This page has been accessed  times.

ABSTRACT: A brain-based alternative to reinforcement learning integrates artificial neural networks (ANN) and knowledge based (KB) systems into one unit or agent for goal-oriented problem solving. The agent works under closed-loop control. Its sensory system provides the input to the controller and the controller's output results in the agent's immediate motor actions. The controller can have both inherited and learned ANN and KB systems. The agent has and develops ANN cues to the environment for dimensionality reduction (data compression) in order to ease the problem of combinatorial explosion. A dynamic conceptual model (DCM) builds cue-models of the phenomena in the world, designs dynamic action sets and makes them compete at a spreading-activation neuronal stage to come to a decision. DCM can create concepts or subgoals. For the production of subgoals two things are important: (i) DCM has an inherited goal system and (ii) subgoals are created by the agent's experiences encountered during the interaction with the environment. Concepts allow the use of rule-based systems for control. The agent's experiences transform during learning into a rule system. Concept generation reduces memory and time requirements. It also improves the system's ability to handle unknown situations. We examine the capabilities of a simple robotic-like object in a two-dimensional conditionally probabilistic space.
 
 

ABSTRACT: The connectionist decade is drawing to a close. Notwithstanding some significant technological successes (e.g., the IEEE describes Neural Networks as "standard technology" in 1994) connectionist models have had limited impact in the cognitive sciences and neuroscience. Difficulties have come in many forms: (1) confusing technology (hacking) with science (2) dissociation from data (e.g., ignoring behavioral data) (3) taking the brain too seriously (4) temporal control problems As connectionist approaches migrate towards normal science, these and other difficulties will focus our attention on providing a basic framework for the application of connectionist tools to the study of brain function and its organization. I will discuss these issues in the context of specific variations in connectionist networks that I have worked on and their applications in defining the appropriate level of analysis, controlling technology and finding behavioral phenonmena that can reveal important properties of brain function.
 
 

ABSTRACT: The theory of minds as "semiotic" (sign-using) systems offers a promising approach to the problem of relating consciousness and cognition, mind and body. The approach is highly compatible with connectionist (or "neural network") models of the brain and has many advantages over alternative theories.
 
 

ABSTRACT: Darwin's theory of evolution has been used to argue that because we humans are fitted by natural selection for survival and reproduction, and associated understanding, it is unlikely that we are made to understand remoter matters, such as consciousness, the origin of the universe, of life, etc. The difficulty is logical, because it is the quality not the quantity of human intelligence that is deficient, some varieties of sceptical naturalism have suggested. How might this scepticism be resisted? By a better understanding of Darwinism. The "over-capacity" of human intelligence is both compatible with Darwinism and explicable by it. Does the sceptic's worry really make sense? Scientific problems are collective and projective concerns. Natural selection could only ever deal with problems "set by the world", with the "surface" and "immediate" problems of individual units of selection. What would it be like for other rational beings to have scientific understanding from which we were logically excluded? This is by no means clear. Davidson argued against radically incommensurable conceptual schemes. Where would Popper, a father of naturalised epistemology, have stood in this dispute? His resolute anti-determinism and his stress on creativity in science locate him in the anti-sceptical camp. Naturalism is a new feint in an old joust. Well before Darwin, philosophers sought to set a priori limits to the scope of the human mind. McGinn, who utilized sceptical naturalism to claim that consciousness might be forever mysterious, now asks - Why is it impossible to do philosophy? (Because we are not biologically fitted for it!) I dispute McGinn's contention, but, of course, if he is right, this paper must stand as its own refutation.
 
 

ABSTRACT: Studies of the breakdown of semantic memory in adults have suggested that object concepts are lost when knowledge of the defining features is lost. In particular, when perceptual features are damaged, knowledge of living things is impaired, and when functional features are damaged, knowledge of nonliving things is impaired (Warington and Shallice, 1984; Farah and McClelland, 1991). This talk will discuss an alternative hypothesis which proposes that the loss of conceptual knowledge depends upon how familiar the concepts is in the subject's experience. It is argued that object concepts are not lost because property knowledge is lost, but rather, access to knowledge of properties depends upon the familiarity of the object concept. What makes an object concept familiar will be discussed in relation to language use and every-day experience with objects.

Warrington, E.K. and Shallice, T. (1984). Category-specific semantic impairment. Brain, 107, 829-854.
Farah, M. and McClelland, J. (1991). A computational model of semantic memory impairment: Modality specificity and emergent category-specificity. Journal of Experimental Psychology: General, 120, 330-357.
 
 
 
 

ABSTRACT: Abnormal development poses interesting challenges to the nature/nurture debate, the relationship between language and cognition, and the extent to which the mind/brain is modular. Individuals with Down syndrome usually show across the-board deficits which many have used to argue for a general, domain-neutral learning mechanism which is impaired in these subjects but operates efficiently in normal individuals. By contrast, the development of individuals with autism, Williams syndrome (WS), hydrocephalus with associated myelomeningocele, dyslexia, and specific language impairment, results in very uneven cognitivo-linguistic profiles in which some domains are relatively spared while others are seriously impaired. These different, uneven profiles challenge any view of development solely in terms of an across-the-board domain- general intelligence, and at first blush seem to point to innately-specified, domain-specific, special-purpose modules. A number of theorists have used the case of WS to argue for an innately-specified, strictly modular view of the human mind/brain in which language is preserved in the face of serious impairments in non-verbal domains. However, in-depth studies of WS demonstrate the existence not only of across- domain dissociations, but also of within- domain dissociations in language and in spatial cognition, suggesting that domain- general and domain-specific mechanisms both play a crucial role in language and cognition. By drawing a distinction between computational, architectural, chronotropic and representational biases for processing different inputs, one can postulate the progressive domain-specific specialization of the normal infant mind/brain on the basis of minimal predispositions as development proceeds. In abnormal development, either the initial biases or the process of specialization is impaired.
 
 

ABSTRACT: Recent habituation studies indicate that 9- and 12-month-old infants (but not 6-month-olds) evaluate the rationality of the goal-directed movement of an agent. When the rationality assumption is upheld, infants can infer the most rational future means of action that the agent is likely to perform in a new situation. These results are related to current approaches to the origins of understanding intentional action: Recent models of Agency (Leslie, Premack, Baron-Cohen, Mandler), which assume the primacy of a naive theory of physics to identify self-propelled objects as the proper domain for intentional explanations, will be contrasted with an alternative approach that grounds naive psychology in the rationality assumption, which is independent of the infant's physical theory. Our recent evidence that the rationality of movement is evaluated even without perceptual cues of biological agency supports the alternative view which places the rationality assumption at the core of the development of naive psychology.
 
 

ABSTRACT: Simple neural networks may implement cognitive functioning. The traditional model framework can be extended by taking into account the electrophysiological details of single cell dynamics. In the other direction, properly defined learning rules can implement associative memory phenomena. The olfactory bulb is a prototype of neural systems that show a set of characteristic dynamic behavior as oscillation and chaos. The need to integrate structural, functional and dynamic approaches is emphasised.
 
 

ABSTRACT: In artificial intelligence systems knowledge is often represented as a set of rules to be interpreted by an expert system. This rule base is garnered from the knowledge of a domain expert, and, for a variety of reasons, may be incomplete, contradictory, or inaccurate. A system that merely uses these rules is then static, and unable to learn new rules or to modify existing rules in the problem domain. On the other hand, Artificial Neural Networks, (ANNs), have been shown to be able to perform inductive learning from a set of domain examples and to modify their behavior in the light of new training examples. Various techniques have been devised to extract symbolic rules from trained ANNs. In this talk I will give an overview of rule extraction techniques from trained ANNs, with specific focus on RULEX. I will then show how RULEX can be used to preconfigure an ANN according to a given set of symbolic rules. This ability to encode existing knowledge into the network, train, and then extract accurate rules makes RULEX the basis for a rule refinement system which may be used modify/maintain the rule base of an expert system.
 
 

ABSTRACT: Language is a product of three varieties of selection. Each requires variations among the self-replicating units upon which it operates and involves different mechanisms by which environments select surviving variants. First, prerequisite physiological attributes (e.g., vocal tract structure, neural organization) were selected phylogenically (Darwinian natural selection). Second, languages are acquired by individuals through ontogenic selection (the selection of behavior by its consequences within an individual lifetime, as when native but not nonnative speech sounds survive in a child's developing vocalizations). Ontogenic selection accounts for features that vary across languages (e.g., particular vocabularies, grammars, phonemic structures). Third, languages are perpetuated by human groups through cultural selection (selection of behavior as it is passed on among individuals). Languages selected ontogenically cannot be effective without cultural selection because they function within communities of speakers and listeners who pass language on from one to another. Thus, accounts of language origins must show not only how ontogenic selection could have been selected phylogenically but also how ontogenic and cultural selection could have evolved concurrently. This paper considers the language units and selection mechanisms involved at the three different levels of selection and argues that one critical event in the evolution of language was the transition from phytogenic to ontogenic selection of vocalizations.
 
 

ABSTRACT: Most children acquire language with no problem but some, though they they have no serious cognitive, neurological or auditory impairment, are unable to acquire language in the normal way. Over the past several years there has been converging evidence from epidemiological studies that show that developmental language impairment aggregates in families and is more concordant in monozygotic twins than in dizygotic twins. These data suggest that this disorder is likely to be heritable and provide us with a natural experiment for investigating how a genetic disorder can affect the development of language. This paper will report on an extensive cross-linguistic study of this disorder in English, Japanese and Greek. The analysis of spontaneous speech and tests of comprehension, production and grammaticality judgement converge to show that these subjects are not able to construct abstract, implicit rules in their grammars. They can compensate for this by storing inflected forms in their mental lexicon as unanalysed whole words. In some cases they can simulate the implicit, procedural rules of inimpaired grammars by learning simpler explicit rules. Cumulative data from several years of testing across several languages show that a genetic disorder can affect the ability to build a normal, rule-governed grammar and that in the absence of this ability subjects resort to other cognitive strategies in order to construct a language-like system.
 
 

ABSTRACT: This kinematic theory that can be used to analyze rapid human movements. A synergy of agonist and antagonist neuromuscular systems is involved in the production of these movements. These systems have a lognormal impulse response that results from the limiting behavior of a large number of interdependent neuromuscular networks, as predicted by the central limit theorem. The delta-lognormal law that follows from this model is very general and can reproduce almost perfectly the complete velocity patterns of an end effector. The theory accounts for the invariance and rescalability of these patterns, as well as for the experimental data on the change in maximum and mean velocities, time to maximum velocity, etc., under different experimental conditions. One of the most striking predictions of the theory is related to a quadratic law predicting movement time as a function of the various parameters describing the neuromuscular synergy. Conditions for a simplified description of the process, using a power law, are also presented. Movement time can be controlled at the input level by the ratio of agonist to antagonist commands or at the system level by modifying the total logtime delay or the logresponse time of the agonist and antagonist neuromuscular networks. Adapting this approach to the specific case of movements executed under different spatial accuracy demands, it is found that movement time is linked to the inverse of the relative spatial error by similar laws. The whole approach is used to explain within a single framework all the observations that have been reported concerning speed/accuracy tradeoffs. Strategies for controlling movement amplitude and duration are analyzed. Some new points of view brought up by the delta-lognormal theory in our general understanding of rapid movement will be discussed in the context of perceptuomotor interaction.
 
 

ABSTRACT: I will sketch a theory of what concepts of individuals, certain kinds, and stuffs are. These concepts are abilities to recognize these various things, in varous ways, when one encounters them, rather than, say, methods of classifying things. I will explain the difference. Then I will argue that believing what you hear said is surprisingly like believing what you see, so that concepts of this kind can be had through language alone without the ability to recognize these individuals, kinds or stuffs in the flesh.
 
 

ABSTRACT: In the early human evolution which preceded language, various behavioural means emerged to facilitate the prediction of the behaviour of fellow group members, to synchronize their behavior for common interests and to communicate thoughts. Mimesis and rituals are examples of such means. For about fourteen thousand years, dogs have been selected to live in an environment in which it is very important to predict the behaviour of the human master, to synchronize behaviour with his, and sometimes to communicate thoughts to him. If these assumptions are correct, then we should find the traces of this parallel evolution in present-day dog populations. In our Institute we have begun a program to test this through studies on human-dog communication in family dogs and various working dogs.
 
 

ABSTRACT: The bulk of animals' communications draws attention to the communicator and/or wards off other creatures. Some communicative signs can make another organism aware of both the communicator and of something in the environment besides that communicator. The vocal signs of vervet monkeys may be described this way. In their case the coordination achieved takes the form of combined visual attention to various sources of external threat. That is a significant behaviour. It does not seem to be referential; but it underpins a subsequent capacity for reference to environmental events and objects which emerges in evolutionary terms. In the case of the monkeys, the signs which achieve coordination are not, themselves, objects of the monkeys' attention. A context would have to arise in which coordinated activity that could induce combined attention is addressed to signs as such (signifiers), not just to what they signify. Referential communication results when coordinated activity takes the form of combined attention to external objects in association with signs that are also the combined (and "externalised") objects of attention.

ABSTRACT: An astonishingly large number of both popular and scholarly sources assert that the Eskimos have an interestingly large number of words for snow. Yet the claims about how many are all different. So are the conclusions we are apparently supposed to draw about language and thought. Laura Martin first drew attention to this in a paper on the anthropology of anthropologists [American Anthropologist 88.2 (1986) 418-423]. I subsequently published a tongue-in-cheek essay that mocked linguists for being unaware of her work and generally doing less than nothing to fight the spread of "The Great Eskimo Vocabulary Hoax" [the eponymous essay in my 1991 book] --- and have found, to my dismay, that it now gets cited as research on Eskimo lexicology. But in truth, although assorted lists of possibly relevant material have appeared on the LINGUIST mail list and the sci.lang newsgroup, no one has yet published a linguistic analysis and assessment of the relevant facts, or attempted to assess the logic of the multitude of morals for our theories of language and thought that have been implicitly attached to the supposed facts. In this work-in-progress talk, based on joint research with Laura Martin, a start is made on this project. The talk is relatively nontechnical, but departs from most previous work by actually looking at lexical material from some of the eight Eskimo languages.
 
 

ABSTRACT: Although general anaesthesia is believed to produce loss of consciousness, hypnotic regression and implicit memory tests have demonstrated that intraoperative events may be perceived and registered by anaesthetized patients. Recent studies on this topic, however, have produced conflicting results, raising questions under what specific circumstances positive evidence for intraoperative information processing can be found. It is plausible that that the inconsistent results are related to different states of anaesthesia during the moment of information presentation, which has not been taken into account in most studies. Moreover, there is still no reliable method to determine intraoperatively whether an auditory stimulus is actually perceived or not. The recording of Auditory Evoked Potentials (AEPs) provide a method for monitoring the transmission and processing of auditory stimuli from the cochlea to the cortex. In our studies, changes in long latency AEP components, following the presentation of frequent and infrequent tones of different pitch, were studied during several periods of cardiac surgery with propofol/alfentanil anesthesia. These components occur between 100 to 300 ms after stimulus presentation and reflect early discrimination processes. They are known to be affected by stimulus characteristics as well as by the subjects' attention. After the operation a recognition test with EEG recording was performed to examine whether intraoperatively presented words would provoke a (covert) recognition reaction in the brain. Results of our most recent study will be presented. They indicate that perceptual processing may continue during surgery, especially in the period before cardiopulmonary bypass. In addition, AEP amplitude differences for the two types of tones revealed a preserved ability to detect stimulus deviance. Although some patients in a previous study showed indications for covert recognition of intraoperatively presented material, this could not be replicated in the current study. Implications for clinical practice and parallels with sleep studies will be discussed.
 
 

ABSTRACT: There are two general approaches to the relationship between evolution and cognition. The first is based on the assumption (banal, according to Popper) that cognitive capabilities have evolved. There is almost universal agreement that this must be so. The implications of the assumption are not banal. The second approach, much less well known yet almost as old as the first, is that the transformation of neural and cognitive states in time are the result of the operation of evolutionary processes. Both approaches will be reviewed, the emphasis being on feasible connections.

Henry Plotkin came to England from South Africa in the 1960s, worked for the MRC and completed a doctorate in London. After a period at Stanford University (1970-1972) he returned to England and joined the department of psychology at UCL. Now Professor of Psychobiology and Head of the Department of Psychology at UCL, Plotkin's work has been largely on nonhuman species (flatworms, beetles, monkeys and various small rodents), but recently also on humans. His "Darwin Machines and the Nature of Knowledge" was published by Penguin 1995; the hardback versions are entitled "Darwin Machines" (Harvard University Press) and "The Nature of Knowledge" (Allen Lane) 1994
 
 

ABSTRACT: People with brain injury may lose the ability to recognise other people's faces, or to interpret their facial expressions (Young, Newcombe, de Haan, Small and Hay, 1993).
I will describe work on the perception of emotion in facial expressions, carried out as part of a collaborative project between myself and Andy Calder at MRC APU, David Perrett and Duncan Rowland at St. Andrews, and Nancy Etcoff of Harvard Medical School. We are using images interpolated ('morphed') between prototype expressions from the Ekman and Friesen (1976) series, to create photographic-quality continua running from one emotion to another (e.g., from happiness to sadness). We have used these to address three fundamental issues concerning the perception of facial expressions:
1. Are emotional facial expressions interpreted in terms of categories corresponding to basic emotions, or by locating them on a small set of underlying dimensions? Our results fit the categories position; for example, perception of the continuum of morphed images running from happiness to sadness shifts abruptly from one category to the other near the mid-point, and does not pass through any neutral or indeterminate region.
2. Can we find evidence of categorical perception of facial expressions, with better discrimination of images at category boundaries? We have several such findings (Calder, Young, Perrett, Etcoff and Rowland, in press), confirming previous results with line drawings reported by Etcoff and Magee (1992).
3. Is the perception of emotion in the face handled by a mechanism capable of dealing with all emotions, as Bruce and Young (1986) supposed, or are there discrete systems tuned to specific emotions? Our follow-up studies of the consequences of bilateral amygdala damage (Young, Aggleton, Hellawell, Johnson, Broks and Hanley, 1995) suggest that different emotions may have different neural substrates.

Bruce, V. and Young, A. (1986). Understanding face recognition. British Journal of Psychology, 77, 305-327.
Calder, A.J., Young, A.W., Perrett, D.I., Etcoff, N.L. and Rowland, D. (in press). Categorical perception of morphed facial expressions. Visual Cognition.
Ekman, P. and Friesen, W.V. (1976). Pictures of facial affect. Palo Alto, California: Consulting Psychologists Press.
Etcoff, N.L. and Magee, J.J. (1992). Categorical perception of facial expressions. Cognition, 44, 227-240.
Young, A.W., Aggleton, J.P., Hellawell, D.J., Johnson, M., Broks, P. and
Hanley, J.R. (1995). Face processing impairments after amygdalotomy. Brain, 118, 15-24.
Young, A.W., Newcombe, F., de Haan, E.H.F., Small, M. and Hay, D.C. (1993). Face perception after brain injury: selective impairments affecting identity and expression. Brain, 116, 941-959.
 
 
 

ABSTRACT: When people receive misleading suggestions about details in a previously witnessed event, they often later erroneously report the suggested details when asked about the event itself. At one time, such errors were taken as prima facie evidence that subjects believed they had actually witnessed the suggested details, but subjects might instead be aware that their responses are based on the postevent information rather than on the event itself (e.g., they might think "I don't remember what kind of weapon I saw, but I remember that the experimenter later said that it was a knife, so I'll say 'knife'"). The aim of the work to be reported in this talk is to determine if misled subjects sometimes genuinely believe that they remember witnessing things that were merely suggested to them and, if so, to explore the factors that determine the likelihood of such errors.
 
 

ABSTRACT: Recent studies have shown that lesions to the frontal lobes can produce a form of memory characterised by the production of excessive levels of false alarms in a range of recognition tests. In this talk I will describe two patients, JB and MR, who both exhibit this disorder following damage to the left pre-frontal cortex. Both patients appear to show relatively normal recognition on forced choice tests but memory deteriorates under more strenuous conditions. Further experiments indicate that this excessive false alarm rate extends beyond the use of standard verbal materials to incorporate nonwords, unfamiliar faces, and even conditions where no targets appear in the reocgnition sequence. It is suggested that these patients' deficit arises from an inability to set an appropriate criterion for recognition. However, the setting of this criterion is somehow linked to the conditions prevailing at encoding because the imposition of orienting tasks greatly reduces the level of false alarms - a finding that is consistent with recent PET data indicating the involvement of the left prefrontal cortex in memory encoding.
 
 

ABSTRACT: More effective research on design would result in more effective human-computer interactions (HCI). This talk will specify some of the relations between research and the design of human-computer interactions. The relations between Cognitive Science and the understanding of natural and artificial forms of intelligence will also be discussed from the standpoint of coherence, completeness and fitness-for-design (as opposed to fitness-for-understanding).

ABSTRACT: There is a large literature showing that auditory Event-Related Potentials (ERPs) are sensitive indicators of attention and information processing. ERPs are accordingly useful in studying the effects of state changes on attentiveness and individual differences in attentional capacity. The most studied component of the auditory ERP is the P3, which is traditionally associated with rare stimuli or stimuli requiring some response. Recent studies suggest that neither of these attributes is an underlying factor in P3 amplitude; the findings instead support a salience model for P3 amplitude: the P3 in the standard paradigm reflects the allocation of attention to the stimulus and not the inherent processing of it. Salience also accounts better for the late positive component in recognition memory and memory-for-source studies.
We have also done studies involving subtle reductions in the ability to allocate attention without reductions in automatic sensory processing. A low dose (.04% BAC) of alcohol affects P3 components and not earlier N1, P2, and N2 components. Caffeine, in contrast, has more complex stimulative effects on the ERP. We also find dramatic effects of P3 attenuation despite no other obvious current symptoms in individuals with a mild but long-past head injury

Dywan, J. & Segalowitz, S.J. (1996). Measures of adaptive functioning and their relationship to attentional processes. Journal of Head Trauma Rehabilitation, 11, 81-98.
Segalowitz, S.J. & Lawson, S. (1995). Subtle symptoms assocated with self-reported mild head injury. Journal of Learning Disabilities, 28, 309-319.
Unsal, A. & Segalowitz, S.J. (1995). Sources of P300 attenuation after head injury: Single trial amplitude, latency jitter, and EEG power. Psychophysiology, 32(3), 249-256.
Segalowitz, S.J. & Berge, B.E. (1995). Functional asymmetries in infancy and early childhood: A review of electrophysiological studies and their implications. In R.J. Davidson & K. Hugdahl (Eds.), Brain asymmetry (pp 579-616). Cambridge, MA: MIT Press.

Sid Segalowitz B.A. McGill U, 1970 Ph.D. Cornell U, 1974 Brock University 1974 - present. Currently at the MRC-APU for 5 months. Areas of interest: I have worked in the past in laterality studies and neurolinguistics, and am currently involved in developmental neuropsychology. Throughout all of this, I have been working with ERPs and EEG in the study of attention, the control of attention, and their relationship to arousal systems and the risk factors such as mild head injury.
 
 

ABSTRACT: Voicing is perhaps the most basic articulatory and acoustic distinction in speech yet we know little about the underlying auditory mechanisms of its perception, beyond the fact that it is "categorical," i.e. perception changes abruptly from `voiced' to `unvoiced' as voice onset time (VOT) is uniformly increased, and discrimination is better between categories than within a category. We have a computer simulation (Pont and Damper, 1991; Damper et al, forthcoming) capable of reproducing faithfully the important behavioural effects associated with the perception of VOT in initial stops: the warping of similarity space underlying the voiced/voiceless distinction and the boundary shift effect, whereby the boundary between categories shifts with place of articulation. The simulation consists of a (spiking) model of the peripheral auditory system whose `neurogram' output is fed to a trainable connectionist network acting as a synthetic listener. The use of a connectionist system to bridge between the sensory and perceptual/cognitive worlds is interesting, in light of Harnad's (1990, 1992) suggestions that such systems are "one natural candidate for the mechanism that learns the invariant features underlying categorical perception, thereby connecting names to the ... objects they stand for." The ability of the simulation to mimic speech CP phenomena in precise detail (Damper, 1994) is potentially very important, as parts of the model can be purposefully changed and the effect on synthetic behaviour assessed. In this way, we can hope to understand better the nature of VOT perception. This talk will detail the model manipulations we have studied recently, and the implications of the findings for a theory of speech CP.

DAMPER, R.I. (1994) ``Connectionist models of categorical perception of speech'', Proceedings of IEEE International Symposium on Speech, Image Processing and Neural Networks, Vol.1, 101--104, Hong Kong.
DAMPER, R.I., PONT, M.J. and ELENIUS, K. (forthcoming) ``Representation of initial stop consonants in a computational model of the dorsal cochlear nucleus'', in Advances in Speech, Hearing and Language Processing, Vol.3, W.A. Ainsworth (ed.), JAI Press, Greenwich, CT, in press.
HARNAD, S. (1992) ``Connecting object to symbol in modeling cognition'', in Connectionism in Context, A. Clarke and R. Lutz (eds.), pp.75--90, Springer Verlag, Berlin.
HARNAD, S. (1990) ``The symbol grounding problem'', Physica D, Vol.42, 335--346.

Bob Damper is a Senior Lecturer in Electronics and Computer Science here at Southampton, where he has taught and researched for the last 16 years. He gained his PhD from the Department of Electrical Engineering at Imperial College in the area of computational modelling of the auditory system. He has extensive research interests in speech science and technology (including human factors aspects of speech I/O), neural computation applied to computer speech and vision, rehabilitation engineering (especially information ergonomics), cognitive modelling for `artificial perception', and intelligent systems engineering. He is the author of the undergraduate text ``Introduction to Discrete-Time Signals and Systems'' and co-editor (with Wendy Hall and John Richards) of ``Multimedia Technologies and Future Applications''.
 

ABSTRACT: Experimental observations of the brain activity that follows sensory stimulation of animals show that sensory cortices engage in construction of activity patterns in response to stimuli. The operation is not that of filter, retrieval, or correlation mechanisms. It is a state transition by which a cortex switches abruptly from one basin of attraction to another, thereby to change one spatial pattern to another like frames in a cinema. The transitions in the primary sensory cortices are shaped by interactions with the limbic system, which express the goal-directed nature of percepts. They result from intentional actions in time and space. Each transition involves learning, so that cumulatively a trajectory is formed by each brain over its lifetime. Each spatial pattern as it occurs reflects the entire content of individual experience. It is a meaning and not the representation of a meaning. It is the basis for consciousness.
It follows that each brain creates its own frames of reference for time, space and associations, which are not directly accessible by any other brain. How, then, can two or more brains be shaped by learning, so as to form cooperative pairs for reproduction and groups for survival? Evolution has provided a biological, mechanism that first came under scientific scrutiny in the form of Pavlovian 'brain washing'. Under now well known conditions of stress in the internal and external environments, a global transition takes place, following which brains sustain a remarkable period of malleability. I believe that Pavlov manipulated a mammalian mechanism of pair bonding, already evolved for the nurture of altricial young through sexual orgasm and lactation, which is mediated by oxytocin and other neuropeptides. Our remote ancestors evolved by adapting this mechanism for tribal bonding through dance, chanting, rituals, and evangelical conversions (Sargant 1957). These dimensions of human experience can be encompassed by a theory of neurodynamics, but not by theories of representation.

Freeman WJ (1995) Societies of Brains. Hillsdale NJ, Lawrence Erlbaum
Sargant W (1957) Battle for the Mind. Westport CT, Greenwood Press

Walter Freeman studied electronics, physics and mathematics at the Massachusetts Institute of Technology, and literature and philosophy at the University of Chicago. He received his medical degree cum laude from Yale University in 1954, and took postdoctoral training in internal medicine at Johns Hopkins University, and neurobiology at UCLA as Fellow of the Foundations Fund for Research in Psychiatry. Since 1959 he has taught neuroscience in the University of California at Berkeley. He received the A. E. Bennett Award of the Society of Biological Psychiatry (1964), Guggenheim Award (1966), Titulaire de la Chaire Solvay, University of Brussels (1974), the MERIT Award from NIMH (1990), the Pioneer Award from the Neural Network Council of the IEEE(1991), and the Spinoza Chair of the University of Amsterdam (1995). He was President of the International Neural Network Society in 1994. His research interests lie in mathematical modeling of nonlinear neurodynamics, based on his experimental measurements of brain activity in behaving animals, and the application of these models in biology, neurology, psychiatry, philosophy, and industry.
 
 

ABSTRACT: From both an applied and scientific psychophysiological point of view, the related but different ideas of using physiological measures to detect and differentiate deception are of considerable potential interest. The detection of deception is of obvious practical social significance, while deception as a psychological process is of great interest in its own right, if only because of its evolutionary significance as an alternative to the fight/flight response. A disinterested approach to both detection and differentiation is one that employs the fundamental terms in a logically consistent way, and that bases its conclusions on empirical evidence interpreted in a logical rather than rhetorical way. I shall suggest that, by and large, the research psychophysiological community, as represented by the Society for Psychophysiological Research (SPR), has failed to measure up to the standards of disinterestedness both with respect to the detection and to the differentiation of deception. Instead, SPR has adopted an uninterested perspective. This has allowed the interested community of professionals who employ that peculiarly North American flight of technological fancy, the so- called "Control" Question so-called "Test" (CQT) polygraph, to hood-wink both themselves and others (including the American Psychological Association) that the CQT is a controversial, but scientifically-based, test for detecting deception. The tradition of disinterested enquiry is relatively strong among British experimental psychophysiologists, who also tend to be less pressured by vested polygraphic interests than their North American colleagues. British psychophysiologists, therefore, are in a good position to exert their influence so that the concept of disinterestedness is applied by more psychophysiologists to the detection and differentiation of deception.

1. Furedy, J.J. (1993). The `control' question `test' (CQT) polygrapher's dilemma: Logico-ethical considerations for psychophysiological practitioners and researchers. International J. of Psychophysiology, 15, 263-75.
2. Furedy, J.J. (1989). The NOrth American CQT polygraph and the legal profession: A case of Canadian credulity and a cause for cultural concern. Criminal Law Quarterly, 31, 431-51.
3. O'Gorman, J. (1994). Review of Ben-Shakhar and Furedy's Theories and Applications in the Detection of Deception, Australian Journal of Psychology, 46, 59-60.

ABSTRACT: During the last few years Canadian campuses, in order to combat the so-called problem of "chilly climate" have instituted speech codes which go beyond freedom-of-speech restrictions of Canadian civil law (which is already totalitarian enough with it's "hate" laws), and which establish comfort as the criterion for determining whether opinion is permissible. The resultant "velvet" totalitarianism that has resulted shares all the features of twentieth century totalitarian systems like German Nazism, Italian Fascism, and Soviet Communism, except for the severity of the punishments meted out for deviancy. Similarly, except for the severity of the penalty for being found "guilty", the 1995 summer events at the University of British Columbia bear all the features of the Salem witch hunts, including the feature that the attack on freedom of speech comes from within rather than outside the community. I shall discuss both the spectacular and the more subtle effects of the culture of comfort on members of the academic community (not only faculty, but also students), and shall suggest that apathy about these matters is an irrational as well as a cowardly strategy.
1. Furedy, J.J. (1995). Academic freedom, opinions and acts: The Voltaire-Mill perspective applied to current Canadian cases. University of New Brunswick Law Review, 44, 131-4.
2. Furedy, J.J. (1994). Ice stations academe: Is an Iron Curtain of speech being erected in NOrth American universities? A personal perspective from the Unversity of Toronto. Gravitas, Fall Issue, 18-22.
 
 

ABSTRACT: This talk will compare three recent pragmatic accounts of bridging reference. Data are drawn from an informal questionnaire (Matsui 1995), in which subjects were asked to assign reference to the italicised expressions in examples like (1), where the prior text provides access to several candidate referents: (1) Kevin moved from New Zealand to England. He loves the sheep. All three accounts share the following assumptions: (a) Candidate referents differ in their order of accessibility, and are therefore evaluated in a certain order; (b) They are evaluated in terms of some criterion of pragmatic acceptability that the interpretation is supposed to meet. Different pragmatic theories offer criteria of pragmatic acceptability. I will compare theories offering truth-based (Lewis 1983), coherence-based (Hobbs 1979, Asher & Lascarides 1993) and relevance-based (Sperber & Wilson 1995) criteria, and argue that the relevance-based criterion is preferable on both descriptive and explanatory grounds.
 
 

ABSTRACT : The study of discourse and discourse understanding is dominated by the view that the production and interpretation of utterances in discourse depends on the identification of coherence relations. I shall argue that coherence can be explained in terms of the speaker's search for relevance, and that this account provides a framework for the analysis of a range of so- called discourse connectives whether they occur in linguistic or non-linguistic contexts.
 
 
 

ABSTRACT: The human species is a prime example of a goal-directed system with an emergent 'read-write' dynamics (in the sense discussed by Pattee). This dynamical systems assumption contrasts sharply with the neoPlatonist view that language emerged from an evolutionary accident and that linguistic phenomena are best characterised in purely formalistic terms. The latter view leaves linguistic transactions semantically ungrounded in the everyday situations where the social transactions they support take place. Without a theory of what grounds it, language phenomena present an ontological mystery rather than just a scientific puzzle. Of special interest, in this regard, is the role language plays in the promotion of individual and social agency (i. e., co-operative or competitive intentional acts). The nature of intentional acts and language as an instrument for such acts will be explored from the radical pragmatic point of view that "a thing is what it means" (Gibson, 1979).

Robert Shaw, Professor of Psychology at the University of Connecticut, is a co-founder and current president of the International Society of Ecological Psychology. He is the author of approximately one-hundred theoretical and research articles, including three books, on topics ranging from psycholinguistics, child development, to perception and action. He is an editor of a book series, Resources for Ecological Psychology, and on the editorial boards of the journals Ecological Psychology and Intelligent Systems. Currently his interests center on developing a method, called 'intentional dynamics', aimed at modelling goal-directed systems in situated learning and situated problem-solving tasks.

Gibson, J. J. (1979) An ecological approach to visual perception. Boston: Houghton Mifflin Shaw, R. & Kinsella-Shaw, J. (1988). Ecological Mechanics: A Physical Geometry for Intentional Constraints, Human. Movement Science., 7, 155.
Shaw, R., Kadar, E., Sim, M., & Repperger, D. (1992). The intentional spring: A strategy for modelling systems that learn to perform intentional acts. Journal of Motor Behavior, 1, No.24, 3-28.
Shaw, R., Flascher, O., & Mace, W (1996). Dimensions of event perception. In W. Prinz & B. Bridgeman (Eds.) Handbook of Perception. and Action: Volume 1. London: Academic Press.
Sim, M., Shaw, R. E., & Turvey, M. T. (in press). Intrinsic and required dynamics of a simple bat-ball skill. Journal of Experimental Psychology: Human Perception and Performance

ABSTRACT: A rational analysis and heuristic processes for syllogistic reasoning are proposed. An informational ordering over quantified statements suggests simple information-based heuristics for syllogistic reasoning. The most important is the "min-heuristic," that people choose the type of the least informative premise as the type of the conclusion. The rationality of this heuristic is confirmed by a probabilistic rational analysis of syllogistic reasoning, which treats logical inference as a limiting case of probabilistic inference. A meta-analysis of past experiments reveals close fits with the rational heuristics model (RHM). RHM also compares favorably with alternative accounts, including mental logics, mental models, and deduction as verbal reasoning. Crucially, RHM extends naturally to generalized quantifiers, such as Most and Few, which have not been characterized logically and are consequently beyond the scope of current mental logic and mental model theories. Two experiments confirm the novel predictions of RHM when generalized quantifiers are used in syllogistic arguments. RHM suggests that syllogistic reasoning performance may be determined by simple but rational informational strategies justified by probability theory rather than by logic.

Nick Chater is Professor of Psychology at the University of Warwick. As an undergraduate, he studied philosophy and psychology at Trinity College, Cambridge. He took a PhD in Cognitive Science at the University of Edinburgh, working on information-based semantics for mental representations, as well as conducting empirical and computational work. He has been a lecturer in the Psychology Departments at University College London, Edinburgh University and Oxford University. His research interests concern computational and mathematical models of cognition, particularly concerning inference, categorization, learning and language processing.
 
 
 

ABSTRACT: Unlike when one hears speech, classifying speech from a seen face can make use of displays that are not dynamically structured. Is this epiphenomenal? Is "real speechreading" no more and no less than the tracking of the dynamic trajectories of the seen articulators, with no contribution from visual form processing? In this talk I suggest that although perceived facial movement is vital to effective speechreading, it cannot, on its own, deliver useful speech information; visual form processing mechanisms are likewise required. Some neural correlates of speechreading will also be discussed.

Campbell et al. 1986, Brain; Phil Trans R.Soc Lond B, 1992
Rosenblum & Saldana, JEP 1996

Ruth Campbell obtained her BA in Psychology at Reading University(1970) and her PhD (1979) from the University of London, where Max Coltheart was her superviser. Her thesis was on facial asymmetries - perceptual and expressive. She has been interested in speechreading as a way of exploring how face processing and language processing converge and diverge. She uses experimental and neuropsychological tasks to explore this. Her other interests include reading and writing (especially in development) and face processing, including face processing in special groups, including people with autism. A new interest is in Cognition and Deafness. She was appointed Professor of Communication Disorders at University College London, Department of Human Communication Science, in October 1996.
 
 

ABSTRACT: In a fraction of a second, people are generally able to recognize most objects and scenes, even from novel images. Is this extraordinary capacity based on generalization from templates of similar objects, as assumed by recently proposed "Appearance Theories"? Or is a descriptive system required that specifies an object's simple parts (or geons) and relations, both based on nonaccidental properties of object edges? A critical review and recent experimental results provides little support for appearance-based accounts. As long as two or three distinctive geons in their specified invariant relations can be extracted from the image, fast basic- and most subordinate-level classifications will almost always be successful despite drastic variations in viewpoint, lighting, orientation, and noise. The special case of face recognition and its comparison to object recognition will also be discussed.

Biederman, I. (1995). Visual object recognition. In S. F. Kosslyn and D. N. Osherson (Eds.). An Invitation to Cognitive Science, 2nd edition, Volume 2., Visual Cognition. MIT Press. Chapter 4, pp. 121-165.
Biederman, I., Gerhardstein, P.C. , Cooper, E. E., & Nelson, C. A. (1997). High Level Object Recognition Without an Anterior Inferior Temporal Cortex. Neuropsychologia, in press.
Biederman, I., & Gerhardstein, P. C. (1995). Viewpoint-dependent mechanisms in visual object recognition: Reply to Tarr and Bülthoff (1995). Journal of Experimental Psychology: Human Perception and Performance, 21, 1506-1514.
Biederman, I., & Gerhardstein, P. C. (1993). Recognizing depth-rotated objects: Evidence and conditions for 3D viewpoint invariance. Journal of Experimental Psychology: Human Perception and Performance, 19, 1162-1182.
Hummel, J. E., & Biederman, I. (1992). Dynamic binding in a neural network for shape recognition. Psychological Review, 99, 480-517.

Irving Biederman is the Keck Professor of Cognitive Neuroscience at the University of Southern California where he directs the Image Understanding Laboratory. He is a member of the Departments of Psychology and Computer Science and is an active member of the Neuroscience Program. Prior to his appointment at the University of Southern California, he was the Fesler-Lampert Professor of Artificial Intelligence and Cognitive Science at the University of Minnesota. Author of over 100 scientific publications, Professor Biederman has developed a theory of real-time human object and scene recognition that posits a representation specifying simple viewpoint-invariant shape primitives, termed geons. Professor Biederman has been a member of panels for the National Science Foundation, National Research Council, and the Air Force Office of Scientific Research, where he served as the first Program Manager (consulting) for the Cognitive Science Program.
 
 
 

ABSTRACT Spatial quantisation of visual images (e.g., faces) with varying grey levels into isoluminant, square-shaped pixels has a detrimental effect on the efficiency of identifying the original image. The effect of the spatial scale level of quantisation is "catastrophic": at certain intermediate-to-coarse levels of quantisation only a minor further increment in pixel size leads to an abrupt decrement in identification. If the quantisation is performed at intermediate or coarse levels, identification can be impaired from a variety of causes: low-pass spatial frequency filtering, masking of low frequencies by the high-frequency content of the pixels' edges and corners, distortion of the internal relational metrics of the elements of image configuration, attentional competition between two variants of image interpretation (face versus "mosaic"). Experiments with valid attentional, peripheral precuing of quantised images have revealed the relative cost of covert attentional orientation on identification. The findings are best explained by a microgenetic theory of the gradual, spatial scale-dependent activation of attentional filters that are tuned to progressively finer-scale levels of image representation, the latter having been formed preattentively.

1. Bachmann,T. (1988). Time course of the subjective contrast enhancement for a second stimulus in successively paired above-threshold forms: perceptual retouch instead of forward masking. Vision Research, 28(11), 1255-1261.
2. Bachmann,T. (1994) "Psychophysiology of Visual Masking", Commack, New York: Nova Science Publishers (298 p.).
3. Bachmann,T. (1996) Fast dynamics of subjective experience of the paired, transient visual signals: some evidence for "perceptual retouch" as a semiautonomous process of assigning conscious quality. Journal of Consciousness Studies (Supplement): "Towards a Science of Consciousness 1996 ("Tucson II")), 62-63.
4. Bachmann,T. (1991) Identification of spatially quantised tachistoscopic images of faces: How many pixels does it take to carry identity? European Journal of Cognitive Psychology, 3(1), 87-103.
5. Bachmann,T. & Kahusk,N. (1996). The cost of spatial-attentional precuing on the perception of spatially quantised visual images. Perception, 25, Supplement, 1-2.

Talis Bachmann, Cand.Psych.,Ph.D. Graduated from Tartu University, Estonia, 1974. Candidate's thesis 1979, Russian Academy of Sci-es. Doctor of Psychology thesis 1989, Moscow University. Has held posts in Tartu University and Tallinn University of Social and Educational Sciences (both in Estonia). Currently works in the Department of Psychology, University of Portsmouth. Has been and/or is a member of the editorial and advisory boards to: European Journal of Cognitive Psychology, Psychological Research/Psychologische Forschung, Consciousness and Cognition, Attention and Performance, Stichting. Primary research interests: visual masking, processing quantised images, spatial scale effects, spatial attention, psychophysiological correlates of visual awareness.
 
 
 

ABSTRACT: The existence of optic aphasia has been taken as strong evidence for a fractionation of knowledge representations into modality-specific systems in addition to verbal semantics. In this talk, the central issues concerning optic aphasia will be reassessed by interpreting case studies within a framework that allows for a fractionation of object knowledge. Particular consideration will be given to the retrieval of knowledge concerning colours of objects. A (semi) direct route will be proposed from the object's structural description to the output lexican without access to functional or associative object knowledge.

Davidoff, J. and Masterson, J. (1996) The development of picture naming: Differences between nouns and verbs. Journal of Neurolinguistics, 9, 69-83.
Luzzatti, C. and Davidoff, J. (1994) Impaired retrieval of object-colour knowledge with preserved colour naming. Neuropsychologia, 32, 933-950.
Davidoff, J. and De Bleser, R. (1993). Optic aphasia: a review of past studies and reappraisal. Aphasiology, 7, 135-154.

Professor Jules Davidoff studied psychology at University College, London and has taught the subject at the Universities of Wales (Swansea), Edinburgh, London (LSE), and Ghana before being appointed to the new Department of Psychology at the University of Essex. His main research interests are in impairments to cognitive processes and has worked on these issues at the MRC Neuropsychology Unit, University of Oxford and at the National Hospital for Neurology and Neurosurgery. He has a special interest in higher-level colour processing and written a text for MIT press called "Cognition through Color".
 
 

ABSTRACT: Is knowledge of number and simple arithmetic part of "ordinary" semantic memory, mixed up with general knowledge and organised according to the usual principles - acquisition history, multidimensional similarity, etc? Or is it a separate mental and neural domain, with its own organisational principles? Our studies of neuropsychological patients show that number abilities dissociate from general cognitive abilities, from language ability, from reading abilities and from Short-Term Memory. Of course, the lesion sites for the number deficits is quite different from deficits of language and semantic memory. Within the number domain, patient studies suggest functional subdivisions for transcoding numbers vs calculation; and within calculation, for arithmetic facts, conceptual knowledge and calculation procedures. There is evidence that the organising principles of stored number knowledge is indeed domain specific.

Cipolotti, L., Butterworth, B. & Denes, G. (1991) A specific deficit for numbers in a case of dense acalculia. Brain, 114, 2619-2637

Cipolotti, L. & Butterworth, B. (1995) Towards a model a multiroute model of number processing: impaired number transcoding with the preservation of calculation skills. Journal of Experimental Psychology: General. 124, 375-390

Butterworth, B., Cipolotti, L. & Warrington, E.K. (1996) Short-term memory impairments and arithmetical ability. Quarterly Journal of Experimental Psychology. 49A, 251-262

Professor Brian Butterworth is currently, Professor of Cognitive Neuropsychology at UCL and Editor of the journal "Mathematical Cognition".He is on the Executive Committee of the new Institute of Cognitive Neurscience at UCL. His main areas of research are speech production and its disorders, reading and its disorders (especially of non-alphabetic scripts), and mathematical cognition. He is very keen to contact anyone who has great difficulty with numbers.
 
 
 

ABSTRACT: How can any physical process give rise to subjective experience? A compelling case can be made for the emergence of consciousness - in phylogeny, ontogeny, and from day to day - in neural networks meeting specific requirements for the generation of electrochemical fields having appropriate spatio-temporal properties.

Professor Michael Corner was formerly Department Head, Comparative and Development Physiology, Institute for Brain Research, The Netherlands. He is currently Project Leader for the Biology Research Council section in Neurobiology; 'Special' Professor of Developmental Physiology, Department of Zoology, University of Amsterdam and Visiting Investigator, Netherlands Institute for Brain Research, Amsterdam.
 
 

ABSTRACT: Dr Murre will present a model of amnesia (the TraceLink model) based on a review of its neuropsychology, neuroanatomy, and connectionist modelling. The model consists of three subsystems: (1) a trace system (neocortex), (2) a link system (hippocampus), and (3) a modulatory system (basal forebrain nuclei, controlled by among others hippocampus). The hippocampus is assigned a double role, being involved in both the link system and the modulatory system. The model is able to account for many of the characteristics of semantic dementia, retrograde amnesia (AA) and anterograde amnesia (RA), including: Ribot gradients, transient-global amnesia, patterns of shrinkage and recovery from AA, correlations between AA and RA or the absence thereof (e.g., as in isolated RA). In ddition we derive certain testable predictions concerning implicit memory, forgetting gradients, and neuroanatomy.

Murre, J.M.J. (1992). Categorization and learning in modular neural networks. Hemel Hempstead: Harvester Wheatsheaf, and Hillsdale, NJ: Lawrence Erlbaum.

Murre, J.M.J., & D.P.F. Sturdy (1995). The connectivity of the brain: multi-level quantitative analysis. Biological Cybernetics, 73, 529-545.

Murre, J.M.J. (1996). TraceLink: a model of amnesia and consolidation of memory. Hippocampus, 6, 675-684.

Murre, J.M.J. (1997). Implicit and explicit memory in amnesia: some explanations and predictions by the TraceLink model. Memory, 5, 213-232.

Dr. Jaap Murre holds a research fellowship by the Netherlands RoyalAcademy of Arts and Sciences. From 1992-1995 he worked at the MRC Applied Psychology Unit in Cambridge. His interests include connectionist modelling of memory, amnesia, recovery from brain-damage in general, as well as some of the more technical aspects of neural network research, such as the development of neurosimulators and neurocomputers. Most modelling work is aimed to be at least 'biologically informed', and one of his recent projects was on modelling of the quantitative neuroanatomy of the connectivity of the brain.

This page has been accessed  times.