which can potentially advance theory development.

The present paper focuses on pseudopartitive constructions headed by quantifier, collective, or container nouns (like a lot of senators, a group of students, a bottle of pills) followed by a singular or a plural verb. We compared these structures with superficially similar adnominal structures of the form NP1[−PL] prep NP2[PL] (e.g., the level of the lakes is/are) in Italian in an acceptability judgment study (Experiment 1), a forced-choice task (Experiment 2), and an eye tracking reading study (Experiment 3). Two major findings were consistent across all studies. First, verb agreement in pseudopartitives always patterned differently from controls. Second, albeit an overall preference for singular verbs was observed, a gradient difference emerged between adnominal controls and pseudopartitives, and among pseudopartitives headed by different nouns. We explain such variability in terms of the availability of a measure interpretation (e.g., pills in the measure of a bottle vs. a bottle containing pills) which is linked to the type of the pseudopartitive’s head noun. While in non-pseudopartitive adnominal structures only one parse is allowed by the grammar, in pseudopartitives a given head noun may admit or block a structural configuration in which the plural feature of the embedded constituent (e.g., of students, modifying a group) can determine the plurality of the subsequent verb. We conclude that verb agreement in pseudopartitives is a grammatical phenomenon and, as such, it refers to speakers’ grammatical competence and cannot be reduced to agreement attraction of the plural intervener.

Several recent experimental studies have investigated the hypothesis that the passive construction is associated with the semantics “[B] (mapped onto the surface [passive] subject) is in a state or circumstance characterized by [A] (mapped onto the by-object or an understood argument) having acted upon it”. (Pinker, Lebeaux & Frost, 1987). In the present work, we extend this method to a new language, Hebrew, and conduct a Bayesian mixed-effects meta-analytic synthesis which draws together the findings of similar studies conducted for English, Indonesian, Mandarin and Balinese. For Hebrew, we found that native adult speakers’ (N=60) acceptability ratings for passives with each of 56 different verbs were significantly predicted by verb-semantic-affectedness ratings provided by a separate group of 16 native adult speakers. Both for Hebrew and across languages, we found that (a) these semantic-affectedness ratings predict verbs’ acceptability in both passive and non-passive constructions, but (b) the effect is bigger for passives than non-passives. These findings raise the possibility that a passive construction that denotes undergoer-affectedness may approach the status of a semantic universal. 

Findings of previous behavioural studies suggest that the semantic nature of what is known as the ‘masculine generic’ in Modern Standard German is indeed not generic but biased towards a masculine reading. Such findings are the cause of debates within and outside linguistic research, as they run counter to the grammarian assumption that the masculine generic form is gender-neutral. The present paper aims to explore the semantics of masculine generics, relating them to  those  of  masculine  and  feminine  explicit  counterparts.  To  achieve  this  aim,  an  approach  novel  to  this  area  of  linguistic  research  is  made  use  of:  discriminative  learning.  Analysing  semantic  vectors  obtained  via  naive  discriminative  learning,  semantic  measures  calculated  via  linear  discriminative  learning,  and  taking  into  account  the  stereotypicality  of  the  words  under investigation, it is found that masculine generics are semantically much more similar to masculine explicits than to feminine explicits. The results presented in this paper thus support the notion of a masculine bias in masculine generics. Further, new insights into the semantic representations of masculine generics are provided and it is shown that stereotypicality does not modulate the masculine bias.

Despite the increased use of different types of gender-fair forms in French, studies investigating how they are interpreted when presented in a sentence remain few. To fill this gap, we conducted a pre-registered study using a timed sentence evaluation task to examine the possibility of speakers’ establishing an anaphoric relationship between a gendered anaphoric expression (femmes ‘women’ or hommes ‘men’) and non-stereotyped role nouns as antecedents. The antecedents were presented in their masculine form or in one out of three different gender-fair forms (complete double forms: les voisines et voisins ‘the neighbours._FEM and neighbours._MASC’, contracted double forms: les voisin·es ‘the neighbours._MASC·FEM’, or gender-neutral forms: le voisinage ‘the neighbourhood’). In line with previous findings, the masculine form led to a male bias in the participants’ mental representations of gender. All three examined gender-fair forms resolved this bias, but comparisons of the different forms to each other revealed differences between them. The results show that complete double forms lead to equally balanced mental representations of gender while contracted double forms slightly favour representation of women. Finally, gender-neutral forms result in a small male bias, although significantly smaller than the one produced by the masculine form. The results are discussed in relation to the mental models theory and provide new and important insights on how gender-fair forms in French are interpreted. 

Language comprehension involves the derivation of the meaning of sentences by combining the meanings of their parts. In some cases, this can lead to ambiguity. A sentence like Every hiker climbed a hill allows two logical representations: One that specifies that every hiker climbed a different hill and one that specifies that every hiker climbed the same hill. The interpretations of such sentences can be primed: Exposure to a particular reading increases the likelihood that the same reading will be assigned to a subsequent similar sentence. Feiman and Snedeker (2016) observed that such priming is not modulated by overlap of the verb between prime and target. This indicates that mental logical representations specify the compositional structure of the sentence meaning without conceptual meaning content. We conducted a close replication of Feiman and Snedeker’s experiment in Dutch and found no verb-independent priming. Moreover, a comparison with a previous, within-verb priming experiment showed an interaction, suggesting stronger verb-specific than abstract priming. A power analysis revealed that both Feiman and Snedeker’s experiment and our Experiment 1 were underpowered. Therefore, we replicated our Experiment 1, using the sample size guidelines provided by our power analysis. This experiment again showed that priming was stronger if a prime-target pair contained the same verb. Together, our experiments show that logical representation priming is enhanced if the prime and target sentence contain the same verb. This suggests that logical representations specify compositional structure and meaning features in an integrated manner.

Number agreement attraction in comprehension is a phenomenon that has been documented in various typologically diverse languages. This evidence has led to claims about the cross-linguistic uniformity of agreement attraction effects and its independence from the grammatical features of a particular language. However, recent research has shown that in Czech, number agreement attraction effects are either absent or negligible in size. This directly contradicts the cross-linguistic uniformity hypothesis. The current paper aims to further corroborate this finding and presents a direct experimental comparison of Czech and English. Two comparable self-paced reading experiments were conducted using stimuli that were translation equivalents in Czech and English. Our analyses demonstrate a preference for the null model in Czech (no agreement attraction), unlike in English, where an interaction between verb number and attractor number was preferred. Moreover, when we compare the data from the two experiments directly, we find that the interaction between language and attraction was also preferred. In sum, we provide evidence against the cross-linguistic uniformity hypothesis for agreement attraction effects and suggest that processing patterns may differ between languages even for almost identical structures, such as agreement relations.

A language user must rely on several different abilities to carry out a conversation, e.g., the ability to acknowledge the conversational contributions of others, to respond appropriately, to stay on topic, etc. There are many aspects of the development of conversational conduct that are yet unknown. In this study, the longitudinal development of conversational conduct, as in acknowledging one's interlocutor's previous turn, was traced from age 5;0 to 7;2. We also investigated whether conversational conduct was predicted by core language skill, executive functions, and specific pragmatic abilities. Previous findings of productive morphosyntactic accuracy were replicated, while findings concerning longitudinal receptive vocabulary were not. We also found connections between children's conversational responses and executive functions, working memory, and the comprehension of conversational implicatures. The results suggest that conversational conduct is dependent on inferring communicative intentions, as well as being able to keep track of others' contributions and how they relate to previous turns.

Though each and every are both distributive universal quantifiers, a common theme in linguistic and psycholinguistic investigations into them has been that each is somehow more individualistic than every. We offer a novel explanation for this generalization: each has a first-order meaning which serves as an internalized instruction to cognition to build a thought that calls for representing the (restricted) domain as a series of individuals; by contrast, every has a second-order meaning which serves as an instruction to build a thought that calls for grouping the domain. In support of this view, we show that these distinct meanings invite the use of distinct verification strategies, using a novel paradigm. In two experiments, participants who had been asked to verify sentences like each/every circle is green were subsequently given a change detection task. Those who evaluated each-sentences were better able to detect the change, suggesting they encoded the individual circles' colors to a greater degree. Taken together with past work demonstrating that participants recall group properties after evaluating sentences with every better than after evaluating sentences with each, these results support the hypothesis that each and every call for treating the individuals that constitute their domain differently: as independent individuals (each) or as members of an ensemble collection (every). We situate our findings within a conception of linguistic meanings as instructions for thought building, on which the format of the resulting thought has consequences for how meanings interface with non-linguistic cognition.

Major models of sentence comprehension assume that a verb triggers retrieval of preceding thematic arguments from memory to establish argument-verb dependencies. If so, longer argument-verb distance should lead to higher processing load at the verb (a locality effect),since the representation of the argument should suffer from decay and/or interference.However, verb-final languages have often failed to show the expected argument-verb locality effect. A possible account of the lack of the effect is that arguments and adjuncts before the verb reactivate each other, counteracting memory degradation. In a pair of self-paced reading experiments in Japanese, a verb-final language, we found evidence of such pre-verb reactivation.Specifically, there was a locality effect and a similarity-based interference effect at the head of the adverbial that follows the subject, both of which suggest the retrieval of the subject at that point. The results are difficult to accommodate with other accounts of the lack of locality effect, such as a confounding effect of expectation and the inherent locality-insensitivity of verb-final languages. It is further argued that the constructivist analysis of verbal argument structure, which has been developed in generative syntax, provides an explanation for why such pre-verb reactivation takes place. 

Expectation-driven convergence occurs when speakers shift their speech to approximate the language they expect rather than observe from their interlocutor. In Wade (2022), participants produced more monophthongal /aI/—a salient feature of Southern U.S. English—after hearing other Southern-accented features. Here, by decoupling acoustic and social information with a dialect-label manipulation task, we investigate what types of cognitive associations account for this behavior: indirect socially-mediated associations that rely on recognizing that monophthongal /aI/ and other Southern-accented variants are both associated with the social category “Southern,” or direct associations between variants that rely on tracking their common co-occurrence at the individual level. We find that both acoustic and social-label cues trigger convergence, but in-group speakers from the South rely on acoustic cues, while out-group speakers from outside of the South are best cued by social-category labels. Results indicate a crucial role of dialect experience in the encoding and utilization of sociolinguistic knowledge.

Scientific studies of language span across many disciplines and provide evidence for social,  cultural, cognitive, technological, and biomedical studies of human nature and behavior. As it becomes increasingly empirical and quantitative, linguistics has been facing challenges and limitations of the scientific practices that pose barriers to reproducibility and replicability. One of the  proposed solutions to the widely acknowledged reproducibility and replicability crisis has been the implementation of transparency practices,  e.g., open access publishing, preregistrations, sharing study materials, data, and analyses, performing study replications, and declaring conflicts of interest. Here, we have assessed the prevalence of these practices in 600 randomly sampled journal articles from linguistics across two time points. In line with similar studies in other disciplines, we found that 35% of the articles were published open access and the rates of sharing materials, data, and protocols were below 10%. None of the articles reported preregistrations, 1% reported replications, and 10% had conflict of interest statements. These rates have not increased noticeably between 2008/2009 and 2018/2019, pointing to remaining barriers and the slow adoption of open and reproducible research practices in linguistics. To facilitate adoption of these practices, we provide a range of recommendations and solutions for implementing transparency and improving reproducibility of research in linguistics.