Unsupervised discovery of nonlinear structure using contrastive backpropagation

We describe a way of modeling high-dimensional data vectors by using an unsupervised, nonlinear, multilayer neural network in which the activity of each neuron-like unit makes an additive contribution to a global energy score that indicates how surprised the network is by the data vector. The connection weights that determine how the activity of each unit depends on the activities in earlier layers are learned by minimizing the energy assigned to data vectors that are actually observed and maximizing the energy assigned to "confabulations" that are generated by perturbing an observed data vector in a direction that decreases its energy under the current model.     

Deficient implicit phonological representations in children with dyslexia

This study tested the segmentation hypothesis of dyslexia by measuring implicit phonological representations in reading-disabled 11- to 13-year-olds. Implicit measures included lexical gating, priming, and syllable similarity tasks designed to reduce metalinguistic demands. Children with dyslexia performed consistently worse than CA and RA controls when more segmental representations were required across all three tasks. Implicit phonological representations were correlated with measures of speech perception, phoneme awareness, and phonological short-term memory, but not rapid automatized naming, and accounted for unique variance in predicting reading ability. Results provide strong support for less mature implicit phonological representations in children with dyslexia.     

Phoneme awareness is a key component of alphabetic literacy skills in consistent and inconsistent orthographies: evidence from Czech and English children

Two studies investigated the importance of phoneme awareness relative to other predictors in the development of reading and spelling among children learning a consistent orthography (Czech) and an inconsistent orthography (English). In Study 1, structural equation models revealed that Czech (n=107) and English (n=71) data were fitted well by the same predictors of reading and spelling. Phoneme awareness was a unique predictor in all models. In Study 2, Czech (n=40) and English (n=27) children with dyslexia showed similar deficits on phoneme awareness relative to their age- and spelling-matched control peers. Phoneme awareness appears to be a core component skill of alphabetic literacy, which is equally important for learners of consistent and inconsistent orthographies.     

Dissociations in performance on novel versus irregular items: single-route demonstrations with input gain in localist and distributed models

Four pairs of connectionist simulations are presented in which quasi-regular mappings are computed using localist and distributed representations. In each simulation, a control parameter termed input gain was modulated over the only level of representation that mapped inputs to outputs. Input gain caused both localist and distributed models to shift between regularity-based and item-based modes of processing. Performance on irregular items was selectively impaired in the regularity-based modes, whereas performance on novel items was selectively impaired in the item-based modes. Thus, the models exhibited double dissociations without separable processing components. These results are discussed in the context of analogous dissociations found in language domains such as word reading and inflectional morphology.     

Preschoolers prior formal mathematics education engage numerical magnitude representation rather than counting principles in symbolic ±1 arithmetic: Evidence from the operational momentum effect

In numerical cognition research, the operational momentum (OM) phenomenon (tendency to overestimate the results of addition and/or binding addition to the right side and underestimating subtraction and/or binding it to the left side) can help illuminate the most basic representations and processes of mental arithmetic and their development. This study is the first to demonstrate OM in symbolic arithmetic in preschoolers. It was modeled on Haman and Lipowska's (2021) non-symbolic arithmetic task, using Arabic numerals instead of visual sets. Seventy-seven children (4–7 years old) who know Arabic numerals and counting principles (CP), but without prior school math education, solved addition and subtraction problems presented as videos with one as the second operand. In principle, such problems may be difficult when involving a non-symbolic approximate number processing system, whereas in symbolic format they can be solved based solely on the successor/predecessor functions and knowledge of numerical orders, without reference to representation of numerical magnitudes. Nevertheless, participants made systematic errors, in particular, overestimating results of addition in line with the typical OM tendency. Moreover, subtraction and addition induced longer response times when primed with left- and right-directed movement, respectively, which corresponds to the reversed spatial form of OM. These results largely replicate those of non-symbolic task and show that children at early stages of mastering symbolic arithmetic may rely on numerical magnitude processing and spatial-numerical associations rather than newly-mastered CP and the concept of an exact number.     

The format of representation of recognized words in infants' early receptive lexicon

Eleven-month-olds can recognize a few auditorily presented familiar words in experimental situations where no hints are given by the intonation, the situation, or the presence of possible visual referents. That is, infants of this age (and possibly somewhat younger) can recognize words based on sound patterns alone. The issue addressed in this article is what is the type of mental representations infants use to code words they recognize. The results of a series of experiments with French-learning infants indicate that word representations in 11-month-olds are segmentally underspecified and suggest that they are all the more underspecified when infants engage in recognizing words rather than merely attending to meaningless speech sounds. But underspecification has limits, which were explored here with respect to word-initial consonants. The last two experiments show the way to investigating further these limits for word-initial consonants as well as for segments in other word positions. In French, infants' word representations are flexible enough to allow for structural changes in the voicing or even in the manner of articulation of word-initial consonants. Word-initial consonants must be present, however, for words to be recognized. In conclusion, a parallel is proposed between the emerging capacities to ignore variations that are irrelevant for word recognition in a “lexical mode” and to ignore variations that are phonemically irrelevant in a “neutral mode” of listening to native speech.     

汉字词的正字法深度与阅读时间的研究

采用命名任务探讨了正字法深度对汉字词读音时间的影响。结果发现,（１）正字法深度对汉字词的读音时间有重要影响。对于高频字,汉字的读音反应时的顺序（由快到慢）是规则形声字、象形字、会意字和不规则形声字,但规则形声字、象形字、会意字之间的差异没有达到显著水平,只有规则形声字与不规则形声字之间差异显著。但在低频时,规则形声字与会意字、不规则的形声字,象形字与不规则的形声字的读音反应时的差异均达到显著水平。（２）与单音词相比,多音词的反应时间长,多音词的优势反应与非优势反应的读音时间也存在着显著的差异。（３）高频时,多音双字词与单音双字词差异不显著,低频时双音双字词与单音双字词差异显著,从而显示了语境与词频的双重影响。整个研究表明,对于表意的汉字而言,正字法深度也是存在的,并对汉字词的读音时间产生重要的影响。     

语义知识神经表征的fMRI研究：通道特异性或类别特异性？

在同一实验中同时考虑类别与通道二个因素,并采用典型的语义提取任务应用功能磁共振成像技术探讨了语义知识的神经基础,结果发现不同类型的语义知识的脑激活模式（含左梭形回的激活）无论在范围还是强度上存在很大的相似性;而且,在严格的统计阈值下,不存在通道或是类别特异性激活的脑区。结果提示语义知识的神经表征分布于整个大脑皮层;在提取物体语义知识时都会同时激活该物体的视觉表象。另外,研究观察到的语义提取过程中BA9区的激活进一步证实并扩展了关于“汉语和表音文字在皮层组织上存在重要差异的观点”。     

Font effects of Chinese characters and pseudo-characters on the N400: evidence for an orthographic processing view

Event-related brain potentials (ERPs) were recorded during a Chinese character decision task to examine whether N400 amplitude is modulated by stimulus font. Results revealed large negative-going ERPs in an N400 time window of 300-500ms to stimuli presented in degraded Xing Kai Ti (XKT) font compared with more intact Song Ti (ST) font regardless of whether the stimuli were real or pseudo-characters. ERPs for the pseudo-characters were more negative than for the real characters with similar timing and scalp distribution. The N400-like font effect on amplitude is interpreted as analogous to an N400 stimulus degradation effect, an extension to Holcomb (1993); the degraded perceptual cues provided by XKT supposedly account for this degradation effect. This effect is further interpreted to reflect relative difficulty, which results from orthographic processing difficulty, in retrieving the meaning of XKT stimuli compared with ST stimuli.