Does knowledge of language consist of mentally-represented rules? Rumelhart and McClelland have described a connectionist (parallel distributed processing) model of the acquisition of the past tense in English which successfully maps many stems onto their past tense forms, both regular (walk/walked) and irregular (go/went), and which mimics some of the errors and sequences of development of children. Yet the model contains no explicit rules, only a set of neuron-style units which stand for trigrams of phonetic features of the stem, a set of units which stand for trigrams of phonetic features of the past form, and an array of connections between the two sets of units whose strengths are modified during learning. Rumelhart and McClelland conclude that linguistic rules may be merely convenient approximate fictions and that the real causal processes in language use and acquisition must be characterized as the transfer of activation levels among units and the modification of the weights of their connections. We analyze both the linguistic and the developmental assumptions of the model in detail and discover that (1) it cannot represent certain words, (2) it cannot learn many rules, (3) it can learn rules found in no human language, (4) it cannot explain morphological and phonological regularities, (5) it cannot explain the differences between irregular and regular forms, (6) it fails at its assigned task of mastering the past tense of English, (7) it gives an incorrect explanation for two developmental phenomena: stages of overregularization of irregular forms such as bringed, and the appearance of doubly-marked forms such as ated, and (8) it gives accounts of two others (infrequent overregularization of verbs ending in t/d, and the order of acquisition of different irregula...

An approach to connectionist natural language processing is proposed, which is based on hierarchically organized modular Parallel Distributed Processing (PDP) networks and a central lexicon of distributed input/output representations. The modules communicate using these representations, which are global and publicly available in the system. The representations are developed automatically by all networks while they are learning their processing tasks. The resulting representations reflect the regularities in the subtasks, which facilitates robust processing in the face of noise and damage, supports improved generalization, and provides expectations about possible contexts. The lexicon can be extended by cloning new instances of the items, that is, by generating a number of items with known processing properties and distinct identities. This technique combinatorially increases the processing power of the system. The recurrent FGREP module, together with a central lexicon, is used as a ba...

Supporters of eliminative connectionism have argued for a pattern association based explanation of language learning and language processing. They deny that explicit rules and symbolic representations play any role in language processing and cognition in general. Their argument is based to a large extent on two artificial neural network (ANN) models that are claimed to be able to learn the past tenses of English verbs. (Rumelhart and McClelland, 1986; MacWhinney and Leinbach, 1991). In this article we critically review Rumelhart and McClelland's as well as MacWhinney and Leinbach's ANN-models and conclude that they do not succeed in the assigned task of learning the past tenses of English verbs. In order to answer their challenge to the symbolic processing approach, we present our Symbolic Pattern Associator (SPA) -- a general purpose pattern associator that can learn to associate arbitrary discrete patterns. We carried out several experiments with the SPA using the same set of verbs ...

This paper discusses a perceptron model of the learning and assignment of linguistic stress, using data from nineteen human languages. First, we point out some interesting parallels between aspects of the model and the constructs and predictions of metrical phonology, the linguistic theory of stress. Second, we develop a novel analysis of linguistic stress in terms of ease of perceptron-learnability. These two sets of results suggest that simple statistical learning techniques have the potential to complement, and provide computational validation for, abstract theoretical investigations of language. We then examine why such methodologies should be of interest for linguistic theorizing. Our analysis began at a high level by observing inherent characteristics of various stress systems, much as theoretical linguistics does. However, our explanations changed substantially whenwe included a detailed account of the model's processing mechanisms. Our higher-level, theoretical accou...

Children's (Grades 1 to 5) implicit learning of French orthographic regularities were investigated through nonword judgment (Experiments 1 and 2) and completion (Experiments 3a and 3b) tasks. Children were increasingly sensitive to (a) the frequency of double consonants (Experiments 1, 2 and 3a), (b) the fact that vowels can never be doubled (Experiment 2) and (c) the legal position of double consonants (Experiments 2 and 3b). The later effect transferred to never doubled consonants, although with a decrement in performance. Moreover, this decrement persisted without any trend towards fading even after the massive amounts of experience provided by years of practice. This result runs against the idea that transfer to novel material is indicative of abstract rule-based knowledge, and suggests instead the action of mechanisms sensitive to the statistical properties of the material. A connectionist model is proposed as an instantiation of such mechanisms.

If a person doesn’t give something to me, I won’t give anything to that person. If I’m sitting eating, and someone like that comes by, I say, “Uhn, uhn. I’m not going to give any of this to you. When you have food, the things you do with it make me unhappy. If you even once in a while gave me something nice, I would surely give some of this to you.” Nisa from Nisa: The Life and Words of a!Kung Woman, Shostak, 1981, p. 89 Instead of keeping things, [!Kung] use them as gifts to express generosity and friendly intent, and to put people under obligation to make return tokens of friendship....In reciprocating, one does not give the same object back again but something of comparable value. Eland fat is a very highly valued gift...Toma said that when he had eland fat to give, he took shrewd note of certain objects he might like to have and gave their owners especially generous gifts of fat. Marshall, 1976, pp. 366–369

This paper is concerned with the acquisition of the lexicon, In particular, we propose a method that uses analo. gieal reasoning to hypothesize new polysemous word senses. This method is one of a number of knowledge acquisition devices to be included in DIRC (Domain Independent Retargetable Consultan0. DIRC is a kind of intelligent, natural language-capable consultant kit that can be remrgeted at diftrent domains. DIRC is essentially "emptyUC" CONIX Consultant, Wilensky et al., 1988). DIRC is to include the lzmguage and reasoning mechanisms of UC, plus a large grammar and a ge, neral lexk:on. The user must then add domain kmwledge, user knowledge and lexical howledge for the area of interest 2. 'revous Work h Acqdsifio o ' the Lexicm

The German plural system has become a focal point for conflicting theories of language, both linguistic and cognitive. We present simulation results with three simple classifiers – an ordinary nearest neighbour algorithm, Nosofsky’s ‘Generalized Context Model ’ (GCM) and a standard, three-layer backprop network – predicting the plural class from a phonological representation of the singular in German. Though these are absolutely ‘minimal’ models, in terms of architecture and input information, they nevertheless do remarkably well. The nearest neighbour predicts the correct plural class with an accuracy of 72 % for a set of 24,640 nouns from the CELEX database. With a subset of 8,598 (noncompound) nouns, the nearest neighbour, the GCM and the network score 71.0%, 75.0 % and 83.5%, respectively, on novel items. Furthermore, they outperform a hybrid, ‘pattern-associator + default rule’, model, as proposed by Marcus et al. (1995), on this data set.

This article analyzes the relationship between skill learning and repetition priming, 2 implicit memory phenomena. A number of reports have suggested that skill learning and repetition priming can be dissociated from each other and are therefore based on different mechanisms. The authors present a theoretical analysis showing that previous results cannot be regarded as evidence of a processing dissociation between skill learning and repetition priming. The authors also present a single-mechanism computational model that simulates a specific experimental task and exhibits both skill learning and repetition priming, as well as a number of apparent dissociations between these measures. These theoretical and computational analyses provide complementary evidence that skill learning and repetition priming are aspects of a single underlying mechanism that has the characteristics of procedural memory. One of the most significant developments in the study of human memory over the last two decades has been the discovery of a dissociation between two different kinds of memory systems. An early indication of this dissociation came from studies of amnesia in patients with excision or lesions of the hippocampus (Scoville & Milner, 1957). These patients were dramatically impaired in their

This article discusses a recurrent connectionist network, simulating empirical phenomena usually explained by current dual-process approaches of attitudes, thereby focusing on the processing mechanisms that may underlie both central and peripheral routes of persuasion. Major findings in attitude formation and change involving both processing modes are reviewed and modeled from a connectionist perspective. We use an autoassociative network architecture with a linear activation update and the delta learning algorithm for adjusting the connection weights. The network is applied to well-known experiments involving deliberative attitude formation, as well as the use of heuristics of length, consensus, expertise, and mood. All these empirical phenomena are successfully reproduced in the simulations. Moreover, the proposed model is shown to be consistent with algebraic models of attitude formation (Fishbein & Ajzen, 1975). The discussion centers on how the proposed network model may be used to unite and formalize current ideas and hypotheses on the processes underlying attitude acquisition and how it can be deployed to develop novel hypotheses in the attitude domain.