The Department of Neurophysiology was established in 1982.

Research in the department concentrates on the identification of neuronal processes underlying cognitive functions and is focused on the analysis of the cerebral cortex. Most of the studies are performed in the visual system but other modalities are included for the examination of polysensory integration.

Evidence indicates that cortical processes are distributed, with different aspects of sensory objects being evaluated in parallel in different cortical areas. How these distributed operations are coordinated and how the results of parallel computations are bound together in order to give rise to coherent percepts and coordinated motor acts is still largely unresolved. A major part of the projects pursued in the department is therefore devoted to this binding problem. The central question is how the nervous system encodes relations among distributed responses.

In principle, there are two non-exclusive possibilities to evaluate relations among neuronal activity patterns. One is to connect neurons whose responses are to be bound with a common target cell. By adjusting the coupling strength of the connections the target cell can be made to respond only if a certain set of input cells is active. Hence, the target cell evaluates relations among afferent activity patterns and if it responds it signals the presence of a particular conjunction of input activities. In sensory systems the input cells are tuned to the various elementary features of perceptual objects such as contour borders and hue in vision or tone frequency in audition. By appropriate arrangement of the input connection to the higher order target cells, conjunction specific neurons can be generated whose responses signal the specific constellations of features characterising a perceptual object. This strategy to bind responses by convergence and to represent relations by conjunction specific units permits rapid processing because it can be realised in simple feed-forward architectures and it is robust because a particular cell signals always the same content.

The complementary strategy for the encoding of relations is the formation of cooperatively coupled cell assemblies. Here, neurons responding to the components of perceptual objects temporarily group themselves into functionally coherent assemblies which as a whole signal the presence of the feature constellation characterising a perceptual object. The advantage of this strategy is that a particular neuron can participate in the representation of many different contents because it can be regrouped dynamically with other neurons in ever changing constellations. This economises the number of neurons required for the representation of the immense variety of perceptual objects and permits representation of novel contents by dynamic grouping of cells into novel assemblies. However, this strategy of distributed coding requires a mechanism that tags responses as related once they have become part of the same assembly and assures that the grouped responses are processed conjointly and are not confounded with unrelated responses of other assemblies.

Following the discovery that neurons in the visual cortex can engage in precisely synchronised oscillatory firing patterns we began to pursue the hypothesis that response synchronisation could be exploited by the brain to dynamically group responses into assemblies and to tag them as related. Precisely synchronised neuronal discharges are more effective in driving cells in target networks than temporally dispersed activities. Therefore, synchronisation could be used to raise conjointly and with high temporal resolution the saliency of selected subsets of neuronal responses. This is equivalent with tagging them as related as it assures with high selectivity joint processing of the synchronised responses.

In order to examine the predictions derived from this hypothesis we record simultaneously the activity of large numbers of neurons and study the precise temporal relations among their responses. The goal is to establish correlations between synchronisation phenomena and those cognitive processes that are thought to involve dynamic selection and grouping of responses. This is the case for certain steps in pattern analysis, attention based stimulus selection, grouping operations in short-term memory, polysensory integration and probably also sensory-motor coordination.

Depending on the complexity of the cognitive functions to be studied, we record with multielectrode arrays either from anaesthetised rodents and cats or from awake, behaviourally trained non-human primates. In the former case we apply in addition optical recording methods to obtain a comprehensive overview of the functional organisation of the studied cortical area and to guide the placement of the multiple recording electrodes. In the latter case, information about the neuronal networks involved in a particular task is obtained with functional magnetic resonance imaging (fMRI) prior to the implantation of the recording electrodes. These animal experiments are paralleled by studies in human subjects, applying the non-invasive techniques of electroencephalography (EEG) and fMRI. Whenever possible, the same cognitive functions are investigated as in the animal experiments in order to facilitate comparisons.

If assemblies are defined by transient synchronisation, the association connections which link distributed cortical neurons must have a synchronising effect. In addition, their synapses should be susceptible to use-dependent modifications to permit consolidation of new assemblies by learning whereby the polarity of these synaptic gain changes should depend on the precise temporal relations between the firing patterns of the connected cells. In order to examine these predictions we perform in vitro experiments on slices of the visual cortex of rodents, combining patch-clamp recordings from visually identified neurons with multielectrode recordings and microstimulation.

The search for temporal patterns and correlations in high dimensional time series such as multi-cell recordings poses challenging theoretical and practical problems. Therefore, increasing resources are devoted to the development of new methods for the recording, acquisition and evaluation of data and for simulation studies on computational models.