|
The interconnection of the nerve cells is not arbitrary. There is irresistible evidence from about all known studies in biopsychology on subjects, whether through non-invasive visualising methods of investigation like positron emission tomography (PET) and functional magnetic resonance tomography (fMRT) or through findings about behavioural deficits as a result from surgical operations and trials with animals, that cognition is established by the interplay of distinctive neural network structures which, depending on the functions considered, can be associated to certain brain regions. The well-known picture depicting the distribution of neurons across the neocortex in differing configurations, already gives an indication, although a very coarse one, on the fact that the different regions of the neocortex are associated to different cognitive functions. We can distinguish between functional modules for perceptual processing, functional subsystems for the generation of voluntary actions, and sensori-motor processing. As an example we will briefly give some insight into the function of perceptual processing of the visual modality in the following. A significant part of the necortex is devoted to visual perceptual processing. Once the visual receptor information enters neural information processing, various functionally specialised areas in the brain are involved at the different levels of the hierarchical functional structure of perceptual processing.
Each of these specialised functional assemblies at the different hierarchical levels plays a certain role. It appears that the human perceptual processing is mainly organised as a oneway pathway for sensory inputs from the receptors to the association cortex via the thalamus, the primary sensory cortex, and the secondary sensory cortex. Descending interconnections are much less in numbers, although important, too. Also the interconnections between functional assemblies within each hierarchy level of processing are of great importance for the process of perception. It is the posterior region of the occipital lobe where the primary visual cortex is located. The secondary visual cortex lies in the adjacent areas of the inferior temporal lobes and the anterior region of the occipital lobe surrounding the primary visual cortex, the so-called prestriate cortex. The main stream of visual information goes along the pathway from the primary visual cortex to the areas of the secondary visual cortex, and from there to those parts of the association cortex which receive visual inputs. The largest single area of of association cortex regarding visual inputs is located in the posterior parietal cortex. It signifies the so-called dorsal visual stream which stands for the perception of where the perceived object is located in the environment. The association cortex as part of the inferotemporal cortex is also known to receive visual input. It signifies the so-called ventral visual stream which stands for the perception of what object is perceived. It is the activation of these areas when the human is having conscious experiences about what he/she sees, and it is on these pathways where the combined activity of different interconnected cortical areas, the so-called binding takes place.
Because of the parallel processing of stimuli there have to be certain mechanisms in order to selectively combine responses on elementary features into neural activation patterns at higher processing levels which represent feature formations, e.g. objects. The synaptical setting (unconscious processing knowledge) along these pathways determines what is recognised as objects and what is not. The number of features to account for has to be sufficient to avoid confusion. The features will be increasingly fine-grained the more knowledge is accumulated about possible objects to be recognised. There is experimental evidence that neurons not only report on the elementary feature they encode but that they also indicate with which other neurons they are currently related for the purpose of representing the object the feature belongs to and. These findings suggest that the participating neurons are temporally synchronised in concert with mechanisms like:
• inhibiting and excluding unrelated responses from further processing and
• enhancing of discharge frequency of the seleted responses.
These mechanisms, which are crucial for the perception process as well as others of the total process of cognition, are amazingly efficient and by far better than any technical device we know of so far. They matter not only for the perception of tangible objects, but also for representations of abstract entities like for instance feelings and intended actions. This is done in a way exhibiting an amazing combination of tremendous generalisation capability and equally high perception reliability. This makes the human perceptual processing a highly exceptional one. It can reliably spell out when two identical objects are recognised, or what the dissimilarities are if there are two objects which are not identical. In addition, it can identify, if objects are semantically identical, although being of quite different appearance. Not only neural structures of the sensory cortex in different modalities might be involved but also other brain structures and corresponding functions, including the attention control. Certainly, the process will be somewhat different for recognising known objects, at least when thinking about what we consciously experience as individual features of the object, as opposed to the perception of unknown objects.
One of the fundamental outcomes of the binding process is the way in which we automatically seggregate the viewed scenery into the so-called figure (one or more objects of predominant concern) as a result of the binding process which we try to take into central focus of fovea vision, and everything else, the background. In that respect, the extremely small central area on the retina of high-resolution vision in combination with the ability of fast eye movement control helps a lot to keep the data stream in a manageable size. The figure has been called the Gestalt by the community of researchers who worked on this phenomenon. They proposed a number of simple laws of perceptual organisation. Although there is also a lot of criticism about the Gestalt laws they help a lot to make good use in the design process of display presentations in work systems. This research also stimulated the finding that complex stimuli are normally perceived as integrated wholes, not as combinations of independent attributes. That means that the overall Gestalt may be perceived before the parts of the object stimulus which make up the whole or that the whole is identified consciously before the individual features are. There has not been found an area in the cortex, though, which receives all partial sensory information in order to integrate it to the complete figure what refers to the binding problem as mentioned earlier. The neuropsychological explanation for that phenomenon, as it has been evidenced so far to some degree, is that various neural areas are working in parallel at different locations in the cortex on the perception of an object. They separately analyse the features of the object, thereby forming a so-called assembly for that object which is bound by a certain common activation pattern of all neurons involved. Thus, it is this assembly which represents the object considered. This is a great concept, exploiting the given multitude of neural interconnections in a clever way in order to be able to be responsive to the tremendous variety of objects encountered in real life and to warrant efficient processing. Artificial cognition might make use of this concept in future developments.
This conception is not confined to the visual system. Also the other sensory systems as well as object perception based on parallel sensing through more than one modality show similar phenomena. Finally, referring to the thalamus (lateral geniculate nuclei), it represents a kind of relay for all information on its way into the cortex. Obviously, many other brain structures project into the thalamus through more than 80% of its afferences. Presumably, thereby the information is selectively modulated before it enters the cortex. The thalamus is also involved in the process of attention control, and it decides which of the information coming from sensing organs and other parts of the body is important enough to be forwarded to respective cortical areas. This kind of function of the thalamus, by the way, is true for all senses, except the olfactory one. In many work systems human perception is confined to a great extent to information gathering from displays. This information is provided by technical sensing devices and communication links which take care of all information needed about the environment. Artificial cognitive systems as opposed to humans might even manage with no sensing at all, if there are appropriate communication links to other agents which have got the sensing capability needed. Nevertheless, it is unquestionable that future artificial cognitive systems in work systems will most likely not do without sensing. Great progress has already been made in the field of artificial perception, also making use of the findings about human perception. Just to mention one invention of nature regarding visual perception, for instance, which must not be missing in work systems for computational efficiency reasons, that is dynamic camera alignment control in analogy to human eye movements.