Although heritage language phonology is often argued to be fairly stable, heritage language speakers often sound noticeably different from both monolinguals and second-language learners. In order to model these types of asymmetries, I propose a theoretical framework—an integrated multilingual sound system—based on modular representations of an integrated set of phonological contrasts. An examination of general findings in laryngeal (voicing, aspiration, etc.) phonetics and phonology for heritage languages shows that procedures for pronouncing phonemes are variable and plastic, even if abstract may representations remain stable. Furthermore, an integrated multilingual sound system predicts that use of one language may require a subset of the available representations, which illuminates the mechanisms that underlie phonological transfer, attrition, and acquisition.
Countries: United Kingdom, Italy, Netherlands, Italy, United Kingdom
Project: EC | ACT (289404), EC | ACT (289404)
The current study examined the effects of variability on infant event-related potential (ERP) data editing methods. A widespread approach for analyzing infant ERPs is through a trial-by-trial editing process. Researchers identify electroencephalogram (EEG) channels containing artifacts and reject trials that are judged to contain excessive noise. This process can be performed manually by experienced researchers, partially automated by specialized software, or completely automated using an artifact-detection algorithm. Here, we compared the editing process from four different editors-three human experts and an automated algorithm-on the final ERP from an existing infant EEG dataset. Findings reveal that agreement between editors was low, for both the numbers of included trials and of interpolated channels. Critically, variability resulted in differences in the final ERP morphology and in the statistical results of the target ERP that each editor obtained. We also analyzed sources of disagreement by estimating the EEG characteristics that each human editor considered for accepting an ERP trial. In sum, our study reveals significant variability in ERP data editing pipelines, which has important consequences for the final ERP results. These findings represent an important step toward developing best practices for ERP editing methods in infancy research.
AbstractPeter Godfrey-Smith’s Metazoa and Joseph LeDoux’s The Deep History of Ourselves present radically different big pictures regarding the nature, evolution and distribution of consciousness in animals. In this essay review, I discuss the motivations behind these big pictures and try to steer a course between them.
Abstract Research into representation learning models of lexical semantics usually utilizes some form of intrinsic evaluation to ensure that the learned representations reflect human semantic judgments. Lexical semantic similarity estimation is a widely used evaluation method, but efforts have typically focused on pairwise judgments of words in isolation, or are limited to specific contexts and lexical stimuli. There are limitations with these approaches that either do not provide any context for judgments, and thereby ignore ambiguity, or provide very specific sentential contexts that cannot then be used to generate a larger lexical resource. Furthermore, similarity between more than two items is not considered. We provide a full description and analysis of our recently proposed methodology for large-scale data set construction that produces a semantic classification of a large sample of verbs in the first phase, as well as multi-way similarity judgments made within the resultant semantic classes in the second phase. The methodology uses a spatial multi-arrangement approach proposed in the field of cognitive neuroscience for capturing multi-way similarity judgments of visual stimuli. We have adapted this method to handle polysemous linguistic stimuli and much larger samples than previous work. We specifically target verbs, but the method can equally be applied to other parts of speech. We perform cluster analysis on the data from the first phase and demonstrate how this might be useful in the construction of a comprehensive verb resource. We also analyze the semantic information captured by the second phase and discuss the potential of the spatially induced similarity judgments to better reflect human notions of word similarity. We demonstrate how the resultant data set can be used for fine-grained analyses and evaluation of representation learning models on the intrinsic tasks of semantic clustering and semantic similarity. In particular, we find that stronger static word embedding methods still outperform lexical representations emerging from more recent pre-training methods, both on word-level similarity and clustering. Moreover, thanks to the data set’s vast coverage, we are able to compare the benefits of specializing vector representations for a particular type of external knowledge by evaluating FrameNet- and VerbNet-retrofitted models on specific semantic domains such as “Heat” or “Motion.”
Axel Constant; Alexander Daniel Dunsmoir Tschantz; Alexander Daniel Dunsmoir Tschantz; Beren Millidge; Beren Millidge; Felipe Criado-Boado; Luis M Martinez; Johannes Müeller; Andy Clark; Andy Clark; +1 more
Axel Constant; Alexander Daniel Dunsmoir Tschantz; Alexander Daniel Dunsmoir Tschantz; Beren Millidge; Beren Millidge; Felipe Criado-Boado; Luis M Martinez; Johannes Müeller; Andy Clark; Andy Clark; Andy Clark;
This paper presents an active inference based simulation study of visual foraging. The goal of the simulation is to show the effect of the acquisition of culturally patterned attention styles on cognitive task performance, under active inference. We show how cultural artefacts like antique vase decorations drive cognitive functions such as perception, action and learning, as well as task performance in a simple visual discrimination task. We thus describe a new active inference based research pipeline that future work may employ to inquire on deep guiding principles determining the manner in which material culture drives human thought, by building and rebuilding our patterns of attention. Researchers on this article were supported by an Australian Laureate Fellowship project A Philosophy of Medicine for the 21st Century (Ref: FL170100160) and by a Social Sciences and Humanities Research Council doctoral fellowship (Ref: 752-2019-0065) (AC), by a PhD studentship from the Sackler Foundation and the School of Engineering and Informatics at the University of Sussex (AT); by an EPSRC PhD Studentship (BM), by a GAIN-Xunta de Galiza Groups of Excellence 2020 (FC-B), and by Horizon 2020 European Union ERC Advanced Grant XSPECT - DLV-692739 (AC). AT is grateful to the Mortimer and Theresa Sackler Foundation, which supports the Sackler Centre for Consciousness Science.
Publisher: Multidisciplinary Digital Publishing Institute
Project: EC | CoSaQ (716230)
While the languages of the world vary greatly, they exhibit systematic patterns, as well. Semantic universals are restrictions on the variation in meaning exhibit cross-linguistically (e.g., that, in all languages, expressions of a certain type can only denote meanings with a certain special property). This paper pursues an efficient communication analysis to explain the presence of semantic universals in a domain of function words: quantifiers. Two experiments measure how well languages do in optimally trading off between competing pressures of simplicity and informativeness. First, we show that artificial languages which more closely resemble natural languages are more optimal. Then, we introduce information-theoretic measures of degrees of semantic universals and show that these are not correlated with optimality in a random sample of artificial languages. These results suggest both that efficient communication shapes semantic typology in both content and function word domains, as well as that semantic universals may not stand in need of independent explanation.
AbstractRadical empiricists at the turn of the twentieth century described organisms as experiencing the relations they maintain with their surroundings prior to any analytic separation from their environment. They notably avoided separating perception of the material environment from social life. This perspective on perceptual experience was to prove the inspiration for Gibson’s ecological approach to perceptual psychology. Gibson provided a theory of how the direct perception of the organism-environment relation is possible. Central to his account was the notion of a medium for direct perception. However Gibson provided two mutually inconsistent accounts of the medium leading to problems for his radical empiricism. We develop an account of the medium that does justice to ecological psychology’s radical empiricist roots. To complement this account of the medium we detail a usage-based account of information. Together they allow us to propose a novel radical empiricist view of direct perception. We then return to the notion of medium and expand it to include sociomaterial practices. We show how direct perception happens in the midst of social life, and is made possible by an active achieving and maintaining of a pragmatic relation with the environment.
The study of dreams represents a crucial intersection between philosophical, psychological, neuroscientific, and clinical interests. Importantly, one of the main sources of insight into dreaming activity are the (oral or written) reports provided by dreamers upon awakening from their sleep. Classically, two main types of information are commonly extracted from dream reports: structural and semantic, content-related information. Extracted structural information is typically limited to the simple count of words or sentences in a report. Instead, content analysis usually relies on quantitative scores assigned by two or more (blind) human operators through the use of predefined coding systems. Within this review, we will show that methods borrowed from the field of linguistic analysis, such as graph analysis, dictionary-based content analysis, and distributional semantics approaches, could be used to complement and, in many cases, replace classical measures and scales for the quantitative structural and semantic assessment of dream reports. Importantly, these methods allow the direct (operator-independent) extraction of quantitative information from language data, hence enabling a fully objective and reproducible analysis of conscious experiences occurring during human sleep. Most importantly, these approaches can be partially or fully automatized and may thus be easily applied to the analysis of large datasets.
This study investigates the form of deictic gestures used by speakers of the Quiahije variety of Eastern Chatino (Otomangean, Zapotecan) spoken in Oaxaca, Mexico. An analysis of over six hours of interviews about local landmarks reveals that Quiahije Chatino speakers consistently use the far-is-up strategy to convey target distance in their deictic gestures—the farther the target, the higher and more expansive the form of the gesture. Participants in the study consistently used the far-is-up strategy to modify two types of deictic gestures: points and ‘go’ emblems (a gesture conveying forward motion). For points alone, participants combined the far-is-up strategy with the use of distinct handshapes for pointing to nearby versus distant targets. By systematically examining how deictic gestures are modified in one community in Mexico, this study lays the groundwork for further comparative and typological research on gestural deixis.
According to the psychological literature, implicit motives allow for the characterization of behavior, subsequent success, and long-term development. Contrary to personality traits, implicit motives are often deemed to be rather stable personality characteristics. Normally, implicit motives are obtained by Operant Motives, unconscious intrinsic desires measured by the Operant Motive Test (OMT). The OMT test requires participants to write freely descriptions associated with a set of provided images and questions. In this work, we explore different recent machine learning techniques and various text representation techniques for facing the problem of the OMT classification task. We focused on advanced language representations (e.g, BERT, XLM, and DistilBERT) and deep Supervised Autoencoders for solving the OMT task. We performed an exhaustive analysis and compared their performance against fully connected neural networks and traditional support vector classifiers. Our comparative study highlights the importance of BERT which outperforms the traditional machine learning techniques by a relative improvement of 7.9%. In addition, we performed an analysis of how the BERT attention mechanism is being modified. Our findings indicate that the writing style features acquire higher importance at the moment of accurately identifying the different OMT categories. This is the first time that a study to determine the performance of different transformer-based architectures in the OMT task is performed. Similarly, our work propose, for the first time, using deep supervised autoencoders in the OMT classification task. Our experiments demonstrate that transformer-based methods exhibit the best empirical results, obtaining a relative improvement of 7.9% over the competitive baseline suggested as part of the GermEval 2020 challenge. Additionally, we show that features associated with the writing style are more important than content-based words. Some of these findings show strong connections to previously reported behavioral research on the implicit psychometrics theory.