C5 Population dynamics in multi-stable perception.

Multi-stable stimuli can provide unique insights into visual processing, as changes in perception are decoupled from changes in the stimulus. Single-neuron studies have shown that higher in the visual hierarchy firing rates increasingly correlate with perception. However, very little is known about the underlying populationlevel mechanisms mediating the selection process and whether this process can be interpreted as the result of probabilistic inference when the posterior has more than one mode. We will address these questions on three fronts: a) using multi-electrode recordings in behaving macaques while presenting ambiguous motion stimuli, b) performing human psychophysics with the same stimuli, c) developing a dynamic states spiking neural network (DSNN) for probabilistic population coding to model the results obtained from electrophysiology and psychophysics. This integrative approach will provide novel
insights into multi-stable perception by combining population dynamics and probabilistic inference to uncover the mechanisms of multi-stability in the cortex.

An example of multi-stable stimuli in the motion domain. (a) A plaid consisting of three square-wave components each translating at a different constant velocity (red, green and blue arrows). (b) Velocity space representation of the stimulus. For some velocity combinations, the three constraint lines coincide, leading to a stable percept of coherent motion with velocity A. (c) However, other velocity combinations are inconsistent with a single coherent motion. Under these conditions, observers experience multi-stable motion transparency. At any point in time, two components tend to combine to form a plaid moving in one direction (A, B or C), while the remaining component appears to move in a different direction. The perceptual grouping of the components changes periodically with prolonged viewing. Stimulus variables such as the contrast, spatial frequency and component velocities can be used to systematically modify the relative duration of different interpretations.

Project leaders: Georgios A. Keliris, Wolfgang Rosenstiel, Roland Fleming, Nikos K. Logothetis
Coworkers: Martin Bogdan, Armin Walter, Qing Lin Li*, Hamed Bahmani
* funded by BMBF

Key publications:

  • Keliris GA, Logothetis NK and Tolias AS. The role of the primary visual cortex in perceptual suppression of salient visual stimuli. Journal of Neuroscience (in press).
  • Sheinberg DL and Logothetis NK (1997) The role of temporal cortical areas in perceptual organization. Proc Natl Acad Sci U S A 94:3408-3413.
  • El-laithy K, Bogdan M (2009) Synchrony state generation in recurrent networks with stochastic synapses. ICANN 2009, Part I, LNCS 5768:181-190.
Sponsored by the Federal Ministry of Education and Research