Contra assertions, feedback improves word recognition

Norris and Cutler (in press) revisit their arguments that (lexical-to-sublexical) feedback cannot improve word recognition performance, based on the assumption that feedback must boost signal and noise equally. They also argue that demonstrations that feedback improves performance (Magnuson, Mirman, Luthra, Strauss, & Harris, 2018) in the TRACE model of spoken word recognition (McClelland & Elman, 1986) were artifacts of converting activations to response probabilities. We first evaluate their claim that feedback in an interactive activation model must boost noise and signal equally. This is not true in a fully interactive activation model such as TRACE, where the feedback signal does not simply mirror the feedforward signal; it is instead shaped by joint probabilities over lexical patterns, and the dynamics of lateral inhibition. Thus, even under high levels of noise, lexical feedback will selectively boost signal more than noise. We demonstrate that feedback promotes faster word recognition and preserves accuracy under noise whether one uses raw activations or response probabilities. We then document that lexical feedback selectively boosts signal (i.e., lexically-coherent series of phonemes) more than noise by tracking sublexical (phoneme) activations under noise with and without feedback. Thus, feedback in a model like TRACE does improve word recognition, exactly by selective reinforcement of lexically-coherent signal. We conclude that whether lexical feedback is integral to human speech processing is an empirical question, and briefly review a growing body of work at behavioral and neural levels that is consistent with feedback and inconsistent with autonomous (non-feedback) architectures.

Does signal reduction imply predictive coding in models of spoken word recognition?

Pervasive behavioral and neural evidence for predictive processing has led to claims that language processing depends upon predictive coding. Formally, predictive coding is a computational mechanism where only deviations from top-down expectations are passed between levels of representation. In many cognitive neuroscience studies, a reduction of signal for expected inputs is taken as being diagnostic of predictive coding. In the present work, we show that despite not explicitly implementing prediction, the TRACE model of speech perception exhibits this putative hallmark of predictive coding, with reductions in total lexical activation, total lexical feedback, and total phoneme activation when the input conforms to expectations. These findings may indicate that interactive activation is functionally equivalent or approximant to predictive coding or that caution is warranted in interpreting neural signal reduction as diagnostic of predictive coding.

The processing of typological similar languages among trilingual speakers of Spanish, English and Portuguese

Pesquisas mostram (VAN HESTE, 1999; VAN HELL; DE GROOT, 1998; VAN HEUVEN, DIJSTRA; GRAINGER, 1998) que, quando falantes bilíngues visualizam palavras em diferentes línguas, diferentes níveis de semelhança lexical fazem com que outros vocábulos sejam simultaneamente ativados. Grande parte das pesquisas em processamento lexical se concentra na população bilíngue; enquanto que poucos estudos analisam tal fenômeno entre falantes trilíngues. O presente estudo se propõe a investigar como falantes trilíngues classificados como falantes nativos e de herança em espanhol processam palavras em diferentes idiomas (espanhol, português, alemão e basco) em comparação com falantes trilíngues não-nativos de espanhol. O objetivo é verificar se o terceiro idioma (português) afeta o processamento lexical em um idioma aprendido anteriormente (espanhol). Dois grupos participaram do estudo. O primeiro grupo era composto por falantes nativos de inglês (L1), com segunda língua espanhol (L2), e terceira língua português (L3). O segundo grupo era composto por falantes trilíngues nativos e/ou de herança em espanhol (n = 26) (L1-espanhol, L2-inglês e L3-português). Todos os participantes se encontravam em estágios iniciais de aprendizagem de língua portuguesa (L3). Resultados com base em tempos de reação e precisão demonstraram que falantes trilíngues nativos/de herança processaram palavras em espanhol mais rapidamente que falantes trilingues não nativos, além de processar todos os tipos de palavras (em português, espanhol, alemão e basco) com mais precisão. Os presentes resultados se basearam no modelo Bilingual Interactive Activation Plus (BIA +) de Dijkstra e Van Heuven (2002).

Metameric

An oft-cited shortcoming of Interactive Activation as a psychological model of word reading is that it lacks the ability to simultaneously represent words of different lengths. We present an implementation of the Interactive Activation model, which we call Metameric, that can simulate words of different lengths, and show that there is nothing inherent to Interactive Activation which prevents it from simultaneously representing multiple word lengths. We provide an in-depth analysis of which specific factors need to be present, and show that the inclusion of three specific adjustments, all of which have been published in various models before, lead to an Interactive Activation model which is fully capable of representing words of different lengths. Finally, we show that our implementation is fully capable of representing all words between 2 and 11 letters in length from the English Lexicon Project (31, 416 words) in a single model. Our implementation is completely open source, heavily optimized, and includes both command line and graphical user interfaces, but is also agnostic to specific input data or problems. It can therefore be used to simulate a myriad of other models, e.g., models of spoken word recognition. The implementation can be accessed at www.github.com/clips/metameric.

A few suggestions on broadening the cross-linguistic relevance of the Multilink model

In their keynote article, Dijkstra, Wahl, Buytenhuijs, van Halem, Al-jibouri, De Korte & Rekke (2018) propose a new model that aims to integrate and take care of the possible shortcomings of both the Revised Hierarchical Model (RHM, Kroll & Stewart, 1994) and Bilingual Interactive Activation model (BIA and several of its later versions, Dijkstra & Van Heuven, 1998). They begin their proposal by examining the objections/issues raised by Brysbaert and Duyck (2010) on RHM. It is well known that RHM is a developmental, production-based model which mostly predicted performance on translation-based tasks albeit with different predictions for translation recognition vs. production with regard to second language proficiency; BIA being a connectionist model of bilingual word recognition majorly emphasized on language non –selective selection and parallel language activation. Having been developed in and around Dutch (English as the second language), it took cognate status and orthographic similarity between words very seriously. Cognate status and orthographic similarity as factors won't count much as theoretical constructs around which a hypothesis could be developed if we look round the diverse types of orthographies and phonologies we find around the world.

The architecture of the mental lexicon and the selection of lexical nodes

This study presents the architecture of the mental lexicon of third language learners by focusing on three representation levels: letter, word and language. In particular, this analysis attempts to examine the extent of the influence of the first and second languages known by bilingual learners of English. The study is guided by Dijkstra’s (2003) Multilingual Interactive Activation (MIA) model, and the hypothesis of the language selective or language nonselective access of third language learners is tested. The method involved in this analysis is the word translation task as a tool for investigating the organization of the mental lexicon. The results obtained as a result of the translation task claim that trilingual speakers can operate with three languages during the process of learning.Keywords: multilingual processing, mental lexicon, language typology.