The paper is a commentary on the target article by Christine A. Skarda & Walter J. Freeman, “How brains make chaos in order to make sense of the world”, in the same issue of the journal, pp.161–195. -/- I confine my comments largely to some philosophical claims that Skarda & Freeman make and to the relationship of their model to connectionism. Some of the comments hinge on what symbols are and how they might sit in neural systems.

The majority of commentators agree on one thing: Our network approach might be the prime candidate for offering a new perspective on the origins of mental disorders. In our response, we elaborate on refinements (e.g., cognitive and genetic levels) and extensions (e.g., to Axis II disorders) of the network model, as well as discuss ways to test its validity.

A set of hypotheses is formulated for a connectionist approach to cognitive modeling. These hypotheses are shown to be incompatible with the hypotheses underlying traditional cognitive models. The connectionist models considered are massively parallel numerical computational systems that are a kind of continuous dynamical system. The numerical variables in the system correspond semantically to fine-grained features below the level of the concepts consciously used to describe the task domain. The level of analysis is intermediate between those of symbolic cognitive models (...) and neural models. The explanations of behavior provided are like those traditional in the physical sciences, unlike the explanations provided by symbolic models. (shrink)

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)

This paper is a commentary on the target article by Dana H. Ballard, “Cortical connections and parallel processing: Structure and function”, in the same issue of the journal, pp. 67–120. -/- I raise some issues about the connectionist or neural-network implementation of information and information processing. Issues include the sharing of information by different parts of a connectionist/neural network, the copying of complex information from one place to another in a network, the possibility of connection weights not being synaptic weights, (...) and the possibility of fast communication within single neurons as complex systems in their own right. (shrink)

This paper presents a general computational treatment of how mammals are able to deal with visual objects and environments. The model tries to cover the entire range from behavior and phenomenological experience to detailed neural encodings in crude but computationally plausible reductive steps. The problems addressed include perceptual constancies, eye movements and the stable visual world, object descriptions, perceptual generalizations, and the representation of extrapersonal space.The entire development is based on an action-oriented notion of perception. The observer is assumed to (...) be continuously sampling the ambient light for information of current value. The central problem of vision is taken to be categorizing and locating objects in the environment. The critical step in this process is the linking of visual information to symbolic object descriptions; this is calledindexing, from the analogy of identifying a book from index terms. The system must also identifysituationsand use this knowledge to guide movement and other actions in the environment. The treatment focuses on the different representations of information used in the visual system.The four representational frames capture information in the following forms: retinotopic, head-based, symbolic, and allocentric. The functional roles of the four frames, the communication among them, and their suggested neurophysiological realization constitute the core of the paper. The model is perforce crude, but appears to be consistent with all relevant findings. (shrink)

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)

A neurally plausible model of how the process of visually perceiving a letter in the context of a word is learned, and how such processing occurs in adults is proposed. The model consists of a collection of abstract letter feature detector neurons and their interconnections. The model also includes a learning rule that specifies how these interconnections evolve with experience. The interconnections between neurons can be interpreted as representing the spatially redundant, sequentially redundant, and transgraphemic information in letter string displays. (...) Anderson, Silverstein, Ritz, and Jones's (1977) “Brain‐State‐in‐a‐Box” (BSB) neural mechanism is then used to implement the proposed model. The resulting system makes explicit qualitative predictions using both letter recognition accuracy and reaction time as dependent measures. In particular, the model offers an integrated explanation of some experiments involving manipulations of orthographic regularity, masking, case alternations, and experience with words. The similarities and differences between the model and models proposed by Adams (1979) and McClelland and Rumelhart (1981) are also discussed. (shrink)

Although the connectionist approach has lead to elegant solutions to a number of problems in cognitive science and artificial intelligence, its suitability for dealing with problems in knowledge representation and inference has often been questioned. This paper partly answers this criticism by demonstrating that effective solutions to certain problems in knowledge representation and limited inference can be found by adopting a connectionist approach. The paper presents a connectionist realization of semantic networks, that is, it describes how knowledge about concepts, their (...) properties, and the hierarchical relationship between them may be encoded as an interpreter‐free massively parallel network of simple processing elements that can solve an interesting class of inheritance and recognition problems extremely fast—in time proportional to the depth of the conceptual hierarchy. The connectionist realization is based on an evidential formulation that leads to principled solutions to the problems of exceptions and conflicting multiple inheritance situations during inheritance, and the best‐match or partial‐match computation during recognition. The paper also identifies constraints that must be satisfied by the conceptual structure in order to arrive at an efficient parallel realization. (shrink)

The use of the computer metaphor has led to the proposal of mind architecture (Pylyshyn 1984; Newell 1990) as a model of the organization of the mind. The dualist computational model, however, has, since the earliest days of psychological functionalism, required that the concepts mind architecture and brain architecture be remote from each other. The development of both connectionism and neurocomputational science, has sought to dispense with this dualism and provide general models of consciousness – a uniform cognitive architecture –, (...) which is in general reductionist, but which retains the computer metaphor. This paper examines, in the first place, the concepts of mind architecture and brain architecture, in order to evaluate the syntheses which have recently been offered. It then moves on to show how modifications which have been made to classical functionalist mind architectures, with the aim of making them compatible with brain architectures, are unable to resolve some of the most serious problems of functionalism. Some suggestions are given as to why it is not possible to relate mind structures and brain structures by using neurocomputational approaches, and finally the question is raised of the validity of reductionism in a theory which sets out to unite mind and brain architectures. (shrink)

This article responds to two unresolved and crucial problems of cognitive science: (1) What is actually accomplished by functions of the nervous system that we ordinarily describe in the intentional idiom? and (2) What makes the information processing involved in these functions semantic? It is argued that, contrary to the assumptions of many cognitive theorists, the computational approach does not provide coherent answers to these problems, and that a more promising start would be to fall back on mathematical communication theory (...) and, with the help of evolutionary biology and neurophysiology, to attempt a characterization of the adaptive processes involved in visual perception. Visual representations are explained as patterns of cortical activity that are enabled to focus on objects in the changing visual environment by constantly adjusting to maintain levels of mutual information between pattern and object that are adequate for continuing perceptual control. In these terms, the answer proposed to (1) is that the intentional functions of vision are those involved in the establishment and maintenance of such representations, and to (2) that semantic features are added to the information processes of vision with the focus on objects that these representations accomplish. The article concludes with proposals for extending this account of intentionality to the higher domains of conceptualization and reason, and with speculation about how semantic information-processing might be achieved in mechanical systems. (shrink)