We develop a connectionist approach to processing in quasi-regular domains, as exemplified by English word reading. A consideration of the shortcomings of a previous implementation (Seidenberg & McClelland, 1989, Psych. Rev.) in reading nonwords leads to the development of orthographic and phonological representations that capture better the relevant structure among the written and spoken forms of words. In a number of simulation experiments, networks using the new representations learn to read both regular and exception words, including low-frequency exception words, and yet are still able to read pronounceable nonwords as well as skilled readers. A mathematical analysis of the effects of word frequency and spelling-sound consistency in a related but simpler system serves to clarify the close relationship of these factors in influencing naming latencies. These insights are verified in subsequent simulations, including an attractor network that reproduces the naming latency data directly in its time to settle on a response. Further analyses of the network's ability to reproduce data on impaired reading in surface dyslexia support a view of the reading system that incorporates a graded division-of-labor between semantic and phonological processes. Such a view is consistent with the more general Seidenberg and McClelland framework and has some similarities with-but also important differences from-the standard dual-route account.Many aspects of language can be characterized as quasiregular-the relationship between inputs and outputs is systematic but admits many exceptions. One such task is the mapping between the written and spoken forms of English words. Most words are regular (e.g., GAVE, MINT) in that their pronunciations adhere to standard spelling-sound correspondences. There are, however, many irregular or exception words (e.g., HAVE, PINT) whose pronunciations violate the standard correspondences. To make matters worse, some spelling patterns have a range of pronunciations with none clearly predominating (e.g., OWN in DOWN, TOWN, BROWN, This research was supported financially by the National Institute of Mental Health (Grants MH47566 and MH00385), the National Institute on Aging (Grant Ag10109), the National Science Foundation (Grant ASC-9109215), and the McDonnell-Pew Program in Cognitive Neuroscience (Grant T89-01245-016).We thank Marlene Behrmann, Derek Besner, Max Coltheart, Joe Devlin, Geoff Hinton, and Eamon Strain for helpful discussions and comments. We also acknowledge Derek Besner, Max Coltheart, and Michael McCloskey for directing attention to many of the issues addressed in this paper.Correspondence concerning this paper should be sent to Dr. David C. Plaut, Department of Psychology, Carnegie Mellon University, Pittsburgh, PA 15213-3890, plaut@cmu.edu. CROWN vs. KNOWN, SHOWN, GROWN, THROWN, or OUGH in COUGH, ROUGH, BOUGH, THOUGH, THROUGH). Nonetheless, in the face of this complexity, skilled readers pronounce written words quickly and accurately, and can also use their knowledge...

No abstract

In everyday tasks, selecting actions in the proper sequence requires a continuously updated representation of temporal context. Previous models have addressed this problem by positing a hierarchy of processing units, mirroring the roughly hierarchical structure of naturalistic tasks themselves. The present study considers an alternative framework, in which the representation of context depends on recurrent connections within a network mapping from environmental inputs to actions. The ability of this approach to account for human performance was evaluated by applying it, through simulation, to a specific everyday task. The resulting model learned to deal flexibly with a complex set of sequencing constraints, encoding contextual information at multiple time scales within a single, distributed internal representation. Degrading this representation led to errors resembling those observed both in everyday behavior and in apraxia. Analysis of the model's function yielded numerous predictions relevant to both normal and apraxic performance.

Existing accounts of single-word semantic priming phenomena incorporate multiple mechanisms, such as spreading activation, expectancy-based processes, and postlexical semantic matching. The authors provide empirical and computational support for a single-mechanism distributed network account. Previous studies have found greater semantic priming for low- than for high-frequency target words as well as inhibition following unrelated primes only at long stimulus-onset asynchronies (SOAs). A series of experiments examined the modulation of these effects by individual differences in age or perceptual ability. Third-grade, 6th-grade, and college students performed a lexical-decision task on high- and low-frequency target words preceded by related, unrelated, and nonword primes. Greater priming for low-frequency targets was exhibited only by participants with high perceptual ability. Moreover, unlike the college students, the children showed no inhibition even at the long SOA. The authors provide an account of these results in terms of the properties of distributed network models and support this account with an explicit computational simulation.

Despite a century of research, the mechanisms underlying short-term or working memory for serial order remain uncertain. Recent theoretical models have converged on a particular account, based on transient associations between independent item and context representations. In the present article, the authors present an alternative model, according to which sequence information is encoded through sustained patterns of activation within a recurrent neural network architecture. As demonstrated through a series of computer simulations, the model provides a parsimonious account for numerous benchmark characteristics of immediate serial recall, including data that have been considered to preclude the application of recurrent neural networks in this domain. Unlike most competing accounts, the model deals naturally with findings concerning the role of background knowledge in serial recall and makes contact with relevant neuroscientific data. Furthermore, the model gives rise to numerous testable predictions that differentiate it from competing theories. Taken together, the results presented indicate that recurrent neural networks may offer a useful framework for understanding short-term memory for serial order.