Past events

A major objective of neuroscience is to understand the neural code, namely how the patterns of neuronal signals (e.g. action potentials, membrane potential, calcium concentrations) “represent” physical objects, commands for actions, or psychological phenomena. An successful neural coding scheme has to be robust to noise (i.e. random neuronal activity). We have recently shown that using a small perturbation, an introduction of one “extra”-spike to the activity of a single neuron in the cortex, and studying the consequence of that perturbation we can obtain bounds on the level of noise in the cortex. Theoretical analysis of the data indicates that intrinsic, stimulus-independent variations in membrane potential of cortical neurons are on the order of 2.2–4.5 mV—variations that are pure noise, and so carry no information at all. Such level of noise places severe limitations on the plausibility of neural code based on precise spike timing. Using recent advances in optogentics we can extend the approach of introducing a precisely controlled perturbation. We explore how these perturbations affect the dynamics of activity in the cortex as well as theirs effect on animal performance on a task, to gain further bounds and insights on the neural code.

How do we make sense of sensory inputs? One important clue may be the central role of selective and predictive attention, by focusing limited resources on behaviorally relevant sensory channels and modulating information flow at multiple stages, to improve perception.Our approach is to study the effect of attention on information processing at the single neuron level in the primary auditory cortex (A1) of animals trained on multiple auditory tasks that require selective attention to task-specific salient spectral frequency or temporal cues. Our results demonstrate that when animals actively attend to a task, their auditory cortical neurons can rapidly change their spectrotemporal filter characteristics to improve the animal’s performance. Thus, cortical sensory filters are not fixed, but are highly adaptive, and show dynamic, task-specific transformations during auditory behavior. To study the broader neural circuits involved in attention, we have initiated research on several other components in the network, including secondary auditory cortical areas, nucleus basalis, and the prefrontal cortex (PFC), a brain area known to play a key role in attention and decision-making. In contrast to A1, PFC responses are largely independent of the acoustic properties of sound, and encode an abstract, categorical representation of sound meaning. Recent studies show that electrical stimulation of PFC can elicit receptive field transformations in A1 neurons very similar to the attentional effects observed during behavior. Our working model suggests a top-down instructive role for PFC, and emphasizes the importance of interactions between multiple brain areas during selective attention that lead to matched auditory cortical filters for attended acoustic stimuli, creating a dynamic, evolving neural representation of task-salient sounds and thus optimizing perception on a moment-to-moment basis.

A major goal in systems neuroscience is to characterize normal cortical dynamics. Numerous in vitro and in vivo studies demonstrated that ongoing cortical dynamics are characterized by cascades of activity across many spatial scales, termed neuronal avalanches. Avalanche dynamics are identified by¬ two measures (1) a power law in the size distribution of activity cascades, with an exponent of -3/2 and (2) a branching parameter of 1, which reflects a balance in the propagation of cortical activity at the border of premature termination and potential exponential blow up. Here we analyzed resting state brain activity recorded using MEG from more than 100 healthy human subjects. We identified discrete events in the MEG signal and segmented them into cascades, using multiple timescales. Cascade-size distributions were found to obey power laws. At the timescale where the branching parameter was close to the critical value of 1, the power law exponent was -3/2, in line with expectations for neuronal avalanches. This behavior was robust to scaling of the number of sensors and to coarse-graining the sensor resolution. As controls, phase-shuffled data with the same power spectrum or empty-scanner data did not exhibit neuronal avalanches. These results indicate that normal resting cortical dynamics are well described by a critical branching process. Both theory and experiments suggest that cortical networks with such critical, scale-free dynamics optimize various types of information processing. Neuronal avalanches could thus provide a biomarker for disorders in information processing, paving the way for novel quantification of normal and pathological cortical states.

This second talk will focus on olfactory circuits in the context of learning. I will present first a couple of non-associative phenomena, both linking plasticity and synchrony. I will then describe more recent work connecting STDP and reward signals, indicating that STDP rules are labile and influenced by the context in which they are being used. If time allows, I will present the outlines of the new work that we started at MPI Brain Research, on computation in an ancient cortex.