......................................................................

NJIT Mathematical Biology Seminar

Note special date:
Thursday, February 9, 2006, 4:00pm
Cullimore Hall 611
New Jersey Institute of Technology

......................................................................

Dynamical Characteristics Common to Neuronal Competition Models

Rodica Curtu

Transilvania University of Brasov
Romania

Reciprocal inhibition arhitecture is widely used to describe binocular rivalry, and bistable perception in general. Binocular rivalry occurs when an observer is presented with one image to the right eye and simultaneously a different image to the left eye. The stimulus to the visual system becomes ambiguous and as a consequence an alternation between the two monocular images (or the corresponding percepts) is reported. Moreover, increasing the stimulus strength in both eyes leads to an increase of the alternation frequency, that is to decrease of the mean dominance duration of the two percepts. This property is known as Levelt's Proposition I (LP-I; Levelt 1968) and it is proved by many psychophysics experiments. We investigate two models for binocular rivalry (Wilson 2003, Laing & Chow 2002), and some versions of them, to determine the 'ingredients' necessary to produce the alternation between competing interpretations of the stimuli. In particular, we are interested in examining the validity of LP-I in different regions of the parameter space. We check for qualitative similarities and differencies in dynamics due, for example, to the choice of the slow negative feedback process, like synaptic depression, or subtractive or divisive spike frequency adaptation. We find that all these neuronal competition models show five distinct regimes of behavior: as the stimulus strength decreases, the systems' dynamics change from (I) a single stable high activity level for both units, to: (II) oscillations that obey LP-I; then (III) entering a winner-take-all regime (one representative unit is up forever while the other one is suppressed); then (IV) oscillations which now yield an increasing curve of the dependence of dominance periods on the stimulus strength, contrary to the LP-I statement; and (V) again a single stable state for both units but at a low level. These dynamics seem to be a direct consequence of the hypothesized mutual inhibition between the neural representations of the two percepts of the stimulus. An increase in the mutual inhibition leads to significant deviations from LP-I at small values of input. The result may suggest the ability of the neuronal system to maintain the cross-inhibition strength to a specific level. Alternatively, it may point to the inadequacy of the existing models to describe rivalry behaviors.