This paper introduces a special issue of Cognitive Science initiated on the 25th anniversary of the publication of Parallel Distributed Processing (PDP), a two-volume work that introduced the use of neural network models as vehicles for understanding cognition. The collection surveys the core commitments of the PDP framework, the key issues the framework has addressed, and the debates the framework has spawned, and presents viewpoints on the current status of these issues. The articles focus on both historical (...) roots and contemporary developments in learning, optimality theory, perception, memory, language, conceptual knowledge, cognitive control, and consciousness. Here we consider the approach more generally, reviewing the original motivations, the resulting framework, and the central tenets of the underlying theory. We then evaluate the impact of PDP both on the field at large and within specific subdomains of cognitive science and consider the current role of PDP models within the broader landscape of contemporary theoretical frameworks in cognitive science. Looking to the future, we consider the implications for cognitive science of the recent success of machine learning systems called “deep networks”—systems that build on key ideas presented in the PDP volumes. (shrink)

Humans can pronounce a nonword. Some researchers have interpreted this behavior as requiring a sequential mechanism by which a grapheme-phoneme correspondence rule is applied to each grapheme in turn. However, several parallel-distributed processing models in English have simulated human nonword reading accuracy without a sequential mechanism. Interestingly, the Japanese psycholinguistic literature went partly in the same direction, but it has since concluded that a sequential parsing mechanism is required to reproduce human nonword reading accuracy. In this study, (...) by manipulating the list composition, we demonstrated that past psycholinguistic studies in Japanese have overestimated human nonword reading accuracy. When the more fairly reevaluated human performance was targeted, a newly implemented Japanese PDP model simulated the target accuracy as well as the error patterns. These findings suggest that PDP models are a more parsimonious way of explaining reading across various languages. (shrink)

Parallel distributed processing is transforming the field of cognitive science. Microcognition provides a clear, readable guide to this emerging paradigm from a cognitive philosopher's point of view. It explains and explores the biological basis of PDP, its psychological importance, and its philosophical relevance.

This paper critically examines the claim that parallel distributed processing (PDP) networks are autonomous learning systems. A PDP model of a simple distributed associative memory is considered. It is shown that the 'generic' PDP architecture cannot implement the computations required by this memory system without the aid of external control. In other words, the model is not autonomous. Two specific problems are highlighted: (i) simultaneous learning and recall are not permitted to occur as would be required (...) of an autonomous system; (ii) connections between processing units cannot simultaneously represent current and previous network activation as would be required if learning is to occur. Similar problems exist for more sophisticated networks constructed from the generic PDP architecture. We argue that this is because these models are not adequately constrained by the properties of the functional architecture assumed by PDP modelers. It is also argued that without such constraints, PDP researchers cannot claim to have developed an architecture radically different from that proposed by the Classical approach in cognitive science. (shrink)

My purpose in this brief paper is to consider the implications of a radically different computer architecure to some fundamental problems in the foundations of Cognitive Science. More exactly, I wish to consider the ramifications of the 'Gödel-Minds-Machines' controversy of the late 1960s on a dynamically changing computer architecture which, I venture to suggest, is going to revolutionize which 'functions' of the human mind can and cannot be modelled by (non-human) computational automata. I will proceed on the presupposition that the (...) reader is familiar with some of the fundamentals of computational theory and mathematical logic. (shrink)

The will to power has non-trivial physical models taken from the class of parallel dis¬tributed processing systems, specifically wave-mechanical discrete dynamical systems with cyclical entropy. The will to power is thus linked to research in non-linear self-organizing dynami¬cal systems, includ¬ing oscillons, cellular automata, spin-glasses, Ising systems, and connectionist networks.

In this prcis we focus on phenomena central to the reaction against similarity-based theories that arose in the 1980s and that subsequently motivated the approach to semantic knowledge. Specifically, we consider (1) how concepts differentiate in early development, (2) why some groupings of items seem to form or coherent categories while others do not, (3) why different properties seem central or important to different concepts, (4) why children and adults sometimes attest to beliefs that seem to contradict their direct experience, (...) (5) how concepts reorganize between the ages of 4 and 10, and (6) the relationship between causal knowledge and semantic knowledge. The explanations our theory offers for these phenomena are illustrated with reference to a simple feed-forward connectionist model. The relationships between this simple model, the broader theory, and more general issues in cognitive science are discussed. (shrink)

The efficient processing of large-scale data has very important practical value. In this study, a data mining platform based on Hadoop distributed file system was designed, and then K-means algorithm was improved with the idea of max-min distance. On Hadoop distributed file system platform, the parallelization was realized by MapReduce. Finally, the data processing effect of the algorithm was analyzed with Iris data set. The results showed that the parallel algorithm divided more correct samples than (...) the traditional algorithm; in the single-machine environment, the parallel algorithm ran longer; in the face of large data sets, the traditional algorithm had insufficient memory, but the parallel algorithm completed the calculation task; the acceleration ratio of the parallel algorithm was raised with the expansion of cluster size and data set size, showing a good parallel effect. The experimental results verifies the reliability of parallel algorithm in big data processing, which makes some contributions to further improve the efficiency of data mining. (shrink)

The parallel distributed processing (PDP) perspective brings forward the important point that all semantic phenomena are based on analog underlying mechanisms, involving the weighted summation of multiple inputs by individual neurons. It falls short of indicating, however, how the essentially discrete nature of semantic processing may emerge at the cognitive level. Bridging this gap probably requires attractor networks.

Emotion Review, Volume 14, Issue 2, Page 99-110, April 2022. Researchers often disagree as to whether emotions are largely consistent across people and over time, or whether they are variable. They also disagree as to whether emotions are initiated by appraisals, or whether they may be initiated in diverse ways. We draw upon Parallel-Distributed-Processing to offer an algorithmic account in which features of an emotion instance are bi-directionally connected to each other via conjunction units. We propose that (...) such indirect connections may be innate as well as learned. These ideas lead to the development of the Interactive Activation and Competition framework for Emotion which allows us to specify when emotions are consistent and when they are variable, as well as when they are appraisal-led and when they are input-agnostic. (shrink)

During active vision, the eyes continually scan the visual environment using saccadic scanning movements. This target article presents an information processing model for the control of these movements, with some close parallels to established physiological processes in the oculomotor system. Two separate pathways are concerned with the spatial and the temporal programming of the movement. In the temporal pathway there is spatially distributed coding and the saccade target is selected from a Both pathways descend through a hierarchy of (...) levels, the lower ones operating automatically. Visual onsets have automatic access to the eye control system via the lower levels. Various centres in each pathway are interconnected via reciprocal inhibition. The model accounts for a number of well-established phenomena in target-elicited saccades: the gap effect, express saccades, the remote distractor effect, and the global effect. High-level control of the pathways in tasks such as visual search and reading is discussed; it operates through spatial selection and search selection, which generally combine in an automated way. The model is examined in relation to data from patients with unilateral neglect. (shrink)

This paper provides an overview of developments in a dual processing theory of automatic and controlled processing that began with the empirical and theoretical work described by Schneider and Shiffrin (1977) and Shiffrin and Schneider (1977) over a quarter century ago. A review of relevant empirical findings suggests that there is a set of core behavioral phenomena reflecting differences between controlled and automatic processing that must be addressed by a successful theory. These phenomena relate to: consistency in (...) training, serial versus parallel processing, level of effort, robustness to stressors, degree of control, effects on long‐term memory, and priority encoding. We detail a computational model of controlled processing, CAP2, that accounts for these phenomena as emergent properties of an underlying hybrid computational architecture. The model employs a large network of distributed data modules that can categorize, buffer, associate, and prioritize information. Each module is a connectionist network with input and output layers, and each module communicates with a central Control System by outputting priority and activity report signals, and by receiving control signals. The Control System is composed of five processors including a Goal Processor, an Attention Controller, an Activity Monitor, an Episodic Store, and a Gating & Report Relay. The transition from controlled to automatic processing occurs in this model as the data modules become capable of transmitting their output without mediation by the Control System. We describe recent progress in mapping the components of this model onto specific neuroanatomical substrates, briefly discuss the potential for applying functional neuroimaging techniques to test the model's predictions, and its relation to other models. (shrink)

Connectionism—also known as parallel distributed processing, or neural network modeling—offers promise as a framework to unite clinical and cognitive psychology, and as a tool for studying conscious and unconscious mental activity. This paper describes a neural network model of the case study of Lucy R., from Freud and Breuer's Studies on Hysteria. Though very simple in architecture, the network spontaneously displays analogues of repression and hallucination, corresponding to Lucy R.'s symptoms. Salient elements of Lucy's conscious experience are (...) represented in the model by the activations of neuronlike processors in a fully interconnected network, without hidden units. The model learns to associate elements of experience that were associated in Lucy's case history. Some of these configurations of elements were traumatic for Lucy; trauma is modeled in the network by "emphatic learning," learning accomplished at an abnormally high learning rate. The model suggests that changing associations among conscious elements are sufficient to generate the symptoms Freud observed: apparent repression, hallucination, and recovery through therapy. In the case of Lucy R., Freud's theoretical inference regarding active but unconscious thought is not required by his data. Instead, the unconscious can be understood as a set of complex dispositions embodied in connections between elements of conscious experience. (shrink)

Something remarkable is happening in the cognitive sciences. After a quarter of a century of cognitive models that were inspired by the metaphor of the digital computer, the newest cognitive models are inspired by the properties of the brain itself. Variously referred to as connectionist, parallel distributed processing, or neutral network models, they explore the idea that complex intellectual operations can be carried out by large networks of simple, neuron-like units. The units themselves are identical, very low-level (...) and 'stupid'. Intelligent performance is derived from the pattern of connection strengths between units, and the fundamental cognitive activity is pattern recognition and completion. Connectionism and the Mind provides an introduction to this newly emerging approach to understanding the mind. The first few chapters focus on network architecture, offering accessible treatment of the equations that describe learning and the propagation of activation. Furthermore, the reader is walked step-by-step through the activities of networks engaged in pattern recognition, learning, and cognitive tasks such as memory retrieval and prototype formation. The remainder of the book addresses the implications of connectionism for theories of the mind, both philosophical and psychological. Foe example: What Role is played by pattern recognition and completion as basic as cognitive functions? Connectionist models have particular strength in learning and pattern recognition; should they be limited to those functions, or can they provide an overall account of cognitive functioning? In particular, can connectionist models provide an adequate account of the ability to employ linguistic and other symbol systems, or must an adequate system incorporate symbol processing as a basic cognitive capacity? Finally, Connectionism and the Mind examines the relation of connectionist models to philosophical accounts of propositional attitudes, and to a variety of other inquiries in cognitive psychology, linguistics, developmental psychology, artificial intelligence and neuroscience. (shrink)

Mental representations have continuous as well as discrete, combinatorial properties. For example, while predominantly discrete, phonological representations also vary continuously; this is reflected by gradient effects in instrumental studies of speech production. Can an integrated theoretical framework address both aspects of structure? The framework we introduce here, Gradient Symbol Processing, characterizes the emergence of grammatical macrostructure from the Parallel Distributed Processing microstructure (McClelland, Rumelhart, & The PDP Research Group, 1986) of language processing. The mental representations (...) that emerge, Distributed Symbol Systems, have both combinatorial and gradient structure. They are processed through Subsymbolic Optimization–Quantization, in which an optimization process favoring representations that satisfy well-formedness constraints operates in parallel with a distributed quantization process favoring discrete symbolic structures. We apply a particular instantiation of this framework, λ-Diffusion Theory, to phonological production. Simulations of the resulting model suggest that Gradient Symbol Processing offers a way to unify accounts of grammatical competence with both discrete and continuous patterns in language performance. (shrink)

There has been a lot of strong evidence showing that human cognition works not in a central processing way but in a distributed way. As well known, human brain processes huge information in a parallel and distributed way. Recently cognitive scientists have contended that the minds are embodied in environment. These two ideas of distribution in cognition and embodiment in the mind can go along overall, but there is a tension between them in some specific respects, (...) especially in the matter of representation. The aim of this paper is to examine the possibility of the embodied distributed cognition by focusing on the concept of mental representation. Firstly, I shall examine the nature of embodied mind and distributed cognition. Secondly I shall make a distinction between those ideas that the notion of embodiment can be confined to the mind but the notion of distribution can be applied to both the mind and to its environment. This implies a difference of applicability of those notions. That is, while the suitable application domain of the latter is scientific cognition, that of the former is our mind. This difference can throw light upon untangling the dispute between Churchland's internalism and Giere's externalism of presentation. (shrink)

Let’s consider the following paradox (Fodor [1989], Jackson and Petit [1988] [1992], Drestke [1988], Block [1991], Lepore and Loewer [1987], Lewis [1986], Segal and Sober [1991]): i) The intentional content of a thought (or any other intentional state) is causally relevant to its behavioural (and other) effects. ii) Intentional content is nothing but the meaning of internal representations. But, iii) Internal processors are only sensitive to the syntactic structures of internal representations, not their meanings. Therefore it seems that if we (...) want to defend the idea -absolutely plausible from an intuitive point of view- that mental / intentional states are causally responsible for behavioural outputs and we want to do it on the physicalist basis of any scientific methodology, we will have to give up the conviction that such intentional states qua intentional, i. e. as having a particular meaning, are the ones causally responsible for our behaviour. The path that takes us to mental epiphenomenalism is clear: 1) the causal powers of any event are completely determined by its physical properties; 2) although intentional properties supervene on physical properties, they can’t be identified with them; 3) intentional properties, as intentional, are not causally responsible for behaviour, because they don’t take part in the causal powers of the states to which they belong, i. e., intentional properties are epiphenomenal. Let’s consider now a different yet parallel position to the one just described. There is an important debate in cognitive science about whether the class of mechanisms to which we belong and which the computational modelling project of cognitive processes refers to is best represented by classical or connectionist approaches (McClelland, Rumelhart et. al [1986], Smolensky [1987] [1988], Fodor and Pylyshyn [1988], Pinker and Prince [1988], Clark [1989], Ramsey, Stich and Rumelhart [1991], Clark and Karmiloff-Smith [forthcoming]).In classical, serial processing models, information is encoded in terms of rules that have a linguistic character. In connectionist or parallel distributed processing (PDP) models, the causal relationships among the units that constitute the system determine how the information is processes by the network, although these units don’t have a direct semantic interpretation. But, for both, classical and connectionist models, all the computations can be explained without any reference to the content of the processed information, i.e., in both cases the properties that seem to be responsible for the system’s behaviour are ultimately physical properties nor intentional ones. This situation mirrors thus, within cognitive science, the philosophical discussion concerning the causal efficacy of semantic properties. Now, if in this debate we opt for the classical paradigm, there is a way of finding a solution to the computational version of the epiphenomenalism paradox. This solution is based mainly on the notion of supervenience or, more precisely, on the notion of mereological supervenience (Kim [1984] [1988])1 . The idea of intentional properties supervening on physical properties makes sense within the classical context because there exists an easily isolable supervenience base comprising the syntactic items in the socalled language of thought. But, what happens if we opt for the connectionist paradigm?. The situation here doesn’t seem to favour the use of the same supervenience strategy. For it has been argued (Ramsey, Stich and Garon [1991]) that beliefs, desires, and other mental states are nor, in the connectionist paradigm, individuable as weight or activation states of the system. This is because information is encoded by the network in distributed and superpositional representations, i.e., there are no straightforwardly isolable vehicles at the physical level that can be identified as the articulated supervenience base on which the semantic properties supervene2 . If this is true, then the connectionist not only loses the battle against epiphenomenalism but more drastically, seems to offer a standing invitation to eliminativism, since talk of beliefs and desires, etc. now seems to be floating free of any acceptable scientific underpinning, i.e., she has lost the necessary theoretical apparatus for supporting the intuitive idea that propositional attitudes -beliefs, desires and any mental states with semantic content- are physically realized. This second line of argumentation doesn’t take us to a paradox but to a dilemma: either we accept eliminativism, if connectionist hypothesis are correct or we defend the causal efficacy of mental states with semantic content by showing that, after all, connectionist networks are not plausible cognitive models (Davies [1991]). From my point of view, however, both lines of argumentation need to be revised. The aim of this paper is to find an account of the causal efficacy of content that avoids the aforementioned epiphenomenalist objections and that doesn’t require the discovery of inner symbols in the computational modelling of such contentful mental states. In short, the aim is to find a meeting point where a philosophical story about content and cause and the connectionist computational model can be brought together (Cfr. Clark [1989]). (shrink)

Connectionism is a style of modeling based upon networks of interconnected simple processing devices. This style of modeling goes by a number of other names too. Connectionist models are also sometimes referred to as 'Parallel Distributed Processing' (or PDP for short) models or networks.1 Connectionist systems are also sometimes referred to as 'neural networks' (abbreviated to NNs) or 'artificial neural networks' (abbreviated to ANNs). Although there may be some rhetorical appeal to this neural nomenclature, it is (...) in fact misleading as connectionist networks are commonly significantly dissimilar to neurological systems. For this reason, I will avoid using this terminology, other than in direct quotations. Instead, I will follow the practice I have adopted above and use 'connectionist' as my primary term for systems of this kind. (shrink)

It is claimed that there are pre-objective phenomena, which cognitive science should explain by employing the notion of non-conceptual representational content. It is argued that a match between parallel distributed processing (PDP) and non-conceptual content (NCC) not only provides a means of refuting recent criticisms of PDP as a cognitive architecture; it also provides a vehicle for NCC that is required by naturalism. A connectionist cognitive mapping algorithm is used as a case study to examine the affinities (...) between PDP and NCC. (shrink)

This article argues for a task-based approach to identifying and individuating cognitive systems. The agent-based extended cognition approach faces a problem of cognitive bloat and has difficulty accommodating both sub-individual cognitive systems ("scaling down") and some supra-individual cognitive systems ("scaling up"). The standard distributed cognition approach can accommodate a wider variety of supra-individual systems but likewise has difficulties with sub-individual systems and faces the problem of cognitive bloat. We develop a task-based variant of distributed cognition designed to scale (...) up and down smoothly while providing a principled means of avoiding cognitive bloat. The advantages of the task-based approach are illustrated by means of two parallel case studies: re-representation in the human visual system and in a biomedical engineering laboratory. (shrink)

This is an amended version of material that first appeared in A. Clark, Microcognition: Philosophy, Cognitive Science, and Parallel Distributed Processing (MIT Press, Cambridge, MA, 1989), Ch. 1, 2, and 6. It appears in German translation in Metzinger,T (Ed) DAS LEIB-SEELE-PROBLEM IN DER ZWEITEN HELFTE DES 20 JAHRHUNDERTS (Frankfurt am Main: Suhrkamp. 1999).

It is crucial, first of all, to stress the importance Churchland attaches to the idea that the neural networks whose assemblages he holds to be “engines of reason” must be recurrent. Non-recurrent networks, of the sort best known among philosophers, simply discover patterns in input data presented to them as sets of features. The learning capacities of such networks, extensively discussed since the publication of Rumelhart and McClelland et al., are indeed impressive; and Churchland describes them clearly and gracefully as (...) preparation for introducing recurrent networks. Now, the importance of recurrence as a feature of the networks Churchland hypothesizes as forming the basis of cognitive activity is well motivated. A cognitive system that was an assembly of non-recurrent networks would be, in essence, a stimulus-response machine of the sort that early behaviourists took themselves to be studying when they examined the mind/brain. Such a system could learn to find novel patterns in data. However, it is empirically evident that human— and some non-human—cognitive capacities go well beyond this. People dream, imagine scenarios that never existed, reconceptualize their perceptions, and theorize. All these activities seem to require not merely inferring patterns from data, but reconceiving the nature of the data itself in light of the inferential structures built through learning. For this to be possible, the data cannot be “clamped,” as they are in the case of non-recurrent networks. Hence the importance of recurrence. A recurrent network can be defined as one which takes some or all of its own output and then treats that output as a source of input. This allows for the recognition of meta-patterns. Furthermore, if some such networks are feeding their output as input to other networks with which they are linked, the possibility of reasoning and conceptualizing by analogy arises; and this ability, as Churchland argues, seems to be an essential aspect of genuine creativity in science, art, and the continuously ongoing re-design of the social order. The actual existence of simulated recurrent networks, and the capacities they demonstrate, constitute an impressive possibility argument that a system with enormous cognitive plasticity and behavioural flexibility could be built out of parallel distributed processing networks, so long as at least many of these networks were recurrent. The importance of this possibility argument should not be under-emphasized, but it of course invites an obvious question: how good are our grounds for supposing that the possibility in question is actually realized in the biological domain? (shrink)

Presupposing no familiarity with the technical concepts of either philosophy or computing, this clear introduction reviews the progress made in AI since the inception of the field in 1956. Copeland goes on to analyze what those working in AI must achieve before they can claim to have built a thinking machine and appraises their prospects of succeeding. There are clear introductions to connectionism and to the language of thought hypothesis which weave together material from philosophy, artificial intelligence and neuroscience. John (...) Searle's attacks on AI and cognitive science are countered and close attention is given to foundational issues, including the nature of computation, Turing Machines, the Church-Turing Thesis and the difference between classical symbol processing and parallel distributed processing. The book also explores the possibility of machines having free will and consciousness and concludes with a discussion of in what sense the human brain may be a computer. (shrink)

O'Brien & Opie make unnecessary distinctions between vehicle and process theories and neglect empirically based distinctions between conscious and unconscious processing. We argue that phenomenal experience emerges, not just as a byproduct of input-driven parallel distributed processing, but as a result of constructive processing in recurrent neural networks. Stable network states may be necessary, but are not sufficient, for consciousness.

Presupposing no familiarity with the technical concepts of either philosophy or computing, this clear introduction reviews the progress made in AI since the inception of the field in 1956. Copeland goes on to analyze what those working in AI must achieve before they can claim to have built a thinking machine and appraises their prospects of succeeding.There are clear introductions to connectionism and to the language of thought hypothesis which weave together material from philosophy, artificial intelligence and neuroscience. John Searle's (...) attacks on AI and cognitive science are countered and close attention is given to foundational issues, including the nature of computation, Turing Machines, the Church-Turing Thesis and the difference between classical symbol processing and parallel distributed processing. The book also explores the possibility of machines having free will and consciousness and concludes with a discussion of in what sense the human brain may be a computer. (shrink)

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 & McClelland, 1986, Parallel distributed processing, Vol. 2, Cambridge, MA: MIT (...) Press; MacWhinney & Leinbach, 1991, Cognition, 40, 121-157). 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 that was used in MacWhinney and Leinbach's simulation with more realistic training and testing procedures. The SPA outperformed the connectionist models by a wide margin in the accuracy of learning, and successful inductive generalizations to unseen verbs. Our SPA has very natural and psychologically realistic explanations to many psychological effects such as U-shaped learning curve, and is much closer to human subjects in predicting past tense of the pseudo-verbs. In contrast to ANNs, whose internal representations are entirely opaque, the SPA can represent the acquired knowledge in the form of production rules that allow for further higher-level processing and integration, resulting in linguistically realistic associative templates for irregular verbs and production rules for regular verbs. In the light of these findings, we conclude that eliminative connectionists' vision of cognition as simple pattern association and pattern recognition without symbolic representation is inadequate. Pattern association as such does not imply rule-less or cue-based models of language acquisition or of human learning in general. (shrink)

This article addresses the ability of Parallel Distributed Processing (PDP) networks to generate stagewise cognitive development in accordance with Piaget's theory of cognitive epigenesis. We carried out a replication study of the simulation experiments by McClelland (1989) and McClelland and Jenkins (1991) in which a PDP network learns to solve balance scale problems. In objective tests motivated from catastrophe theory, a mathematical theory of transitions in epigenetical systems, no evidence for stage transitions in network performance was found. (...) It is concluded that PDP networks lack the ability to recover the positive outcomes of analogous catastrophe analyses of real cognitive developmental data. In an attempt to further characterize the learning behaviour of PDP networks, we carried out a second simulation study using the discrimination‐shift paradigm. The results thus obtained indicate that PDP learning is compatible with the learning of stimulus‐response relationships, not with the acquisition of mediating rules such as conceived in (neo‐)Piagetian theory. In closing, we speculate about the feasibility of simulating stagewise development with alternative network architectures. (shrink)

A fundamental claim associated with parallel distributed processing theories of cognition is that knowledge is coded in a distributed manner in mind and brain. This approach rejects the claim that knowledge is coded in a localist fashion, with words, objects, and simple concepts, that is, coded with their own dedicated representations. One of the putative advantages of this approach is that the theories are biologically plausible. Indeed, advocates of the PDP approach often highlight the close parallels (...) between distributed representations learned in connectionist models and neural coding in brain and often dismiss localist theories as biologically implausible. The author reviews a range a data that strongly challenge this claim and shows that localist models provide a better account of single-cell recording studies. The author also contrast local and alternative distributed coding schemes and argues that common rejection of grandmother cell theories in neuroscience is due to a misunderstanding about how localist models behave. The author concludes that the localist representations embedded in theories of perception and cognition are consistent with neuroscience; biology only calls into question the distributed representations often learned in PDP models. (shrink)

Are morphological patterns learned in the form of rules? Some models deny this, attributing all morphology to analogical mechanisms. The dual mechanism model (Pinker, S., & Prince, A. (1998). On language and connectionism: analysis of a parallel distributed processing model of language acquisition. Cognition, 28, 73-193) posits that speakers do internalize rules, but that these rules are few and cover only regular processes; the remaining patterns are attributed to analogy. This article advocates a third approach, which uses (...) multiple stochastic rules and no analogy. We propose a model that employs inductive learning to discover multiple rules, and assigns them confidence scores based on their performance in the lexicon. Our model is supported over the two alternatives by new "wug test" data on English past tenses, which show that participant ratings of novel pasts depend on the phonological shape of the stem, both for irregulars and, surprisingly, also for regulars. The latter observation cannot be explained under the dual mechanism approach, which derives all regulars with a single rule. To evaluate the alternative hypothesis that all morphology is analogical, we implemented a purely analogical model, which evaluates novel pasts based solely on their similarity to existing verbs. Tested against experimental data, this analogical model also failed in key respects: it could not locate patterns that require abstract structural characterizations, and it favored implausible responses based on single, highly similar exemplars. We conclude that speakers extend morphological patterns based on abstract structural properties, of a kind appropriately described with rules. (shrink)

The introduction of connectionist or parallel distributed processing (PDP) systems to model cognitive functions has raised the question of the possible relations between these models and traditional information processing models which employ rules to manipulate representations. After presenting a brief account of PDP models and two ways in which they are commonly interpreted by those seeking to use them to explain cognitive functions, I present two ways one might relate these models to traditional information processing (...) models and so not totally repudiate the tradition of modelling cognition through systems of rules and representations. The proposal that seems most promising is that PDP-type structures might provide the underlying framework in which a rule and representation model might be implemented. To show how one might pursue such a strategy, I discuss recent research by Barsalou on the instability of concepts and show how that might be accounted for in a system whose microstructure had a PDP architecture. I also outline how adopting a multi-leveled view of the mind, where on one level the mind employed a PDP-type system and at another level constituted a rule processing system, would allow researchers to relocate some problems which seemed difficult to explain at one level, such as the capacity for concept learning, to another level where it could be handled in a straightforward manner. (shrink)

This article addresses the ability of Parallel Distributed Processing (PDP) networks to generate stagewise cognitive development in accordance with Piaget's theory of cognitive epigenesis. We carried out a replication study of the simulation experiments by McClelland (1989) and McClelland and Jenkins (1991) in which a PDP network learns to solve balance scale problems. In objective tests motivated from catastrophe theory, a mathematical theory of transitions in epigenetical systems, no evidence for stage transitions in network performance was found. (...) It is concluded that PDP networks lack the ability to recover the positive outcomes of analogous catastrophe analyses of real cognitive developmental data. In an attempt to further characterize the learning behaviour of PDP networks, we carried out a second simulation study using the discrimination‐shift paradigm. The results thus obtained indicate that PDP learning is compatible with the learning of stimulus‐response relationships, not with the acquisition of mediating rules such as conceived in (neo‐)Piagetian theory. In closing, we speculate about the feasibility of simulating stagewise development with alternative network architectures. (shrink)