Search

Caretakers tend to repeat themselves when speaking to children, either to clarify their message or to redirect wandering attention. This repetition also appears to support language learning. For example, words that are heard more frequently tend to be produced earlier by young children. However, pure repetition only goes so far; some variation between utterances is necessary to support acquisition of a fully productive grammar. When individual words or morphemes are repeated, but embedded in different lexical and syntactic contexts, the child has more information about how these forms may be used and combined. Corpus analysis has shown that these partial repetitions frequently occur in clusters, which have been coined variation sets. More recent research has introduced algorithms that can extract these variation sets automatically from corpora with the goal of measuring their relative prevalence across ages and languages. Longitudinal analyses have revealed that rates of variation sets tend to decrease as children get older. We extend this research in several ways. First, we consider a maximally diverse sample of languages, both genealogically and geographically, to test the generalizability of developmental trends. Second, we compare multiple levels of repetition, both words and morphemes, to account for typological differences in how information is encoded. Third, we consider several additional measures of development to account for deficiencies in age as a measure of linguistic aptitude. Fourth, we examine whether the levels of repetition found in child-surrounding speech is greater or less than what would have been expected by chance. This analysis produced a new measure, redundancy, which captures how repetitive speech is on average given how repeititive it could have been. Fifth, we compare rates of repetition in child-surrounding and adult-directed speech to test whether variation sets are especially prevalent in child-surrounding speech. We find that (1) some languages show increases in repetition over development, (2) true estimates of variation sets are generally lower than or equal to random baselines, (3) these patterns are largely convergent across developmental indices, and (4) adult-directed speech is reliably less redundant, though in some cases more repetitive, than child-surrounding speech. These results are discussed with respect to features of the corpora, typological properties of the languages, and differential rates of change in repetition and redundancy over children's development., European Union (EU); European Research Council (ERC), European Union's Seventh Framework Programme (FP7/2007-2013); Swiss National Science Foundation

Eye movements are known to reflect cognitive processes in reading, and psychological reading research has shown that eye gaze patterns differ between readers with and without dyslexia. In recent years, researchers have attempted to classify readers with dyslexia based on their eye movements using Support Vector Machines (SVMs). However, these approaches (i) are based on highly aggregated features averaged over all words read by a participant, thus disregarding the sequential nature of the eye movements, and (ii) do not consider the linguistic stimulus and its interaction with the reader's eye movements. In the present work, we propose two simple sequence models that process eye movements on the entire stimulus without the need of aggregating features across the sentence. Additionally, we incorporate the linguistic stimulus into the model in two ways -- contextualized word embeddings and manually extracted linguistic features. The models are evaluated on a Mandarin Chinese dataset containing eye movements from children with and without dyslexia. Our results show that (i) even for a logographic script such as Chinese, sequence models are able to classify dyslexia on eye gaze sequences, reaching state-of-the-art performance, and (ii) incorporating the linguistic stimulus does not help to improve classification performance.

It has been shown that multilingual transformer models are able to predict human reading behavior when fine-tuned on small amounts of eye tracking data. As the cumulated prediction results do not provide insights into the linguistic cues that the model acquires to predict reading behavior, we conduct a deeper analysis of the predictions from the perspective of readability. We try to disentangle the three-fold relationship between human eye movements, the capability of language models to predict these eye movement patterns, and sentence-level readability measures for English. We compare a range of model configurations to multiple baselines. We show that the models exhibit difficulties with function words and that pre-training only provides limited advantages for linguistic generalization.

We present a new machine learning benchmark for reading task classification with the goal of advancing EEG and eye-tracking research at the intersection between computational language processing and cognitive neuroscience. The benchmark task consists of a cross-subject classification to distinguish between two reading paradigms: normal reading and task-specific reading. The data for the benchmark is based on the Zurich Cognitive Language Processing Corpus (ZuCo 2.0), which provides simultaneous eye-tracking and EEG signals from natural reading of English sentences. The training dataset is publicly available, and we present a newly recorded hidden testset. We provide multiple solid baseline methods for this task and discuss future improvements. We release our code and provide an easy-to-use interface to evaluate new approaches with an accompanying public leaderboard: www.zuco-benchmark.com., Frontiers in Psychology, 13, ISSN:1664-1078

As efficient systems of communication, languages are usually expected to map meanings to forms in a one-to-one way, using for example the same affix form (e.g., -s in English) every time a particular meaning is intended (e.g., plural number), and placing affixes with the same meaning consistently in the same position (e.g., always suffixal). Forms and positional rules extending over contexts with a common meaning (e.g., plural in 1PL, 2PL, 3PL) are thus considered natural, and those extending over contexts with no consistent common meaning (e.g., 1PL and 3SG) are considered unnatural. Natural patterns are most common cross-linguistically, and most learnable in experiments; however, little is yet know about differences between unnatural classes. In this study we explore syncretism (i.e., use of the same form in different functions) and affix position in the domain of person and number agreement in verbs, both cross-linguistically and in artificial language learning experiments. Results from the two approaches and both phenomena converge in finding a gradient of (un)naturalness. Rather than a dichotomous natural/unnatural distinction, we found that both cross-linguistic frequency and learnability are proportional to the amount of shared feature values among the contexts requiring the same form or position. We argue that a cognitive bias towards similarity-based structure explains our experimental results and could be driving the patterns observed in natural languages.

International audience; A critical feature of language is that the form of words need not bear any perceptual similarity to their functionthese relationships can be 'arbitrary'. The capacity to process these arbitrary form-function associations facilitates the enormous expressive power of language. However, the evolutionary roots of our capacity for arbitrariness, i.e. the extent to which related abilities may be shared with animals, is largely unexamined. We argue this is due to the challenges of applying such an intrinsically linguistic concept to animal communication, and address this by proposing a novel conceptual framework highlighting a key underpinning of linguistic arbitrariness, which is nevertheless applicable to non-human species. Specifically, we focus on the capacity to associate alternative functions with a signal, or alternative signals with a function, a feature we refer to as optionality. We apply this framework to a broad survey of findings from animal communication studies and identify five key dimensions of communicative optionality: signal production, signal adjustment, signal usage, signal combinatoriality and signal perception. We find that optionality is widespread in non-human animals across each of these dimensions, although only humans demonstrate it in all five. Finally, we discuss the relevance of optionality to behavioural and cognitive domains outside of communication. This investigation provides a powerful new conceptual framework for the cross-species investigation of the origins of arbitrariness, and promises to generate original insights into animal communication and language evolution more generally.

With the prevalence of GPS tracking technologiees, car insurance companies have started to adopt usage-based insurance policies, which adapt insurance premiums according to the customers' driving behavior. Although many risk models for assessing an individual driver's accident risk based on the history of driving trajectories, driving events, and exposure records exist, these models do not take the geographical context of the driven trajectories and driving events into account. This study explores the influence of enriching the existing purely driving-behavior-based feature set by multiple geographical context features for the task of differentiating between accident and accident-free drivers. Prediction performances of five machine learning classifiers—logistic regression, random forest, XGBoost, feed-forward neural networks (FFNN), and long short-term memory (LSTM) networks—were evaluated on the usage records of over 8000 vehicles in one year from Italy. The results show that the inclusion of geographical information such as weather, points of interest (POIs), and land use can increase the relative predictive performance in terms of AUC by up to 8%, among which land use is the most informative. For the data of this study, XGBoost generally yielded the best performance and made most use out of the geographical information, while logistic regression is only slightly outperformed by more complex models if the proposed geographical information is not available. LSTM did not outperform the other methods, possibly due to the small volume of training data available. The results outline the potential of including the geographical context in usage-based car insurance risk modeling to improve the accuracy, leading to fairer usage-based insurance policies.

In Albanian, the application of a preverbal accusative index, also known as clitic doubling, may appear not only with nominal but also with clausal arguments, a hitherto little-discussed phenomenon. The occurrence of the preverbal object index is at least partly conditional on readiness for individuation of the complement clause and on its information status. Individuation and information status are also involved in the indexing of accusative nominal arguments, which testifies to a system in Albanian that operates across structurally different complement types, treating them alike. The analysis presented in this chapter contributes to more general discussions concerning the syntactic properties of complements.

This paper explores the potential of legal documents for the study of the sociology of Old English. It gives a rationale for the use of legal genres, or charters, and introduces research databases and tools that may elucidate the interconnections between practitioners of legal Old English and their linguistic practices. A series of short case studies on wills illustrates what legal genres tell us about the correlation between linguistic variation, supralocalisation, and change and such variables as archive and gender.

The present paper discusses the concept of gradience and fuzziness from the perspective of Diachronic Construction Grammar. It does so by investigating verb-attached PPs in the history of English, with a focus on their semantic and syntactic functions and features over time. Specifically, the paper uses the Penn-Helsinki Parsed Corpora of Historical English, including texts from Middle English (PPCME2), Early Modern English (PPCEME), and Late Modern English (PPCMBE) to revisit the distinction between adjuncts and complements. In particular, I address the question whether this traditionally binary classification finds support in diachronic data, or whether PPs rather represent a gradience between prototypical adjunct- and complementhood. Furthermore, the paper assesses whether any change in the distribution and features of PPs (specifically an increase in complementhood) can be observed over time. Ultimately, the findings suggest a multi-level network of PPs that is diachronically very stable.

In this study, we examined whether the convergence in interlocutors’ vowel acoustics leads to decreasing discriminability between interlocutors’ voices. Ten pairs of Grison and Zürich German speakers produced lexical items before and after dialogue interactions with evidence of vowel convergence in post-dialogue productions. In Experiment 1, native and non-native Swiss German listeners discriminated pairs of speakers whose speech was obtained pre- and post-dialogue. Results showed that listeners’ sensitivity (A’) was higher for native than non-native listeners, but comparable for pre- and post-dialogue recordings. The observed negative correlation between voice discrimination and the acoustic distance in formant space was mainly driven by a single speaker pair. In Experiment 2, the speaker recognition performance of an i-vector-based software was compared in pre- and post-dialogue speech. Results revealed no difference in the system performance between the two conditions. The findings suggest that vowel convergence does not compromise voice discriminability under the given experimental conditions.

Zusammenfassung The goddess Orthia, whose name is attested by different variants in inscriptions mainly at her sanctuary at Sparta, and who was at some point identified with Artemis, is the subject of an ongoing debate in various fields of ancient studies. As the textual mythology of the goddess is meagre, the etymology of this theonym is of primary importance in bringing to light possible mythological concepts associated with the deity. Drawing on earlier attempts proposed in the literature, the Greek adjective ὀρθός (ved. ūrdhvá- ‚upright‘) is identified as the natural derivational basis for the name, which can in turn be traced back to the PIE root underlying ved. vrādh- and av. uruuad-. In Vedic, ūrdhvá- is used in describing the epiphany of Uṣas. Orthia can therefore be considered to be a descendant of the PIE dawn goddess. Via its etymology, the name of Orthia is related to the Avestan theonym Arəduuī (and probably to Celtic Ardvinna as well). The abundantly documented mythological profile of the Iranian goddess matches up well with the proposed origin of the name, thereby cross-validating the linguistic and mythological origin of Orthia.

The present study provides a survey of the semantics of depictive (in a broad sense, including circumstantials) adjectival compounds in Vedic Sanskrit. Following the typology of depictive constructions developed by Himmelmann & Schultze-Berndt (2005), we structure our classification along the semantic fields that such expressions tend to occur in. Our results show that in Vedic, the use of depictive adjectival compounds spans (almost) the whole gamut of functions reported for depictives in cross-linguistic studies. In Vedic, depictive compounds rank on a par with other strategies of non-finite event elaboration such as participles, verbal adjectives, and action nouns.

Redundant marking of grammatical relations seems to be commonplace across languages, and has been shown to benefit learning as well as robust information transmission. At the same time, languages also exhibit trade-offs between strategies such as case marking or word order, suggesting that redundancy may also be dis-preferred in line with a tendency towards communicative efficiency. In the present paper, we assess redundancy in terms of number of strategies used simultaneously to mark specific relations within individual utterances (syntagmatic redundancy) in light of these competing motivations. Our test case is participant role disambiguation in English and Dutch, specifically the interaction of constituent order, case, prepositional marking, and agreement to distinguish agents and recipients in ditransitive clauses. Using evidence from corpora of Present Day Dutch and English as well as data from Middle English, we find that redundancy is prevalent, albeit within certain limits.&nbsp

The alignment between low-frequency activity in the brain and slow acoustic modulations in the speech signal depicts a core principle in present theories of speech perceptionda process referred to as 'neural speech tracking'. While most older adults, particularly those with highly prevalent age-related hearing loss, have difficulties with speech perception and comprehension, the impact of hearing loss on neural speech tracking is still unclear. In this study we investigated the effects of pure-tone hearing loss and different types of background noise on the neural tracking response in a large sample of older adults (N = 101). Furthermore, we examined whether the neural tracking response was predictive for speech comprehension. For this purpose, we obtained scalp EEG from our participants who had varying degrees of pure-tone hearing loss (7.5-59.6 dB HL for .5-8 kHz pure tones) while they listened to sentences in quiet, pink and multi-talker babble noise. Speech tracking was quantified by computing the cross-correlation between the EEG signal and the amplitude envelope of the sentences heard. A higher degree of pure-tone hearing loss was associated with greater neural speech tracking (i.e., greater cross-correlation). Additionally, neural speech tracking showed a positive association with speech comprehension. This relationship was modulated by the degree of pure-tone hearing loss with hearing-impaired participants benefitting more from greater neural speech tracking. Our results highlight the potential of neural speech tracking as an objective measure of speech comprehension and as a possible target mechanism for clinical interventions such as neurofeedback. Furthermore, the interaction between speech tracking and pure-tone hearing loss suggests a compensatory mechanism by which the hearing-impaired rely more on slow amplitude modulations in the speech signal. (C) 2022 The Author(s). Published by Elsevier Ltd., Cortex, 151, ISSN:0010-9452, ISSN:1973-8102