Editor's Introduction: 2017 Rumelhart Prize Issue Honoring Lila R. Gleitman

This is the Editor's introduction to the Special Issue of TopiCS in honor of Lila R. Gleitman's receipt of the 2017 David E. Rumelhart Prize. The introduction gives an overview of Gleitman's intellectual history and scientific contributions, and it briefly reviews each of the contributions to the issue.     

戊四氮点燃癫痫大鼠空间学习记忆障碍及海马突触素表达减少的研究

目的:探讨戊四氮点燃癫痫对大鼠空间学习记忆的影响及可能的分子机制。方法:戊四氮(pentylenetet-razol,PTZ)点燃建立慢性癫痫(chronic epileptic,CEP)模型,Morris水迷宫进行行为学检测,western blot方法观察大鼠海马突触素(synaptophysin,P38)蛋白的表达变化。结果:水迷宫试验检测癫痫组大鼠空间学习记忆能力受损(P<0.05,P<0.01);western blot结果表明海马P38蛋白表达较对照组明显减少(P<0.05)。结论:戊四氮点燃癫痫大鼠伴有学习记忆功能减退,其海马P38蛋白的表达减少可能参与了空间学习记忆受损。     

Editors’ Introduction: Aligning Implicit Learning and Statistical Learning: Two Approaches,One Phenomenon

This editors’ introduction provides the background to the special issue. We first outline the rationale for bringing together, in a single volume, leading researchers from two distinct, yet related research strands, implicit learning and statistical learning. The aim of the special issue is to facilitate the development of a shared understanding of research questions and methodologies, to provide a platform for discussing similarities and differences between the two strands, and to encourage the formulation of joint research agendas. We then introduce the new contributions solicited for this special issue and provide our perspective on the agenda setting that results from combining these two approaches.     

Noise Correlations for Faster and More Robust Learning

Distributed population codes are ubiquitous in the brain and pose a challenge to downstream neurons that must learn an appropriate readout. Here we explore the possibility that this learning problem is simplified through inductive biases implemented by stimulus-independent noise correlations that constrain learning to task-relevant dimensions. We test this idea in a set of neural networks that learn to perform a perceptual discrimination task. Correlations among similarly tuned units were manipulated independently of an overall population signal-to-noise ratio to test how the format of stored information affects learning. Higher noise correlations among similarly tuned units led to faster and more robust learning, favoring homogenous weights assigned to neurons within a functionally similar pool, and could emerge through Hebbian learning. When multiple discriminations were learned simultaneously, noise correlations across relevant feature dimensions sped learning, whereas those across irrelevant feature dimensions slowed it. Our results complement the existing theory on noise correlations by demonstrating that when such correlations are produced without significant degradation of the signal-to-noise ratio, they can improve the speed of readout learning by constraining it to appropriate dimensions.SIGNIFICANCE STATEMENT Positive noise correlations between similarly tuned neurons theoretically reduce the representational capacity of the brain, yet they are commonly observed, emerge dynamically in complex tasks, and persist even in well-trained animals. Here we show that such correlations, when embedded in a neural population with a fixed signal-to-noise ratio, can improve the speed and robustness with which an appropriate readout is learned. In a simple discrimination task such correlations can emerge naturally through Hebbian learning. In more complex tasks that require multiple discriminations, correlations between neurons that similarly encode the task-relevant feature improve learning by constraining it to the appropriate task dimension.     

Aligning Developmental and Processing Accounts of Implicit and Statistical Learning

A long‐standing question in child language research concerns how children achieve mature syntactic knowledge in the face of a complex linguistic environment. A widely accepted view is that this process involves extracting distributional regularities from the environment in a manner that is incidental and happens, for the most part, without the learner's awareness. In this way, the debate speaks to two associated but separate literatures in language acquisition: statistical learning and implicit learning. Both fields have explored this issue in some depth but, at present, neither the results from the infant studies used by the statistical learning literature nor the artificial grammar learning tasks studies from the implicit learning literature can be used to fully explain how children's syntax becomes adult‐like. In this work, we consider an alternative explanation—that children use error‐based learning to become mature syntax users. We discuss this proposal in the light of the behavioral findings from structural priming studies and the computational findings from Chang, Dell, and Bock's (2006) dual‐path model, which incorporates properties from both statistical and implicit learning, and offers an explanation for syntax learning and structural priming using a common error‐based learning mechanism. We then turn our attention to future directions for the field, here suggesting how structural priming might inform the statistical learning and implicit learning literature on the nature of the learning mechanism.     

More Than Words: The Role of Multiword Sequences in Language Learning and Use

The ability to convey our thoughts using an infinite number of linguistic expressions is one of the hallmarks of human language. Understanding the nature of the psychological mechanisms and representations that give rise to this unique productivity is a fundamental goal for the cognitive sciences. A long-standing hypothesis is that single words and rules form the basic building blocks of linguistic productivity, with multiword sequences being treated as units only in peripheral cases such as idioms. The new millennium, however, has seen a shift toward construing multiword linguistic units not as linguistic rarities, but as important building blocks for language acquisition and processing. This shift—which originated within theoretical approaches that emphasize language learning and use—has far-reaching implications for theories of language representation, processing, and acquisition. Incorporating multiword units as integral building blocks blurs the distinction between grammar and lexicon; calls for models of production and comprehension that can accommodate and give rise to the effect of multiword information on processing; and highlights the importance of such units to learning. In this special topic, we bring together cutting-edge work on multiword sequences in theoretical linguistics, first-language acquisition, psycholinguistics, computational modeling, and second-language learning to present a comprehensive overview of the prominence and importance of such units in language, their possible role in explaining differences between first- and second-language learning, and the challenges the combined findings pose for theories of language.     

A Single Paradigm for Implicit and Statistical Learning

Implicit learning generally refers to the acquisition of structures that, like knowledge of natural language grammar, are not available to awareness. In contrast, statistical learning has frequently been related to learning language structures that are explicitly available, such as vocabulary. In this paper, we report an experimental paradigm that enables testing of both classic implicit and statistical learning in language. The paradigm employs an artificial language comprising sentences that accompany visual scenes that they represent, thus combining artificial grammar learning with cross‐situational statistical learning of vocabulary. We show that this methodology enables a comparison between acquisition of grammar and vocabulary, and the influences on their learning. We show that both grammar and vocabulary are promoted by explicit information about the language structure, that awareness of structure affects acquisition during learning, and awareness precedes learning, but is not distinctive at the endpoint of learning. The two traditions of learning—implicit and statistical—can be conjoined in a single paradigm to explore both the phenomenological and learning consequences of statistical structural knowledge.     

Spatial and numerical processing in children with non-verbal learning disabilities

Consistently with the idea that numbers and space interact with each other, the present paper aimed to investigate the impact of non-verbal learning disabilities (NVLD) on spatial and numerical processing. In order to do so, 15 NVLD and 15 control children were required to perform different spatial (the line bisection and Simon tasks) and numerical tasks (the number bisection, number-to-position and numerical comparison tasks). In every task, NVLD children presented lower accuracy scores than the control group. While both groups manifested similar pseudo-neglect and Simon effects, they however differed in the numerical comparison task: while control children presented the standard SNARC effect in the uncrossed and crossed postures, no SNARC effect was observed in the NVLD group. Our results therefore suggest that NVLD affects the accuracy and the nature of the mental number line by decreasing its precision and the saliency of its left-to-right orientation.     

Implicit Learning of Sequential Regularities and Spatial Contexts in Corticobasal Syndrome

The present study investigated two forms of implicit learning in patients with corticobasal syndrome (CBS): contextual cueing and sequence learning. The former primarily implicates the medial temporal lobe system, and the latter, fronto-striatal-cerebellar circuits. Results revealed relatively preserved contextual cueing in patients with CBS. By contrast, sequence learning showed impairments, which seemed to reflect inability to execute correct responses in the presence of intact learning of the sequence. These findings provide the first characterization of implicit learning systems in CBS, and show that the two systems are differentially affected in patients with CBS.     

Spatial and temporal influences on extinction

This study investigated the spatial and temporal characteristics of the attentional deficit in patients exhibiting extinction to determine the extent to which these characteristics can be explained by a theory of an underlying gradient resulting from the differential contribution of interacting cell populations. The paradigm required the identification of two letters whose spatial location was varied both within and across hemifields. Additionally, the interval between the appearances of the two stimuli was manipulated by changing the stimulus onset asynchrony (SOA). A final variable, that of expectancy, was introduced by making the stimulus location more or less predictable and examining the effect of this top–down contingency on performance. The findings were consistent across two patients and indicated the joint contribution of both spatial and temporal factors: the contralesional stimulus was maximally extinguished when it was preceded by the ipsilesional stimulus by 300–900 ms, but this extinction was reduced when the stimuli appeared further ipsilesionally. Interestingly, there was increased extinction of the contralesional stimulus when location was predictable. These findings support the hypothesis that the attentional deficit in extinction patients arises from a contralesional-to-ipsilesional gradient of cell populations that interact in a mutually inhibitory manner.     

Editors' Review and Introduction: Learning Grammatical Structures: Developmental,Cross‐Species,and Computational Approaches

Human languages all have a grammar, that is, rules that determine how symbols in a language can be combined to create complex meaningful expressions. Despite decades of research, the evolutionary, developmental, cognitive, and computational bases of grammatical abilities are still not fully understood. “Artificial Grammar Learning” (AGL) studies provide important insights into how rules and structured sequences are learned, the relevance of these processes to language in humans, and whether the cognitive systems involved are shared with other animals. AGL tasks can be used to study how human adults, infants, animals, or machines learn artificial grammars of various sorts, consisting of rules defined typically over syllables, sounds, or visual items. In this introduction, we distill some lessons from the nine other papers in this special issue, which review the advances made from this growing body of literature. We provide a critical synthesis, identify the questions that remain open, and recognize the challenges that lie ahead. A key observation across the disciplines is that the limits of human, animal, and machine capabilities have yet to be found. Thus, this interdisciplinary area of research firmly rooted in the cognitive sciences has unearthed exciting new questions and venues for research, along the way fostering impactful collaborations between traditionally disconnected disciplines that are breaking scientific ground.     

Structured Sequence Learning: Animal Abilities,Cognitive Operations,and Language Evolution

Human language is a salient example of a neurocognitive system that is specialized to process complex dependencies between sensory events distributed in time, yet how this system evolved and specialized remains unclear. Artificial Grammar Learning (AGL) studies have generated a wealth of insights into how human adults and infants process different types of sequencing dependencies of varying complexity. The AGL paradigm has also been adopted to examine the sequence processing abilities of nonhuman animals. We critically evaluate this growing literature in species ranging from mammals (primates and rats) to birds (pigeons, songbirds, and parrots) considering also cross‐species comparisons. The findings are contrasted with seminal studies in human infants that motivated the work in nonhuman animals. This synopsis identifies advances in knowledge and where uncertainty remains regarding the various strategies that nonhuman animals can adopt for processing sequencing dependencies. The paucity of evidence in the few species studied to date and the need for follow‐up experiments indicate that we do not yet understand the limits of animal sequence processing capacities and thereby the evolutionary pattern. This vibrant, yet still budding, field of research carries substantial promise for advancing knowledge on animal abilities, cognitive substrates, and language evolution.     

Learning deficits in congenitally hydrocephalic rats and prevention by early shunt treatment

Shunt surgery is the usual treatment for infantile hydrocephalus; however, the extent to which it avoids subsequent neurological deficits is uncertain. The effect of early-onset hydrocephalus was tested in H-Tx rats using the Morris water maze. Spatial learning was assessed at 21 days after birth in control (n=18), hydrocephalic (n=18) and hydrocephalic rats shunt-treated at 4–5 (n=7) or at 10–12 days of life (n=13). The time taken to find a hidden platform was measured in five trials on 2 consecutive days and the data analyzed by one-and two-way ANOVA and t-tests. The latencies of the control rats decreased significantly between the first and second trial on the 1 st day, and learning was retained until the 2nd day. The hydrocephalic group had longer latencies than controls on both days, with no significant decrease between any trials. Performance was not significantly different between the two shunt groups. Overall, the shunted rats had latencies which were not significantly different from controls but were significantly lower than hydrocephalics. Despite this, the shunted rats did not perform as well as the controls. It is concluded that, although shunt treatment improved learning, some effects of early-onset hydrocephalus may not be reversible and/or a longer recovery time is required.     

Ibotenate Lesions of Hippocampus and/or Subiculum: Dissociating Components of Allocentric Spatial Learning

This study examined the effects of ibotenic acid-induced lesions of the hippocampus, subiculum and hippocampus +/- subiculum upon the capacity of rats to learn and perform a series of allocentric spatial learning tasks in an open-field water maze. The lesions were made by infusing small volumes of the neurotoxin at a total of 26 (hippocampus) or 20 (subiculum) sites intended to achieve complete target cell loss but minimal extratarget damage. The regional extent and axon-sparing nature of these lesions was evaluated using both cresyl violet and Fink - Heimer stained sections. The behavioural findings indicated that both the hippocampus and subiculum lesions caused impairment to the initial postoperative acquisition of place navigation but did not prevent eventual learning to levels of performance almost as effective as those of controls. However, overtraining of the hippocampus + subiculum lesioned rats did not result in significant place learning. Qualitative observations of the paths taken to find a hidden escape platform indicated that different strategies were deployed by hippocampal and subiculum lesioned groups. Subsequent training on a delayed matching to place task revealed a deficit in all lesioned groups across a range of sample choice intervals, but the subiculum lesioned group was less impaired than the group with the hippocampal lesion. Finally, unoperated control rats given both the initial training and overtraining were later given either a hippocampal lesion or sham surgery. The hippocampal lesioned rats were impaired during a subsequent retention/relearning phase. Together, these findings suggest that total hippocampal cell loss may cause a dual deficit: a slower rate of place learning and a separate navigational impairment. The prospect of unravelling dissociable components of allocentric spatial learning is discussed.     

Spatial learning and memory impairment and increased locomotion in a transgenic amyloid precursor protein mouse model of Alzheimer's disease

This study provides an examination of spatial learning and a behavioral assessment of irritability and locomotion in TgCRND8 mice, an amyloid precursor protein transgenic model of Alzheimer's disease. Performance was assessed using the Barnes maze, the touch escape test, and an open-field test. While past research focused primarily on 2-5-month-old TgCRND8 mice, the present study used an older age cohort (9-month-old female mice), in addition to a 4-month-old cohort of both transgenic (Tg) and wildtype female mice. Both younger and older Tg mice displayed poor spatial learning in the Barnes maze task compared to their wildtype littermates, as demonstrated by significantly longer latencies and more errors both during acquisition and at a 2-week retest. No differences in irritability were found between Tg and control mice in the younger cohort; however, older Tg mice displayed significantly higher irritability compared with wildtype littermates, as measured by the touch escape test. Additionally, Tg mice of both age cohorts showed increased locomotion and slowed habituation during a 60-min open-field test over 3 days of testing. These results demonstrate that TgCRND8 mice show significant deficits in spatial and nonspatial behavioral tasks at advanced stages of amyloid pathology.     

Impairments in Probabilistic Prediction and Bayesian Learning Can Explain Reduced Neural Semantic Priming in Schizophrenia

It has been proposed that abnormalities in probabilistic prediction and dynamic belief updating explain the multiple features of schizophrenia. Here, we used electroencephalography (EEG) to ask whether these abnormalities can account for the well-established reduction in semantic priming observed in schizophrenia under nonautomatic conditions. We isolated predictive contributions to the neural semantic priming effect by manipulating the prime’s predictive validity and minimizing retroactive semantic matching mechanisms. We additionally examined the link between prediction and learning using a Bayesian model that probed dynamic belief updating as participants adapted to the increase in predictive validity. We found that patients were less likely than healthy controls to use the prime to predictively facilitate semantic processing on the target, resulting in a reduced N400 effect. Moreover, the trial-by-trial output of our Bayesian computational model explained between-group differences in trial-by-trial N400 amplitudes as participants transitioned from conditions of lower to higher predictive validity. These findings suggest that, compared with healthy controls, people with schizophrenia are less able to mobilize predictive mechanisms to facilitate processing at the earliest stages of accessing the meanings of incoming words. This deficit may be linked to a failure to adapt to changes in the broader environment. This reciprocal relationship between impairments in probabilistic prediction and Bayesian learning/adaptation may drive a vicious cycle that maintains cognitive disturbances in schizophrenia.     

Canalization of Language Structure From Environmental Constraints: A Computational Model of Word Learning From Multiple Cues

There is substantial variation in language experience, yet there is surprising similarity in the language structure acquired. Constraints on language structure may be external modulators that result in this canalization of language structure, or else they may derive from the broader, communicative environment in which language is acquired. In this paper, the latter perspective is tested for its adequacy in explaining robustness of language learning to environmental variation. A computational model of word learning from cross-situational, multimodal information was constructed and tested. Key to the model's robustness was the presence of multiple, individually unreliable information sources to support learning. This “degeneracy” in the language system has a detrimental effect on learning, compared to a noise-free environment, but has a critically important effect on acquisition of a canalized system that is resistant to environmental noise in communication.     

Spatial bias in symbolic and non-symbolic numerical comparison in neglect

When asked to bisect mentally numerical intervals, neglect patients show a displacement of the numerical midpoint similar to the one observed in physical line bisection. This spatial-numerical bias has been taken as evidence of the spatial nature of numerical magnitude representations. However, to date, neuropsychological studies in neglect patients have only used symbolic numerical material. Here, we compare the results of patients with right-hemisphere damage with and without unilateral left neglect and age-matched healthy control participants in two numerical comparison tasks using symbolic and non-symbolic materials, in order to determine whether the representation of non-symbolic numerosities was altered or not by the presence of neglect. When asked to judge if an Arabic digit or a sequence of flashed dots was smaller or larger than a reference value (i.e., 5), the responses of neglect patients to smaller magnitudes (i.e., 4) were impaired. Moreover, only neglect patients presented an asymmetrical distance effect (i.e., an enhanced effect only for stimuli of smaller numerical magnitude than the reference). These results provide the first direct evidence of a spatial bias in non-symbolic numerosity in neglect patients, and support the existence of common processing mechanisms and/or a representational system for symbolic and non-symbolic inputs.     

Learning word meanings during reading by children with language learning disability and typically-developing peers

This study investigated whether children with language learning disability (LLD) differed from typically-developing peers in their ability to learn meanings of novel words presented during reading. Fifteen 9–11-year-old children with LLD and 15 typically-developing peers read four passages containing 20 nonsense words. Word learning was assessed through oral definition and multiple-choice tasks. Variables were position of informative context, number of exposures, part of speech, and contextual clues. The LLD group scored lower than same-aged peers on oral definition (p < .001) and multiple-choice (p < .001) tasks. For both groups, there was no effect for position of informative context (p = .867) or number of exposures (p = .223). All children benefitted from contextual clues. The findings suggested difficulty inferring and recalling word meanings during reading and pointed to the need for vocabulary intervention in the upper elementary years for children with LLD.     

How age of bilingual exposure can change the neural systems for language in the developing brain: A functional near infrared spectroscopy investigation of syntactic processing in monolingual and bilingual children

Is the developing bilingual brain fundamentally similar to the monolingual brain (e.g., neural resources supporting language and cognition)? Or, does early-life bilingual language experience change the brain? If so, how does age of first bilingual exposure impact neural activation for language?We compared how typically-developing bilingual and monolingual children (ages 7–10) and adults recruit brain areas during sentence processing using functional Near Infrared Spectroscopy (fNIRS) brain imaging. Bilingual participants included early-exposed (bilingual exposure from birth) and later-exposed individuals (bilingual exposure between ages 4–6).Both bilingual children and adults showed greater neural activation in left-hemisphere classic language areas, and additionally, right-hemisphere homologues (Right Superior Temporal Gyrus, Right Inferior Frontal Gyrus). However, important differences were observed between early-exposed and later-exposed bilinguals in their earliest-exposed language. Early bilingual exposure imparts fundamental changes to classic language areas instead of alterations to brain regions governing higher cognitive executive functions. However, age of first bilingual exposure does matter. Later-exposed bilinguals showed greater recruitment of the prefrontal cortex relative to early-exposed bilinguals and monolinguals.The findings provide fascinating insight into the neural resources that facilitate bilingual language use and are discussed in terms of how early-life language experiences can modify the neural systems underlying human language processing.