this article may be sent to David Plaut, Mellon Institute 115-- CNBC, Carnegie Mellon University, 4400 Fifth Avenue, Pittsburgh PA 15213--2683; email: plaut@cmu.edu

This study explores the self-organizing neural network as a model of lexical and morphological acquisition. We examined issues of generalization, representation, and recovery in a multiple feature-map model as implemented in DISLEX (Miikkulainen, 1997). Our results indicate that self-organization and Hebbian learning are two important computational principles that can account for the psycholinguistic processes of semantic representation, morphological generalization, and recovery from generalizations in the acquisition of reversive prefixes such as un- and dis-. These results attest to the utility of self-organizing neural networks in the study of language acquisition. Introduction Language learning is characterized by the learner's ability to generalize beyond what is heard in the input. One current debate on connectionist models of language acquisition concerns the issue of generalization (Elman, 1998). Probably the best-known example in this debate has to do with the acquisition o...

A long-standing debate regarding the representation of semantic knowledge is whether such knowledge is represented in a single, amodal system or whether it is organised into multiple subsystems based on modality of input or type of information. The current paper presents a distributed connectionist model of semantics that constitutes a middle ground between these unitary- versus multiple-semantics accounts. In the model, semantic representations develop under the pressure of learning to mediate between multiple input and output modalities in performing various tasks. The system has a topographic bias on learning that favours short connections, leading to a graded degree of modality-specific functional specialisation within semantics. The model is applied to the specific empirical phenomena of optic aphasia—a neuropsychological disorder in which patients exhibit a selective deficit in naming visually presented objects that is not attributable to more generalised impairments in object recognition (visual agnosia) or naming (anomia). As a result of the topographic bias in the model, as well as the relative degrees of systematicity among tasks, damage to connections from vision to regions of semantics near phonology impairs visual object naming far more than visual gesturing or tactile naming, as observed in optic aphasia. Moreover, as in optic aphasia, the system is better at generating the name of an action associated with an object than at generating the name of the object itself, because action naming receives interactive support from the activation of action representations. The ability of the model to account for the pattern of performance observed in optic aphasia across the full range of severity of impairment provides support for the claim that semantic representations exhibit graded functional specialisation rather than being entirely amodal or modality-specific.

Current connectionist models of bilingual language processing have been largely restricted to localist stationary models. To fully capture the dynamics of bilingual processing, we present SOMBIP, a self-organizing model of bilingual processing that has learning characteristics. SOMBIP consists of two interconnected selforganizing neural networks, coupled with a recurrent neural network that computes lexical co-occurrence constraints. Simulations with our model indicate that (1) the model can account for distinct patterns of the bilingual lexicon without the use of language nodes or language tags, (2) it can develop meaningful lexical-semantic categories through self-organizing processes, and (3) it can account for a variety of priming and interference effects based on associative pathways between phonology and semantics in the lexicon, and (4) it can explain lexical representation in bilinguals with different levels of proficiency and working memory capacity. These capabilities of our model are due to its design characteristics in that (a) it combines localist and distributed properties of processing, (b) it combines representation and learning, and (c) it combines lexicon and sentences in bilingual processing. Thus, SOMBIP serves as a new model of bilingual processing and provides a new perspective on connectionist bilingualism. It has the potential of explaining a wide variety of empirical and theoretical issues in bilingual research.

This study uses self-organizing feature maps to model the acquisition of lexical and grammatical aspect. Previous research has identified a strong association between lexical aspect and grammatical aspect in child language, on the basis of which some researchers proposed innate semantic categories (Bickerton, 1984) or prelinguistic semantic space (Slobin, 1985). Our simulations indicate that this association can be modeled by self-organization and Hebbian learning principles in a feature-map model, without making particular assumptions about the structure of innate knowledge. In line with results from Li (1999), our study further attests to the utility of self-organizing neural networks in the study of language acquisition. Introduction Most linguistic theories of tense and aspect recognize two kinds of aspect: lexical aspect refers to the inherent temporal meanings of a verb, whereas grammatical aspect refers to a particular viewpoint toward the described situation. For ...

In this paper we present DevLex-II, a self-organizing neural network model of early word production. It consists of three self-organizing feature maps (a semantic layer, a phonological layer and a phonemic layer) that are connected via associative links trained by Hebbian learning. We use this model to simulate the early stages of lexical acquisition in children. The simulating results indicate a number of important effects in determining the timing and function of children’s word production, such as word frequency and word length effects. In addition, results from lesioned models indicate developmental plasticity in the network’s recovery from damage. Plasticity occurs at early stages, and changes with time in a non-monotonic and nonlinear fashion. These simulated patterns are due to the nonlinear dynamic properties of the network and match up with data from empirical studies of children.

Despite gspite interests in the acquisition of Chinese orthography, few studies have modeled the acquisition process using connectionist networks. This study uses a self-organizing connectionist model to simulate children's learning of Chinese characters. There are two major goals of our study: (1) To evaluate the degree to which connectionist models can inform us of the complex structural and processing properties of the Chinese orthography. One of the most difficult tasks in achieving this goal is is how to faithfully capture the orthographic...

We present a recurrent multimodal model of binding written words to mental objects or concepts and investigate the capability of the network in reading misspelt but categorically related words. Our model consists of three mutually interconnected association modules which store mental objects, represent their written names and bind these together to form mental concepts. A controllable feedback gain term controlling top-down influence is incorporated into the model architecture and it is shown that correct settings for this during map formation and simulated reading experiments is necessary for correct interpretation and semantic binding of the written words. 1

2 Brief review of the state of the art 2