We recognize today's deep neural network (DNN) models of language behaviors as engineering achievements. However, what we know intuitively and scientifically about language shows that what DNNs are and how they are trained on bare texts, makes them poor models of mind and brain for language organization, as it interacts with infant biology, maturation, experience, unique principles, and natural law.

Jacques Mehler and I started Cognition, to break the grip of empiricist associationist behaviorism and stultifying style requirements on scientific discourse, and to liberate Cognitive Psychology's contributions to social issues. The journal provided opportunities for upcoming generations to expand Cognitive Psychology, publishing new concepts unhindered by established topics and standard formats. As the journal matured, Jacques kept. it fresh for 4 decades, seeking young scientists, novel ideas, and elegant writing, as it midwifed the emergence from Reductionist Behaviorism through Cognitive Psychology to rationalist Cognitive Science. The journal now has opportunities to nurture further progress in the future of Cognitive Science. I speculate that the field will keep associationist processes but integrate them with a new kind of non-reductionist theory that eschews detailed predictions, and which which interprets the brain as an enactor of thought, but not its structural cause. Ideally it will provide a set of constraints on the action of brain and mind that subsume and explain behavioral regularities, the role of frequency, how the brain externalizes those constraints and how the externalization processes emerge developmentally as a function of innate factors, structures unique to the mind and brain, experience and natural law.

Every language-learning child eventually automatically segments the organization of word sequences into natural units. Within the natural units, processing of normal conversation reveals a disconnect between listener’s representation of the sound and meaning of utterances. A compressed or absent word at a point early in a sequence is unintelligible until later acoustic information, yet listeners think they perceived the earlier sounds and their interpretation as they were heard. This discovery has several implications: Our conscious unified experience of language as we hear and simultaneously interpret it is partly reconstructed in time-suspended “psychological moments”; the “poverty of the stimulus language learning problem” is far graver than usually supposed; the serial domain where such integration occurs may be the “phase,” which unifies the serial percept with structural assignment and meanings; every level of language processing overlaps with others in a “computational fractal”; each level analysis-by-synthesis interaction of associative-serial and structure dependent processes.

Abstract The Interpretation of Chinese relative clauses has become of significant interest in sentence processing research, since some studies found that Chinese comprehension patterns violate the well-known universal subject relatives preference proposed by NAPH theory and proved by cross-linguistic studies. The current study uses the neuropsychological measure of ERP. It supports such a language-specific phenomenon, with the results showing that both subject-and object-modifying subject relatives eliciting larger ERP components than object relatives. The results suggest that object relatives are easier to process than subject relatives at both modifying positions, challenging the claim of universal subject preference for all languages. In addition, this study casts doubts on CWO and working memory-based DLT models despite results being compatible with them, and concludes that none of the current models are comprehensive enough to account for the data. Finally, this study offers a piece of evidence for the garden path effect caused by the surface NVN word order in the Chinese relative clause. All in all, this study adds to the evidence that processing preference is not universal. It contributes to a comprehensive model of how complex structures are processed.

Christiansen & Chater (C&C) ignore the many linguistic universals that cannot be reduced to processing or cognitive constraints, some of which we present. Their claim that grammar is merely acquired language processing skill cannot account for such universals. Their claim that all other universal properties are historically and culturally based is a nonsequitur about language evolution, lacking data.

aesthetics; depth perception; golden mean; language; vision