Connectionist Models of Word Reading

Abstract— Connectionist models of word reading attempt to explain the computational mechanisms underlying this important skill. The goal of this research is an integrated theory of reading and its brain bases, with the computational model as the interface between the two. The models are governed by computational principles that differ considerably from naive intuitions but nonetheless account for many aspects of normal and impaired (dyslexic) reading.     

Marker-passing over Microfeatures: Towards a Hybrid Symbolic/Connectionist Model

Spreading activation, in the form of computer models and cognitive theories, has recently been undergoing a resurgence of interest in the cognitive science and AI communities. Two different types of cognitive models have been proposed to explain the activation spreading results. One approach, that of marker-passing, concentrates on the spreading of symbolic information through an associative knowledge representation. The other technique, including the work in local connectionism, has focused on the passage of numeric information through a network. In this article, it is shown that these two techniques can be merged. The implementation of a mechanism in which a local-connectionist-like model is integrated with a symbolic marker-passer is described and shows that the combined system is more powerful than either of the separate models alone. Finally, some early steps toward a hybrid model in which a distributed network is used to learn the microfeatures is described.     

Signal Detection Models with Random Participant and Item Effects

The theory of signal detection is convenient for measuring mnemonic ability in recognition memory paradigms. In these paradigms, randomly selected participants are asked to study randomly selected items. In practice, researchers aggregate data across items or participants or both. The signal detection model is nonlinear; consequently, analysis with aggregated data is not consistent. In fact, mnemonic ability is underestimated, even in the large-sample limit. We present two hierarchical Bayesian models that simultaneously account for participant and item variability. We show how these models provide for accurate estimation of participants’ mnemonic ability as well as the memorability of items. The model is benchmarked with a simulation study and applied to a novel data set. This research is supported by NSF grants SES-0095919 and SES-0351523, NIH grant R01-MH071418, a University of Missouri Research Leave grant and fellowships from the Spanish Ministry of Education and the University of Leuven, Belgium.     

Connectionist models of conditioning: A tutorial

Models containing networks of neuron-like units have become increasingly prominent in the study of both cognitive psychology and artificial intelligence. This article describes the basic features of connectionist models and provides an illustrative application to compound-stimulus effects in respondent conditioning. Connectionist models designed specifically for operant conditioning are not yet widely available, but some current learning algorithms for machine learning indicate that such models are feasible. Conversely, designers for machine learning appear to have recognized the value of behavioral principles in producing adaptive behavior in their creations.     

No strong evidence for lateralisation of word reading and face recognition deficits following posterior brain injury

Face recognition and word reading are thought to be mediated by relatively independent cognitive systems lateralised to the right and left hemispheres, respectively. In this case, we should expect a higher incidence of face recognition problems in patients with right hemisphere injury and a higher incidence of reading problems in patients with left hemisphere injury. We tested this hypothesis in a group of 31 patients with unilateral right or left hemisphere infarcts in the territory of the posterior cerebral arteries. In most domains tested (e.g., visual attention, object recognition, visuo-construction, motion perception), we found that both patient groups performed significantly worse than a matched control group. In particular, we found a significant number of face recognition deficits in patients with left hemisphere injury and a significant number of patients with word reading deficits following right hemisphere injury. This suggests that face recognition and word reading may be mediated by more bilaterally distributed neural systems than is commonly assumed.     

Raven's matrices and working memory: A dual-task approach

The cognitive system does not just act as a mirror from the sensory input; instead, it tends to normalize this information. Given that letter processing seems to be much more specialized than digit processing in the cortex, we examined whether the regularization process occurs differently from digits to letters than from letters to digits: We employed a masked priming same/different experiment (e.g., probe, VESZED; prime, V35Z3D ; and target, VESZED). When embedded in letter strings, digits that resemble letters (e.g., 3 and 5 in V35Z3D -VESZED) tend to be encoded in a letter-like manner, whereas when embedded in digit strings, letters that resemble digits (e.g., E and S in 9ES7E2–935732) tend not to be encoded in a digit-like manner.     

Implicit and explicit processing in deep dyslexia: Semantic blocking as a test for failure of inhibition in the phonological output lexicon

The failure of inhibition hypothesis posits a theoretical distinction between implicit and explicit access in deep dyslexia. Specifically, the effects of failure of inhibition are assumed only in conditions that have an explicit selection requirement in the context of production (i.e., aloud reading). In contrast, the failure of inhibition hypothesis proposes that implicit processing and explicit access to semantic information without production demands are intact in deep dyslexia. Evidence for intact implicit and explicit access requires that performance in deep dyslexia parallels that observed in neurologically intact participants on tasks based on implicit and explicit processes. In other words, deep dyslexics should produce normal effects in conditions with implicit task demands (i.e., lexical decision) and on tasks based on explicit access without production (i.e., forced choice semantic decisions) because failure of inhibition does not impact the availability of lexical information, only explicit retrieval in the context of production. This research examined the distinction between implicit and explicit processes in deep dyslexia using semantic blocking in lexical decision and forced choice semantic decisions as a test for the failure of inhibition hypothesis. The results of the semantic blocking paradigm support the distinction between implicit and explicit processing and provide evidence for failure of inhibition as an explanation for semantic errors in deep dyslexia.     

词义通达的三种理论模型及研究简介

本文介绍了视觉词汇识别领域中 ,词义通达的三种理论模型及其各自最新的一些实验证据 ,并提出今后研究中需要注意的一些问题     

Models for a paraconsistent set theory

In this paper the existence of natural models for a paraconsistent version of naive set theory is discussed. These stand apart from the previous attempts due to the presence of some non-monotonic ingredients in the comprehension scheme they fulfill. Particularly, it is proved here that allowing the equality relation in formulae defining sets, within an extensional universe, compels the use of non-monotonic operators. By reviewing the preceding attempts, we show how our models can naturally be obtained as fixed points of some functor acting on a suitable category (stressing the use of fixed-point arguments in obtaining such alternative semantics).     

SUBTLEX-UK: A new and improved word frequency database for British English

We present word frequencies based on subtitles of British television programmes. We show that the SUBTLEX-UK word frequencies explain more of the variance in the lexical decision times of the British Lexicon Project than the word frequencies based on the British National Corpus and the SUBTLEX-US frequencies. In addition to the word form frequencies, we also present measures of contextual diversity part-of-speech specific word frequencies, word frequencies in children programmes, and word bigram frequencies, giving researchers of British English access to the full range of norms recently made available for other languages. Finally, we introduce a new measure of word frequency, the Zipf scale, which we hope will stop the current misunderstandings of the word frequency effect.     

Distal rhythm influences whether or not listeners hear a word in continuous speech: Support for a perceptual grouping hypothesis

Due to extensive variability in the phonetic realizations of words, there may be few or no proximal spectro-temporal cues that identify a word’s onset or even its presence. Dilley and Pitt (2010) showed that the rate of context speech, distal from a to-be-recognized word, can have a sizeable effect on whether or not a word is perceived. This investigation considered whether there is a distinct role for distal rhythm in the disappearing word effect. Listeners heard sentences that had a grammatical interpretation with or without a critical function word (FW) and transcribed what they heard (e.g., are in Jill got quite mad when she heard there are birds can be removed and Jill got quite mad when she heard their birds is still grammatical). Consistent with a perceptual grouping hypothesis, participants were more likely to report critical FWs when distal rhythm (repeating ternary or binary pitch patterns) matched the rhythm in the FW-containing region than when it did not. Notably, effects of distal rhythm and distal rate were additive. Results demonstrate a novel effect of distal rhythm on the amount of lexical material listeners hear, highlighting the importance of distal timing information and providing new constraints for models of spoken word recognition.     

Honneth and the Communitarians: Towards a Recognitive Critical Theory of Community

This paper attempts to sketch a critical model of political community by drawing upon recent contributions to the theory of ‘recognition’, particularly in the work of Axel Honneth. The paper proceeds by, first, delineating key features shared by a range of positions associated with ‘communitarianism’, along with the limitations and problems incurred by these commitments. The second part of the paper attempts to mobilise Honneth’s theoretical work to develop a conception of community that shares a number of the basic premisesvis-á-vis political life associated with communitarianism, but which nevertheless accommodates the reservations expressed by communitarianism’s liberal and radical critics. This revised version was published online in August 2006 with corrections to the Cover Date.     

Phonology: A Review and Proposals from a Connectionist Perspective

A parallel distributed processing (PDP) model of phonological processing is developed, including components to support repetition, auditory processing, comprehension, and language production. From the performance of the PDP reading model of Plaut, McClelland, Seidenberg, and Patterson (1996), it is inferred that the acoustic-articulatory motor pattern associator that supports repetition provides the basis for phonological sequence knowledge. From the observation that many patients make phonemic paraphasic errors in language production, as in repetition, it is argued that there must be a direct link between distributed concept representations (lexical semantic knowledge) and this network representation of sequence knowledge. In this way, both lexical semantic and phonotactic constraints are brought to bear on language production. The literature on phonological function in normal subjects (slip-of-the-tongue corpora) and in patients with aphasia is critically reviewed from this perspective. The relationship between acoustic and articulatory motor representations in the process of phonetic perception is considered. Repetition and reproduction conduction aphasia are reviewed in detail and extended consideration is given to the representation of auditory verbal short-term memory in the model. Finally, the PDP model is reconciled with information processing models of phonological processing, including that of Lichtheim, and with current knowledge of the anatomic localization of phonological processing. Although no simulations of the model were run, a number of simulation studies are proposed.     

Hierarchical Multinomial Processing Tree Models: A Latent-Class Approach

Multinomial processing tree models are widely used in many areas of psychology. Their application relies on the assumption of parameter homogeneity, that is, on the assumption that participants do not differ in their parameter values. Tests for parameter homogeneity are proposed that can be routinely used as part of multinomial model analyses to defend the assumption. If parameter homogeneity is found to be violated, a new family of models, termed latent-class multinomial processing tree models, can be applied that accommodates parameter heterogeneity and correlated parameters, yet preserves most of the advantages of the traditional multinomial method. Estimation, goodness-of-fit tests, and tests of other hypotheses of interest are considered for the new family of models. The author thanks Bill Batchelder, Edgar Erdfelder, Thorsten Meiser, and Christoph Stahl for helpful comments on a previous version of this paper. The author is also grateful to Edgar Erdfelder for making available the data set analyzed in this paper.     

Semiparametric Thurstonian Models for Recurrent Choices: A Bayesian Analysis

We develop semiparametric Bayesian Thurstonian models for analyzing repeated choice decisions involving multinomial, multivariate binary or multivariate ordinal data. Our modeling framework has multiple components that together yield considerable flexibility in modeling preference utilities, cross-sectional heterogeneity and parameter-driven dynamics. Each component of our model is specified semiparametrically using Dirichlet process (DP) priors. The utility (latent variable) component of our model allows the alternative-specific utility errors to semiparametrically deviate from a normal distribution. This generates a robust alternative to popular Thurstonian specifications that are based on underlying normally distributed latent variables. Our second component focuses on flexibly modeling cross-sectional heterogeneity. The semiparametric specification allows the heterogeneity distribution to mimic either a finite mixture distribution or a continuous distribution such as the normal, whichever is supported by the data. Thus, special features such as multimodality can be readily incorporated without the need to overtly search for the best heterogeneity specification across a series of models. Finally, we allow for parameter-driven dynamics using a semiparametric state-space approach. This specification adds to the literature on robust Kalman filters. The resulting framework is very general and integrates divergent strands of the literatures on flexible choice models, Bayesian nonparametrics and robust time series specifications. Given this generality, we show how several existing Thurstonian models can be obtained as special forms of our model. We describe Markov chain Monte Carlo methods for the inference of model parameters, report results from two simulation studies and apply the model to consumer choice data from a frequently purchased product category. The results from our simulations and application highlight the benefits of using our semiparametric approach.     

The Utility of Connectionism for Motor Learning: A Reinterpretation of Contextual Interference in Movement Schemas

Using the general framework of schema theory, and building on it, the present article takes a connectionist approach to motor learning and to contextual interference effects. These phenomena were simulated in an exploratory manner in neural networks. The outcome closely reflects previous research with humans. In a simulated ballistic movement task, networks performed worse during practice but showed better transfer when target movement distances were presented in a random rather than a blocked fashion. Connectionism provides a parsimonious account of the effect in terms of properties inherent in the parallel distributed network.     

The Word Superiority Effect in central and peripheral vision

The Word Superiority Effect (WSE) is a well-known phenomenon in reading research, where words are reported more accurately than single letters or non-words. We report two experiments that investigate the WSE in the central and peripheral visual field, as well as laterality differences in the perception of words and letters, using methods based on the Theory of Visual Attention. The results show a WSE in the central visual field, reflected in mean scores, perception thresholds, and processing speed, whereas the effect is eliminated or reversed in the periphery. This may be caused by crowding, which prevents lexical analysis of a word in the periphery. We conclude that perception of words and letters differs according to location in the visual field. Linking our results to previous studies of crowding effects in patients with reading impairments, we hypothesize that similar mechanisms may limit normal word peripheral processing.     

Chinese-English biliteracy acquisition: cross-language and writing system transfer

This study investigated cross-language and writing system relationship in biliteracy acquisition of children learning to read two different writing systems-Chinese and English. Forty-six Mandarin-speaking children were tested for their first language (Chinese-L1) and second language (English-L2) reading skills. Comparable experiments in Chinese and English were designed focusing on two reading processes-phonological and orthographic processing. Word reading skills in both writing systems were tested. Results revealed that Chinese onset matching skill was significantly correlated with English onset and rime matching skills. Pinyin, an alphabetic phonetic system used to assist children in learning to read Chinese characters, was highly correlated with English pseudoword reading. Furthermore, Chinese tone processing skill contributed a moderate but significant amount of variance in predicting English pseudoword reading even when English phonemic-level processing skill was taken into consideration. Orthographic processing skill in the two writing systems, on the other hand, did not predict each other's word reading. These findings suggest that bilingual reading acquisition is a joint function of shared phonological processes and orthographic specific skills.