Volume 53, Issue 7-8

The criterion that pure natural science can only investigate objective phenomena which can be observed by independent observers sets certain limits to pure natural scientific under­ standing of brain functions. It excludes consciousness and feelings since these are only sub­ jectively accessible and alternatives cannot be decided objectively. The limitations of brain research are discussed by comparing the properties of brains and computers. At least for the time being we do not know of any natural scientific -i.e. physical or chemical -method which allows the objective measurement of consciousness, sensations, and emotions. In addition it is discussed how and in how far our brain understands its surrounding nature.

Our working assumption reads: Every mental property exhibited by a natural system can be exhibited by an artificial system, at least in principle. Is this true? The paper is in two parts. The first part is explicative: What do we mean by ‘natural’ or by ‘artificial’? When are we ready to ascribe thinking to a system? We show that sometimes behavior is enough, sometimes underlying mechanisms are decisive. We study the difference between simulation and realization of a property and the kinds of explanation used in the philosophy of mind. The second part is argumentative. We collect and discuss positive arguments for and critical arguments against the working hypothesis. No decisive counterargument is found

The goal of the present paper is to provide an embodied cognitive science view on representation. Using the fundamental task of category learning, we will demonstrate that this perspective enables us to shed new light on many pertinent issues and opens up new prospects for investigation. The main focus of this paper is on the prerequisites to acquire representations of objects in the real world. We suggest that the main prerequisite is embodiment which allows an agent - human, animal or robot - to manipulate its sensory input such that invariances are generated. These invariances, in turn, are the basis of representation formation. In other words, the paper does not focus on representations per se, but rather discusses the various processes involved in order to make learning and representation acquisition possible. The argument structure is as follows. First we introduce two new perspectives on representation, namely frame-of-reference, and complete agent. Then we elaborate the complete agent perspective and focus in particular on embodiment and situatedness. We argue that embodiment has two main aspects, a dynamic and an information theoretic one. Focusing on the latter, there are a number of implications: Representation can only be understood if the embedding of the neural substrate in the physical agent is known, which includes morphology (shape), positioning and nature of sensors. Because an autonomous mobile agent in the real world is exposed to a continuously changing high-dimensional stream of sensory stimulation, if it is to learn category distinctions, it first needs a focus of attention mechanism, and then it must have a way to reduce the dimensionality of this high-dimensional sensory stream. Learning is very hard because the invariances are typically not found in the sensory data directly - the classical problem of object constancy: it is a so-called type 2 problem. Rather than trying to improve the learning algorithms - which is the standard approach - the embodied cognitive science view suggests a different approach which focuses on the nature of the data: the agent is not passively exposed to a given data distribution, but, by exploiting its body and through the interaction with the environment, it can actually generate the data. More specifically, it can generate correlated data that has the property that it can be easily learned. This learnability is due to redundancies resulting from the appropriate interactions with the environment. Through such interactions, the former type 2 problem is transformed into a type 1 problem, thus reducing the complexity of the learning task by orders of magnitude. By observing the frame-of-reference problem we will discuss to what extent these invariances are reflected - represented - in the “neural substrate”, i.e. the internal mechanisms of the agent. It is concluded, that representation is not a concept that can be studied in the abstract, but should be elaborated in the context of concrete agent-environment interactions. These ideas are all illustrated with examples of natural agents and artificial agents. In particular, we will present a suite of experiments on simulated and real-world artificial agents instantiating the main arguments

Experimental evidence suggests that the hippocampus represents locations within an allocentric representation of space. The environmental inputs that underlie the rat’s representa­tion of its own location within an environment (in the firing of place cells) are the distances to walls, and different walls are identified by their allocentric direction from the rat. We propose that the locations of goals in an environment is stored downstream of the place cells, in the subiculum. In addition to firing rate coding, place cells may use phase coding relative to the theta rhythm of the EEG. In some circumstances path integration may be used, in addition to environmental information, as an input to the hippocampal system. A detailed computational model of the hippocampus successfully guides the navigation of a mobile robot. The model’s behaviour is compared to electrophysiological and behavioural data in rats, and implications for the role of the hippocampus in primates are explored.

The issue of translating the planning of arm movements into muscle forces is discussed in relation to the recent discovery of structures in the spinal cord. These structures contain circuitry that, when activated, produce precisely balanced contractions in groups of muscles. These synergistic contractions generate forces that direct the limb toward an equilibrium point in space. Remarkably, the force outputs, produced by activating different spinal-cord structures, sum vectorially. This vectorial combination of motor outputs might be a mecha­nism for producing a vast repertoire of motor behaviors in a simple manner.

A model of recognition is described based on cell properties in the ventral cortical stream of visual processing in the primate brain. At a critical intermediate stage in this system, ‘Elaborate’ feature sensitive cells respond selectively to visual features in a way that depends on size (± 1 octave), orientation (± 4 5 °) but does not depend on position within central vision (± 5 °). These features are simple conjunctions of 2-D elements (e.g. a horizontal dark area above a dark smoothly convex area). They can arise either as elements of an object’s surface pattern or as a 3-D component bounded by an object’s external contour. By requiring a combination of several such features without regard to their position within the central region of the visual image, ‘Pattern’ sensitive cells at higher levels can exhibit selectivity for complex configurations that typify objects seen under particular viewing conditions. Given that input features to such Pattern sensitive cells are specified in approximate size and orientation, initial cellular ‘representations’ of the visual appearance of object type (or object example) are also selective for orientation and size. At this level, sensitivity to object view (± 6 0 °) arises because visual features disappear as objects are rotated in perspective. Processing is thus viewer-centred and the neurones only respond to objects seen from particular viewing conditions or ‘object instances’. Combined sensitivity to multiple features (conjunctions of elements) independent of their position, establishes selectivity for the configurations of ob­ject parts (from one view) because rearranged configurations of the same parts yield images lacking some of the 2-D visual features present in the normal configuration. Different neural populations appear to be selectively tuned to particular components of the same biological object (e.g. face, eyes, hands, legs), perhaps because the independent articulation of these components gives rise to correlated activity in different sets of input visual features. Generalisation over viewing conditions for a given object can be established by hierarchically pooling outputs of view-condition specific cells with pooling operations dependent on the continuity in experience across viewing conditions. Different object parts are seen together and different views are seen in succession when the observer walks around the object. The view specific coding that characterises the selectivity of cells in the temporal lobe can be seen as a natural consequence of selective experience of objects from particular vantage points. View specific coding for the face and body also has great utility in understand­ing complex social signals, a property that may not be feasible with object-centred processing.

In this contribution, it is argued that the prevailing view of the sensory system as an intricate, but passive processor of external information does not capture the full complexity of brain performance. Instead, we try to reinforce a notion that sees the brain as a system embedded within the environment and actively exploring it. We will attempt to emphasize the bidirectional interaction between brain and environment at all levels of processing and at different scales of the system description. This modified approach to the understanding of the brain has profound consequences for experimental investigations, starting with the experimental design and extending into data analysis and interpretation.

The development of autonomous as well as situated robots is one of the great remaining challenges and involves a number of different scientific disciplines. In spite of recent dramatic progress, it remains worthwhile to examine natural systems, because their abilities are still out of reach. Motivated by research work done in the fields of cognitive systems, visual perception, and psychology of memory we designed and implemented a memory architecture for visual tasks. Structural and functional concepts of the memory architecture were modeled on the ones found in natural systems. We present an efficient implementation based on parallel programming techniques. The memory module is integrated into a distributed system for speech and image analysis, which is currently developed in the Sonderforschungsbereich (SFB) 360, Situated Artificial Communicators, where a hybrid vision system combining neural and semantic networks is used.

Comparative neuroethological research emphasizes that brains of animals have been shaped by the specific demands and constraints imposed by the ecological niche that a species occupies. Since avian species have developed very diverse life styles and occupy extreme ecological niches, bird brains should show many specializations, which may be revealed in species that have survived under high ecological pressures. In this paper, we will give several examples of adaptations, in which we are able to correlate structural and physiological spe­cializations to the specific ecological demands: adaptations found to nocturnal hunting in barn owls, the characteristics of bird song and its underlying neurobiological correlates, retinopetal projections and their relation to peripheral attentional switching, looming detection, and adaptations related to memory capacities of food-storing birds. We stress especially that the analysis of the animal’s ecological situation is important in understanding the factors that shaped both behavior and the neuronal substrate.

Arthropods live in almost any conceivable habitat. Accordingly, structural and functional specialisations have been described in many species which allow them to behave in an adap­tive way with the limited computational resources of their small brains. These adaptations range from the special design of the eyes, the spectral sensitivities of their photoreceptors to the specific properties of neural circuits.

The adaptive control of behaviour requires brain mechanisms for the selection (i.e. activa­tion and suppression) of responses, as well as mechanisms for the modulation of the response vigour. The concept of motivation postulates the existence of brain centres that regulate the selection and strength of behavioural responses. The present paper provides examples from the behavioural neurosciences for brain mechanisms that lead to adaptive changes of an organisms responsiveness to external stimuli. The mammalian startle response is a simple defensive behaviour which is mediated by an oligosynaptic pathway located in the lower brainstem. The startle response is enhanced by aversive states (fear, anxiety) and attenuated by appetitive states (pleasure), which can be regarded as an example of motivational priming. Furthermore, the startle response is inhibited by a weak sensory stimulus presented shortly before the startling stimulus. The suppression of startle by a prepulse is an example of sen­ sorimotor gating, a principle that is important for the hierarchical organisation of behaviour. This paper describes the neuronal mechanisms underlying the modulation (prepulse inhibi­tion and fear potentiation) of the startle response in rats, and discusses the possible adaptive significance of these different phenomena of behavioural plasticity.

Hypotheses are presented on neural peculiarities of the monkey brain that distinguish monkeys from other mammals and man: - The unique’ feature of cortical tissue is that it can be applied serially. It fulfils some requirements of a powerful visual system. - The richness of visual as compared to other signals has yielded a basis for recognizing the bodily similarity of oneself to conspecifics. A visuo/motor coupling trained on oneself but applied to conspecifics (“aping”) evolved. - Long series of (generalized visuo/visual) computations feasible in a very large cortex would produce excessive delays that would not correspond to outer world delays. The new human solution to this is an “offline” system in which temporal relationships are described by excitation patterns. This gives rise to a rapid expansion of the cortex. - In humans, a “meaning” has to be attributed to the mutual (computational) relationships of excitation patterns in the offline system, by a reference to the corresponding relationships that would be valid in the normal (online) system. “Perception” is a corollary of this. - The offline treatment of time, together with “aping”, leads to new types of helping and other social interactions. - Monkeys may be compared to humans not using their offline system in the state of “absent- mindedness”. - Experimental approaches departing from excitation patterns are discussed

In this report we present a general introduction to object recognition. We begin with brief discussions of the terminology used in the object recognition literature and the psychophysi­ cal tasks that are used to investigate object recognition. We then discuss models of shape representation. We dispense with the idea that shape representations are like the 3-D models used in computer aided design and explore instead models of shape representation that are based on feature descriptions. As these descriptions encode only the features that are visible from a particular viewpoint, they are generally viewpoint-specific. We discuss various means of achieving viewpoint-invariant recognition using such descriptions, including reliance on diagnostic features visible from a wide range of viewpoints, storage of multiple descriptions for each object, and the use of transformation mechanisms. Finally, we discuss how differ­ences in viewpoint dependence that are often observed for within-category and between-category recognition tasks could be due to differences in the types of features that are natu­rally available to distinguish among different objects in these tasks.

In the early steps of visual information processing motion is one of the most important queues for the development of spatial representations. Obstacle detection and egomotion estimation are only two examples of the powerfulness of visual motion detection systems. The underlying process of information extraction has to be active due to the observer’s capabilities of egomotion. This means that the observer’s motion has an impact on the pro­jected retinal motion field. Therefore one of the challenging tasks for biological as well as for technical vision systems is to couple retinal motion and egomotion and to uncouple egomotion and object motion. The following sections describe a model that couples visual motion processing with the egomotion parameters of a moving observer. Beneath a theoreti­cal introduction of the model an application to traffic scene analysis is presented. A t last the paper relates the model to biological motion processing systems.

A simulated network for controlling a six-legged, insect-like walking system is proposed. The network contains internal recurrent connections, but important recurrent connections utilize the loop through the environment. This approach leads to a subnet for controlling the three joints of a leg during its swing which is arguably the simplest possible solution. The task for the stance subnet appears more difficult because the movements of a larger and varying number of joints (9 -18: three for each leg in stance) have to be controlled such that each leg contributes efficiently to support and propulsion and legs do not work at cross purposes. Already inherently non-linear, this task is further complicated by four factors: 1) the combination of legs in stance varies continuously, 2) during curve walking, legs must move at different speeds, 3) on compliant substrates, the speed of the individual leg may vary unpredictably, and 4) the geometry of the system may vary through growth and injury or due to non-rigid suspension of the joints. This task appears to require some kind of “motor intelligence”. We show that an extremely decentralized, simple controller, based on a combi­nation of negative and positive feedback at the joint level, copes with all these problems by exploiting the physical properties of the system.

In different disciplines such as philosophy of mind, dynamical systems theory, and connec­tionism the term ‘emergence’ has different jobs to perform. Therefore, various concepts of emergence are developed and examined. While weaker versions are compatible with prop­erty reductionism, stronger versions are not. Within philosophy of mind, particularly within the qualia debate there is a need for a strong notion of emergence, while in discussions of emergent properties of connectionist nets or of dynamical systems one can do with weaker notions of emergence.

Neuronal Activity, Emergence of Spatio-Temporal Patterns This paper explores if dynamic modulation of coherent firing serves cortical functions. We recorded neuronal activity in the frontal cortex of behaving monkeys and found that tempo­ ral coincidences of spikes firing of different neurons can emerge within a fraction of a second in relation to the animal behavior. The temporal patterns of the correlation could not be predicted from the modulations of the neurons firing rate and finally, the patterns of correla­ tion depend on the distance between neurons. These findings call for a revision of prevailing models of neural coding that solely rely on firing rates. The findings suggest that modification of neuronal interactions can serve as a mechanism by which neurons associate rapidly into a functional group in order to perform a specific computational task. Increased correlation between members of the groups, and decreased or negative correlation with others, enhance the ability to dissociate one group from concurrently activated competing groups. Such modu­ lation of neuronal interactions allows each neuron to become a member of several different groups and participate in different computational tasks.

One example of “emergence” is the development, as a result of neural ontogeny and living experience, of cortical networks capable of representing and retaining cognitive information. A large body of evidence from neuropsychology, electrophysiology and neuroimaging indi­cates that so-called working memory and long-term memory share the same neural substrate in the cerebral cortex. That substrate consists in a system of widespread, overlapping and hierarchically organized networks of cortical neurons. In this system, any neuron or group of neurons can be part of many networks, and thus many memories. Working memory is the temporary activation of one such network of long-term memory for the purpose of executing an action in the near future. The activation of the network may be brought about by stimuli that by virtue of prior experience are in some manner associated with the cognitive content of the network, including the response of the organism to those stimuli. The mechanisms by which the network stays activated are presumed to include the recurrent re-entry of impulses through associated neuronal assemblies of the network. Consistent with this notion is the following evidence: (1) working memory depends on the functional integrity of cortico-corti-cal connective loops; and (2) during working memory, remarkable similarities -including “attractor behavior” -have been observed between firing patterns in real cortex and in an artificial recurrent network.

Non-linear dynamical models of brain activity can describe the spontaneous emergence of large-scale coherent structures both in a temporal and spatial domain. We discuss a number of discrete time dynamical neuron models that illustrate some of the mechanisms involved. Of special interest is the phenomenon of spatio-temporal stochastic resonance in which co­herent structures emerge as a result of the interaction of the neuronal system with external noise at a given level punitive data. We then discuss the general role of stochastic noise in brain dynamics and how similar concepts can be studied in the context of networks of con­nected brains on the Internet.

In view of the enormous capacity and complexity of mammalian brains it is evident that a detailed account of their anatomy and physiology alone cannot lead to a complete under­ standing of their function. Computer simulation and mathematical analysis of abstract, yet biologically realistic models for neurons and networks yield additional and useful information about the interplay of the underlying anatomical structures, the physiological processes oper­ating on the neuronal substrate, and the resulting brain functions during the performance of behavioral tasks. We discuss some contributions of such experiment-guided theory to the issue of accurate spike synchronization.

Visual cortical neurons are broadly tuned to one or a few feature dimensions, like color and motion. This is advantageous because broadly tuned neurons can contribute to the repre­sentation of many visual scenes. However, if there are multiple objects in a visual scene, the cortex is at risk to combine features of different objects as if they belong to a single object. The term “binding problem” was introduced to refer to the difficulties that may occur in sorting out those responses that are evoked by a single perceptual object. The present article reviews proposals suggesting that the binding problem is solved by labelling an assembly of neurons that is responsive to a single perceptual obejct. Evidence is reviewed in favor of two possible assembly-labels: rate enhancement due to visual attention and neuronal synchrony. Assembly-labels should be spread through the cortical network to all neurons that have to participate in an assembly The present article tries to shed light on the mechanisms that subserve such a selective spread of assembly labels. Moreover, it is suggested that assembly labels may fulfill an equivalent role in the motor system, since binding problems can also occur during the generation of useful patterns of motor activity.

Gene Regulation, Brain Evolution, Neural Network, Empathy, Evolution of Altruism A network of gene regulation organized in a hierarchical and combinatorial manner is crucially involved in the development of the neural network, and has to be considered one of the main substrates of genetic change in its evolution. Though qualitative features may emerge by way of the accumulation of rather unspecific quantitative changes, it is reasonable to assume that at least in some cases specific combinations of regulatory parts of the genome initiated new directions of evolution, leading to novel capabilities of the brain. These notions are applied, in this paper, to the evolution of the capability of cognition-based human empa­thy. It is suggested that it has evolved as a secondary effect of the evolution of strategic thought. Development of strategies depends on abstract representations of one’s own pos­sible future states in one’s own brain to allow assessment of their emotional desirability, but also on the representation and emotional evaluation of possible states of others, allowing anticipation of their behaviour. This is best achieved if representations of others are con­nected to one’s own emotional centres in a manner similar to self-representations. For this reason, the evolution of the human brain is assumed to have established representations with such linkages. No group selection is involved, because the quality of strategic thought affects the fitness of the individual. A secondary effect of this linkage is that both the actual states and the future perspectives of others elicit vicarious emotions, which may contribute to the motivations of altruistic behaviour.

In this paper, I introduce the philosophical notion of strong emergence and argue that it is almost exclusively applied to properties related to (conscious) subjective experience. Con­trary to the still common attitude of refraining anxiously from these topics, I argue that we have a promising scientific approach to them: the cognitive neurosciences. I list a spectrum of interesting, potentially strongly emergent properties already investigated, and discuss the example of volition in more detail. Psychiatric disorders, like antisocial personality disorder, schizophrenia, and obsessive compulsive disorder, open new ways for understanding aspects of volition such as willed action, decision making, and agency. I conclude that the notion of strong emergence is only a preliminary label which, however, might be understood as a challenge for empirical scientists to explain and understand phenomena related to subjectivity and consciousness.