In a seminal 1977 article, Rumelhart argued that perception required the simultaneous use of multiple sources of information, allowing perceivers to optimally interpret sensory information at many levels of representation in real time as information arrives. Building on Rumelhart's arguments, we present the Interactive Activation hypothesis—the idea that the mechanism used in perception and comprehension to achieve these feats exploits an interactive activation process implemented through the bidirectional propagation of activation among simple processing units. We then examine the interactive activation (...) model of letter and word perception and the TRACE model of speech perception, as early attempts to explore this hypothesis, and review the experimental evidence relevant to their assumptions and predictions. We consider how well these models address the computational challenge posed by the problem of perception, and we consider how consistent they are with evidence from behavioral experiments. We examine empirical and theoretical controversies surrounding the idea of interactive processing, including a controversy that swirls around the relationship between interactive computation and optimal Bayesian inference. Some of the implementation details of early versions of interactive activation models caused deviation from optimality and from aspects of human performance data. More recent versions of these models, however, overcome these deficiencies. Among these is a model called the multinomial interactive activation model, which explicitly links interactive activation and Bayesian computations. We also review evidence from neurophysiological and neuroimaging studies supporting the view that interactive processing is a characteristic of the perceptual processing machinery in the brain. In sum, we argue that a computational analysis, as well as behavioral and neuroscience evidence, all support the Interactive Activation hypothesis. The evidence suggests that contemporary versions of models based on the idea of interactive activation continue to provide a basis for efforts to achieve a fuller understanding of the process of perception. (shrink)

I consider the role of cognitive modeling in cognitive science. Modeling, and the computers that enable it, are central to the field, but the role of modeling is often misunderstood. Models are not intended to capture fully the processes they attempt to elucidate. Rather, they are explorations of ideas about the nature of cognitive processes. In these explorations, simplification is essential—through simplification, the implications of the central ideas become more transparent. This is not to say that simplification has no downsides; (...) it does, and these are discussed. I then consider several contemporary frameworks for cognitive modeling, stressing the idea that each framework is useful in its own particular ways. Increases in computer power (by a factor of about 4 million) since 1958 have enabled new modeling paradigms to emerge, but these also depend on new ways of thinking. Will new paradigms emerge again with the next 1,000‐fold increase? (shrink)

The study of human intelligence was once dominated by symbolic approaches, but over the last 30 years an alternative approach has arisen. Symbols and processes that operate on them are often seen today as approximate characterizations of the emergent consequences of sub- or nonsymbolic processes, and a wide range of constructs in cognitive science can be understood as emergents. These include representational constructs (units, structures, rules), architectural constructs (central executive, declarative memory), and developmental processes and outcomes (stages, sensitive periods, neurocognitive (...) modules, developmental disorders). The greatest achievements of human cognition may be largely emergent phenomena. It remains a challenge for the future to learn more about how these greatest achievements arise and to emulate them in artificial systems. (shrink)

In this interview, James L. McClelland responds to questions regarding connectionist models of cognition, a theory inspired by information processing in the brain. McClelland explains the distinction between symbolic and non-symbolic processing for a better understanding of mental processes. He argues that connectionist models can perform the computations which we know the brain can perform. In addition, he responds to several general questions on the perspectives of computational models of cognition.

We share with Anderson & Lebiere (A&L) (and with Newell before them) the goal of developing a domain-general framework for modeling cognition, and we take seriously the issue of evaluation criteria. We advocate a more focused approach than the one reflected in Newell's criteria, based on analysis of failures as well as successes of models brought into close contact with experimental data. A&L attribute the shortcomings of our parallel-distributed processing framework to a failure to acknowledge a symbolic level of thought. (...) Our framework does acknowledge a symbolic level, contrary to their claim. What we deny is that the symbolic level is the level at which the principles of cognitive processing should be formulated. Models cast at a symbolic level are sometimes useful as high-level approximations of the underlying mechanisms of thought. The adequacy of this approximation will continue to increase as symbolic modelers continue to incorporate principles of parallel distributed processing. (shrink)

Different developmental anomalies produce contrasting deficits in a single, integrated system. In a network that inflects regular and exception verbs correctly, a disproportionate deficit with exceptions occurs if connections are deleted, whereas a disproportionate deficit with regulars occurs when an auditory deficit impairs perception of the regular inflection. In general, contrasting deficits do not license the inference of underlying modularity.

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)

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 effects of lexical context on phonological processing are pervasive and there have been indications that such effects may be modulated by attention. However, attentional modulation in speech processing is neither well documented nor well understood. Experiment 1 demonstrated attentional modulation of lexical facilitation of speech sound recognition when task and critical stimuli were identical across attention conditions. We propose modulation of lexical activation as a neurophysiologically plausible computational mechanism that can account for this type of modulation. Contrary to the (...) claims of critics, this mechanism can account for attentional modulation without violating the principle of interactive processing. Simulations of the interactive TRACE model extended to include two different ways of modulating lexical activation showed that each can account for attentional modulation of lexical feedback effects. Experiment 2 tested conflicting predictions from the two implementations and provided evidence that is consistent with bias input as the mechanism of attentional control of lexical activation. (shrink)

In this article we present symmetric diffusion networks, a family of networks that instantiate the principles of continuous, stochastic, adaptive and interactive propagation of information. Using methods of Markovion diffusion theory, we formalize the activation dynamics of these networks and then show that they can be trained to reproduce entire multivariate probability distributions on their outputs using the contrastive Hebbion learning rule (CHL). We show that CHL performs gradient descent on an error function that captures differences between desired and obtained (...) continuous multivariate probability distributions. This allows the learning algorithm to go beyond expected values of output units and to approximate complete probability distributions on continuous multivariate activation spaces. We argue that learning continuous distributions is an important task underlying a variety of real‐life situations that were beyond the scope of previous connectionist networks. Deterministic networks, like back propagation, cannot learn this task because they are limited to learning average values of independent output units. Previous stochastic connectionist networks could learn probability distributions but they were limited to discrete variables. Simulations show that symmetric diffusion networks can be trained with the CHL rule to approximate discrete and continuous probability distributions of various types. (shrink)

Page's proposal to stipulate representations in which individual units correspond to meaningful entities is too unconstrained to support effective theorizing. An approach combining general computational principles with domain-specific assumptions, in which learning is used to discover representations that are effective in solving tasks, provides more insight into why cognitive and neural systems are organized the way they are.

The commentaries reflect three core themes that pertain not just to our theory, but to the enterprise of connectionist modeling more generally. The first concerns the relationship between a cognitive theory and an implemented computer model. Specifically, how does one determine, when a model departs from the theory it exemplifies, whether the departure is a useful simplification or a critical flaw? We argue that the answer to this question depends partially upon the model's intended function, and we suggest that connectionist (...) models have important functions beyond the commonly accepted goals of fitting data and making predictions. The second theme concerns perceived in-principle limitations of the connectionist approach to cognition, and the specific concerns these perceived limitations raise for our theory. We argue that the approach is not in fact limited in the ways our critics suggest. One common misconception, that connectionist models cannot address abstract or relational structure, is corrected through new simulations showing directly that such structure can be captured. The third theme concerns the relationship between parallel distributed processing (PDP) models and structured probabilistic approaches. In this case we argue that there the difference between the approaches is not merely one of levels. Our PDP approach differs from structured statistical approaches at all of Marr's levels, including the characterization of the goals of cognitive computations, and of the representations and algorithms used. (shrink)

Mitchell et al. describe many fascinating studies, and in the process, propose what they consider to be a unified framework for human learning in which effortful, controlled learning results in propositional knowledge. However, it is unclear how any of their findings privilege a propositional account, and we remain concerned that embedding all knowledge in propositional representations obscures the tight interdependence between learning from experiences and the use of the results of learning as a basis for action.