The cerebral cortex is a rich and diverse structure that is the basis of intelligent behavior. One of the deepest mysteries of the function of cortex is that neural processing times are only about one hundred times as fast as the fastest response times for complex behavior. At the very least, this would seem to indicate that the cortex does massive amounts of parallel computation.This paper explores the hypothesis that an important part of the cortex can be modeled as a (...) connectionist computer that is especially suited for parallel problem solving. The connectionist computer uses a special representation, termed value unit encoding, that represents small subsets of parameters in a way that allows parallel access to many different parameter values. This computer can be thought of as computing hierarchies of sensorimotor invariants. The neural substrate can be interpreted as a commitment to data structures and algorithms that compute invariants fast enough to explain the behavioral response times. A detailed consideration of this model has several implications for the underlying anatomy and physiology. (shrink)

The cerebral cortex is a rich and diverse structure that is the basis of intelligent behavior. One of the deepest mysteries of the function of cortex is that neural processing times are only about one hundred times as fast as the fastest response times for complex behavior. At the very least, this would seem to indicate that the cortex does massive amounts of parallel computation.This paper explores the hypothesis that an important part of the cortex can be modeled as a (...) connectionist computer that is especially suited for parallel problem solving. The connectionist computer uses a special representation, termed value unit encoding, that represents small subsets of parameters in a way that allows parallel access to many different parameter values. This computer can be thought of as computing hierarchies of sensorimotor invariants. The neural substrate can be interpreted as a commitment to data structures and algorithms that compute invariants fast enough to explain the behavioral response times. A detailed consideration of this model has several implications for the underlying anatomy and physiology. (shrink)

Intersemiotic translation (IT) was defined by Roman Jakobson (The Translation Studies Reader, Routledge, London, p. 114, 2000) as “transmutation of signs”—“an interpretation of verbal signs by means of signs of nonverbal sign systems.” Despite its theoretical relevance, and in spite of the frequency in which it is practiced, the phenomenon remains virtually unexplored in terms of conceptual modeling, especially from a semiotic perspective. Our approach is based on two premises: (i) IT is fundamentally a semiotic operation process (semiosis) and (ii) (...) IT is a deeply iconic-dependent process. We exemplify our approach by means of literature to dance IT and we explore some implications for the development of a general model of IT. (shrink)

Throughout what is now the more than 50-year history of the computer many theories have been advanced regarding the contribution this machine would make to changes both in the structure of society and in ways of thinking. Like other theories regarding the future, these should also be taken with a pinch of salt. The history of the development of computer technology contains many predictions which have failed to come true and many applications that have not been foreseen. While we must (...) reserve judgment as to the question of the impact on the structure of society and human thought, there is no reason to wait for history when it comes to the question: what are the properties that could give the computer such far-reaching importance? The present book is intended as an answer to this question. The fact that this is a theoretical analysis is due to the nature of the subject. No other possibilities are available because such a description of the properties of the computer must be valid for any kind of application. An additional demand is that the description should be capable of providing an account of the properties which permit and limit these possible applications, just as it must make it possible to characterize a computer as distinct from a) other machines whether clocks, steam engines, thermostats, or mechanical and automatic calculating machines, b) other symbolic media whether printed, mechanical, or electronic and c) other symbolic languages whether ordinary languages, spoken or written or formal languages. This triple limitation, however, (with regard to other machines, symbolic media and symbolic languages) raises a theoretical question as it implies a meeting between concepts of mechanical-deterministic systems, which stem from mathematical physics, and concepts of symbolic systems which stem from the description of symbolic activities common to the humanities. The relationship between science and the humanities has traditionally been seen from a dualistic perspective, as a relationship between two clearly separate subject areas, each studied on its own set of premises and using its own methods. In the present case, however, this perspective cannot be maintained since there is both a common subject area and a new - and specific - kind of interaction between physical and symbolic processes. (shrink)

The format of mental representation is the way information is organized in the mind. The discussion surrounding the format of representation addresses the problem of what representational primitives are and the rules of information processing. In philosophy, the discussion is dominated by the distinction between analog and digital representational systems. It is thought that this distinction can bring us closer to an understanding of the nature of perceptual and discursive representations. I argue that the analog-digital distinction cannot meet that expectation. (...) The analog-digital distinction is neither sufficient nor necessary to explain the distinction between perceptual and discursive representations. I propose an alternative interpretation of the concept of representational format which provides us a better understanding of the difference between iconic and discursive representations. I explain the differences between formats of representations in terms of differences in information processing. I demonstrate, how this alternative interpretation of the concept of the representational format can explain the constraints put on the contents of representational systems. (shrink)

Characterizing different kinds of representation is of fundamental importance to cognitive science, and one traditional way of doing so is in terms of the analog–digital distinction. Indeed the distinction is often appealed to in ways both narrow and broad. In this paper I argue that the analog–digital distinction does not apply to representational schemes but only to representational systems, where a representational system is constituted by a representational scheme and its user, and that whether a representational system is analog or (...) non-analog depends on facts about that user. This aspect of the distinction has gone unnoticed, and I argue that the failure to notice it can be an impediment to scientific progress. (shrink)

I argue that analogue mental representations possess a canonical decomposition into privileged constituents from which they compose. I motivate this suggestion, and rebut arguments to the contrary, through reflection on the approximate number system, whose representations are widely expected to have an analogue format. I then argue that arguments for the compositionality and constituent structure of these analogue representations generalize to other analogue mental representations posited in the human mind, such as those in early vision and visual imagery.

It is often claimed that pre-attentive vision has an ‘iconic’ format. This is seen to explain pre-attentive vision's characteristically high processing capacity and to make sense of an overlap in the mechanisms of early vision and mental imagery. But what does the iconicity of pre-attentive vision amount to? This paper considers two prominent ways of characterising pre-attentive visual icons and argues that neither is adequate: one approach renders the claim ‘pre-attentive vision is iconic’ empirically false while the other obscures its (...) ability to do the explanatory work, which motivates positing pre-attentive visual icons in the first place. With this noted, I introduce the notion of an ‘Analogue Map’ and argue that it provides a superior characterisation of pre-attentive vision's iconicity. I then argue that this forces a reassessment of debates which have traditionally presupposed the iconicity of pre-attentive vision, emphasising ramifications for the viability of a format-based perception-thought border. (shrink)

In this paper, I argue for three claims. The first is that the difference between analog and digital representation lies in the format and not the medium of representation. The second is that whether a given system is analog or digital will sometimes depend on facts about the user of that system. The third is that the first two claims are implicit in Haugeland's (1998) account of the distinction.

Amit's Hebbian model risks being overexplanatory, since it does not depend on specific physiological modelling of cortical ANNs, but concentrates on those phenomena which are modelled by a large class of ANNs. While offering a strong demonstration of the presence of Hebb's “cell assemblies,” it does not offer an equal account of Hebb's “phase sequence” concept.

Noam Chomsky and other linguists and psychologists have suggested that human linguistic behavior is somehow governed by a mental representation of a transformational grammar. Challenges to this controversial claim have often been met by invoking an explicitly computational perspective: It makes perfect sense to suppose that a grammar could be represented in the memory of a computational device and that this grammar could govern the device's use of a language. This paper urges, however, that the claim that humans are such (...) a device is unsupported and that it seems unlikely that linguists and psychologists really want to claim any such thing. Evidence for the linguists' original claim is drawn from three main sources: the explanation of language comprehension and other linguistic abilities; evidence for formal properties of the rules of the grammar; and the explanation of language acquisition. It is argued in this paper that none of these sources provides support for the view that the grammar governs language processing in something like the way a program governs the operation of a programmed machine. The computational approach, on the contrary, suggests ways in which linguistic abilities can be explained without the attribution of any explicit representation of rules governing linguistic behavior. (shrink)

Patricia Smith Churchland's Neurophilosophy argues that a mind is the same thing as the complex patterns of neural activity in a human brain and, furthermore, that we will be able to find out interesting things about the mind by studying the brain. I basically agree with this stance and my comments are divided into four sections. First, comparisons between human and non?human primate brains are discussed in the context, roughly, of where one should locate higher functions. Second, I examine Churchland's (...) views on reduction and levels of organization, which I find mostly congenial. Third, a key point of disagreement about the relationship and importance of language to specifically human cognition is taken up. I like Churchland's critique of certain sentential paradigms, but I try to show using an analogy with cellular coding systems why we need to get a better theory of ?sentences?. Finally, I discuss how the models introduced in the last chapter might be extended to make better contact with neurobiology and language. (shrink)

Using four examples of models and computer simulations from the history of psychology, I discuss some of the methodological aspects involved in their construction and use, and I illustrate how the existence of a model can demonstrate the viability of a hypothesis that had previously been deemed impossible on a priori grounds. This shows a new way in which scientists can learn from models that extends the analysis of Morgan (1999), who has identified the construction and manipulation of models as (...) those phases in which learning from models takes place. (shrink)

Cortical reverberations may induce synaptic changes that underlie developmental plasticity as well as long-term memory. They may be especially important for the consolidation of synaptic changes. Reverberations in cortical networks should have particular significance during development, when large numbers of new representations are formed. This includes the formation of representations across different sensory modalities.

A distinction should be made between the formation of stimulus-driven associations and cognitive concepts. To test the learning mode of a neural network, we propose a simple and classic input-output test: the discrimination shift task. Feed-forward PDP models appear to form stimulus-driven associations. A Hopfield network should be extended to apply the test.

Amit defines cell assemblies aslocal cortical neuron populationswith strong internal connections. However, Hebb himself proposed that cell assemblies are distributed over different cortical areas (nonlocal ortranscortical assemblies). We review evidence from cognitive neuroscience and neuropsychology supporting the assumption that cell assemblies are transcortical.

The theory of existential graphs, which Peirce ultimately divided into four quadrants , is a rich method of analysis in the philosophy of logic. Its $$\upbeta $$ β -part boasts a diagrammatic theory of quantification, which by 1902 Peirce had used in the logical analysis of natural-language expressions such as complex donkey-type anaphora, quantificational patterns describing new mathematical concepts, and cognitive information processing. In the $$\upbeta $$ β -quadrant, he came close to inventing independence-friendly logic, the idea of which he (...) found indispensable in fulfilling the tasks –. (shrink)

Defending or attacking either functionalism or computationalism requires clarity on what they amount to and what evidence counts for or against them. My goal here is not to evaluate their plausibility. My goal is to formulate them and their relationship clearly enough that we can determine which type of evidence is relevant to them. I aim to dispel some sources of confusion that surround functionalism and computationalism, recruit recent philosophical work on mechanisms and computation to shed light on them, and (...) clarify how functionalism and computationalism may or may not legitimately come together. (shrink)

We begin by distinguishing computationalism from a number of other theses that are sometimes conflated with it. We also distinguish between several important kinds of computation: computation in a generic sense, digital computation, and analog computation. Then, we defend a weak version of computationalism—neural processes are computations in the generic sense. After that, we reject on empirical grounds the common assimilation of neural computation to either analog or digital computation, concluding that neural computation is sui generis. Analog computation requires continuous (...) signals; digital computation requires strings of digits. But current neuroscientific evidence indicates that typical neural signals, such as spike trains, are graded like continuous signals but are constituted by discrete functional elements (spikes); thus, typical neural signals are neither continuous signals nor strings of digits. It follows that neural computation is sui generis. Finally, we highlight three important consequences of a proper understanding of neural computation for the theory of cognition. First, understanding neural computation requires a specially designed mathematical theory (or theories) rather than the mathematical theories of analog or digital computation. Second, several popular views about neural computation turn out to be incorrect. Third, computational theories of cognition that rely on non-neural notions of computation ought to be replaced or reinterpreted in terms of neural computation. (shrink)

Some philosophers have conflated functionalism and computationalism. I reconstruct how this came about and uncover two assumptions that made the conflation possible. They are the assumptions that (i) psychological functional analyses are computational descriptions and (ii) everything may be described as performing computations. I argue that, if we want to improve our understanding of both the metaphysics of mental states and the functional relations between them, we should reject these assumptions. # 2004 Elsevier Ltd. All rights reserved.

Since the cognitive revolution, it has become commonplace that cognition involves both computation and information processing. Is this one claim or two? Is computation the same as information processing? The two terms are often used interchangeably, but this usage masks important differences. In this paper, we distinguish information processing from computation and examine some of their mutual relations, shedding light on the role each can play in a theory of cognition. We recommend that theorists of cognition be explicit and careful (...) in choosing notions of computation and information and connecting them together.Keywords: Computation; Information processing; Computationalism; Computational theory of mind; Cognitivism. (shrink)

According to pancomputationalism, everything is a computing system. In this paper, I distinguish between different varieties of pancomputationalism. I find that although some varieties are more plausible than others, only the strongest variety is relevant to the philosophy of mind, but only the most trivial varieties are true. As a side effect of this exercise, I offer a clarified distinction between computational modelling and computational explanation.<br><br>.

I offer an explication of the notion of computer, grounded in the practices of computability theorists and computer scientists. I begin by explaining what distinguishes computers from calculators. Then, I offer a systematic taxonomy of kinds of computer, including hard-wired versus programmable, general-purpose versus special-purpose, analog versus digital, and serial versus parallel, giving explicit criteria for each kind. My account is mechanistic: which class a system belongs in, and which functions are computable by which system, depends on the system's mechanistic (...) properties. Finally, I briefly illustrate how my account sheds light on some issues in the history and philosophy of computing as well as the philosophy of mind. (shrink)

Amit's “Attractor Neural Network” perspective on cognition raises difficult technical problems already met by prior dynamical models. This commentary sketches briefly some of them concerning the internal topological structure of attractors, the constituency problem, the possibility of activating simultaneously several attractors, and the different kinds of dynamical structures one can use to model brain activity: point attractors, strange attractors, synchronized arrays of oscillators, synfire chains, and so forth.

Starting in the 1950s, computer programs for simulating cognitive processes and intelligent behaviour were the hallmark of Good Old-Fashioned Artificial Intelligence and ‘cognitivist’ cognitive science. This article examines a somewhat neglected case of simulation pursued by one of the founding fathers of simulation methodology, Herbert A. Simon. In the 1970s and 1980s, Simon had repeated contacts with Marxist countries and scientists, in the context of which he advanced the idea that cognitivism could be used as a framework for simulating dialectical (...) materialism. Simon's idea was, in particular, to represent dialectical processes through a ‘symbolic’ version of dialectical logic. This article explores the context of Simon's interaction with Marxist countries—China and the USSR—and also assesses the outcome of the simulation. The difficulty with simulating distinctive features of dialectical materialism is read in light of the underlying assumptions of cognitivism and, ultimately, in light of the attempt to tame a rival world view. (shrink)

Recent studies indicate that the processing of an unexpected word is costly when the initial, disconfirmed prediction was strong. This penalty was suggested to stem from commitment to the strongly predicted word, requiring its inhibition when disconfirmed. Additional studies show that comprehenders rationally adapt their predictions in different situations. In the current study, we hypothesized that since the disconfirmation of strong predictions incurs costs, it would also trigger adaptation mechanisms influencing the processing of subsequent strong predictions. In two experiments, participants (...) made speeded congruency judgments on two-word phrases in which the first word was either highly constraining or not. We manipulated the proportion of disconfirmed predictions in highly constraining contexts between participants. The results provide additional evidence of the costs associated with the disconfirmation of strong predictions. Moreover, they show a reduction in these costs when participants experience a high proportion of disconfirmed strong predictions throughout the experiment, indicating that participants adjust the strength of their predictions when strong prediction is discouraged. We formulate a Bayesian adaptation model whereby prediction failure cost is weighted by the participant’s belief about the likelihood of encountering the expected word, and show that it accounts for the trial-by-trial data. (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)

The Miyashita experiments are very interesting and the results should be examined from a viewpoint of attractor dynamics. Amit's target article shows a path toward realistic modeling by artificial neural networks (ANN), but it is not necessarily the only one. I introduce another model that can explain a substantial part of the empirical observations and makes an interesting prediction. This model consists of such units that have nonmonotonic input-output characteristics with local inhibition neurons.