Subthemes
During the second BMBF funding period, six sub-themes fostered intense scientific and methodological interactions and strongly interlinked the center’s Principial Research Projects: Projects concerning Invariant Representations, Population Codes and Multimodal Interactions addressed key concepts of neural computation and set the stage for understanding representations of space-time on the single-cell, network, and system’s level. Projects concerning Closed-Loop Technologies, Hearing & Neuroprostheses and Navigation forged methodological links across projects and covered important application aspects.
Invariant Representations in Space & Time
Organisms operate in constantly changing environments. To do so, stimulus attributes need to be extracted that are invariant with respect to natural stimulus variations, such as the identity of a speaker regardless of the distance and acoustic environment of the sound source or the speed of a moving object irrespective of changes in viewing conditions. In fact, the very notion of an “object” implies that high-dimensional spatio-temporal input data are reduced to a low-dimensional description reflecting functional invariances. Studying invariant stimulus representations in space and time is thus a core research focus in the Bernstein Center Munich. The proper transformations may be innate or result from learning processes, link space and time as well as different modalities and act differently on the single-neuron versus systems level. Investigating these different aspects using various model organisms as well as whole-brain studies in humans we foster tight interactions between the respective projects and stimulate the entire center.
Parallel Information Processing and Population Coding
To deal with the complexity of spatio-temporal stimuli, different stimulus attributes are often processed in parallel by neural populations. This raises the question how the computational load is distributed between the elementary units of larger processing modules, how the results of these computations are integrated across space and time to create consistent neural representations and how this is accomplished for dynamically varying input stimuli. The broad range of projects at the center allows us to compare population codes at various processing levels – from the sensory periphery and ascending pathways to central processing stages and (pre-)motor areas. Similarly, we are able to contrast coding principles in different auditory, visual, vestibular, somatosensory, and electrosensory model systems. Joint research about the role of feed-forward versus feedback processing also touches upon another aspect of the Center’s overarching research theme: how do neural systems synchronize the timing of internal information processing with the progression of time in the external world?
Multimodal Interactions
Combining vision, audition, olfaction, and touch allows the brain to form a unified representation of the outside world that goes beyond simply summing up the different sensory modalities separately. As different senses reach the brain with different external and internal delays, the nervous system must bring the different signals into spatio-temporal register. A whole range of multimodal interactions is the subject of study in the subtheme Multimodal Interactions, spanning both biological and man-made systems. Comparing the individual strategies to combine information from different modalities (including strategies to deal with the respective uncertainties and potential cross-modal inconsistencies) helps to reveal general principles underlying neuronal representations of space-time. These investigations profit substantially from the newly established Bernstein Virtual Reality Facility with its two parallel setups for rodent neurophysiology and human psychophysics.
Closed-Loop Technologies
Recent hard- and software advances make it possible to integrate rapid online data analysis, adaptive sampling techniques, and computational modeling into the design of neuroscience experiments. The broad range of such approaches at the Bernstein Center Munich – from dynamic clamp and iso-response methods to virtual reality (see also Bernstein Virtual Reality Facility) – makes it possible to discuss methodological issues and transfer expert knowledge between different levels of investigation. The primary goal is to improve the design of experiments, addressing issues such as optimal spatio/temporal averaging of data and the stability of experimental feedback loops, which, in turn, may lead to new insight about closed-loop situations in neurobiological and artificial systems, the interaction of visual processing and eye movements, and course control.
Auditory Processing and Neuroprostheses
We use dynamic clamp and iso-response techniques to study auditory coding. In parallel, we investigate visual influences on auditory processing. Using results from on auditory performance in reverberant environments (e.g., in rooms), these projects improve techniques for directly stimulating the auditory nerve and the design of coding strategies for cochlear implants and help elucidate age-related changes in spatial hearing. Advances in the development of methods and scientific integration are applied to understand biological mechanisms, which in turn support clinical investigations and technological innovations to be spearheaded by our industry partner MED-EL.
Navigation in Animals, Men and Machines
Navigation by biological and man-made systems relies on coding of space, object localization, spatial memory, and motion detection, and utilizes invariant stimulus representations and multimodal interaction for choosing the appropriate actions.
