ABSTRACT
Synchronized signals are functionally important because they can activate a neurone more effectively than an uncorrelated input, especially if the input’s correlation is over a time span not broader than the temporal window of the neurone’s integration. In active cortices, integration times can be in the range of milliseconds, so that precisely synchronized signals seem important for cortical processing. We analysed such high frequency activities in visual cortex of cats and monkeys by multiple microelectrode recordings. Transiently changing scene segments were represented by synchronized population activities that were often non-rhythmic and phase-coupled to visual stimuli. During sustained activation, as in periods of stable retinal input, activities were more oscillatory (30-100 Hz) and dominated by cortical dynamics. Such oscillations were synchronized within a vertical column across cortical layers, among different columns of the same cortical area within some millimeters (“synchronization field”), and between different cortical areas amongst neurones with neighbouring receptive fields. Oscillation frequencies were highly variable, while average phase-delays, even in separate locations, were narrowly distributed around zero. Experimental data and related computer simulations support a comparably simple explanation: Fast oscillations are generated during states of sustained activation in local populations via local feedback inhibition, while distributed populations are synchronized via mutual facilitatory coupling. Synchronization at near-zero phase can be explained by common input from shared cortical and subcortical sources. The width of cortical synchronization fields with oscillations at zero-phase delay can be explained by temporal properties of cortical circuits shaped by Hebbian learning. Since the synchronization of fast cortical oscillations depends on specific grouping of visual features, its role in scene segmentation, object definition, and other more general association processes are discussed. Finally, it is argued how basic visual processing operations might be carried
out in a stepwise fashion by a modular circuit derived from interactions among simple and complex cells in primary visual cortex. KEYWORDS: Cortex; Gamma oscillations; Cortico-cortical interactions; Population coding; Perceptual grouping; Association field
1. INTRODUCTION
1.1. Synchronized Input Excites Neurones Strongly and Precisely
1.1.1. Definition of “Integration time” and “Synchronization” in the present context
The term synchronization will be used with respect to the duration of a single neurone’s integration time, which is mainly given by the time course of postsynaptic influences in response to an input spike. To a first approximation we might think of the integration time as the half-height duration of an excitatory postsynaptic potential (EPSP) at the neurone’s spike encoder. In cortical neurones these durations span a broad range from about 2 up to 100 milliseconds. Integration times depend on the site and type of synapses, on the different dendro-somatic membranes in different types of neurones, but also on dynamic changes in membrane time constants of single neurones (Häusser and Roth, 1997). One example of variability in integration time is the prolongation of EPSPs from NMDA synapses with increasing levels of membrane depolarization (Fox and Daw, 1992), while another example is the shortening of PSPs with increase in a neurone’s average synaptic activation (Agmon-Snir and Segev, 1993; Nelson, 1994). In the activated awake cortex in which we are interested in this paper dominant integration times are estimated in the range 2 to 10 ms (e.g. König et al., 1995a), where the shorter time constants will be found in “low impedance locations” with several neighbouring synapses activated in parallel, while longer ones will occur at “high impedance locations” when only a single synapse is activated. Neural signal components covarying, with the same sign, during such integration windows will be called here “synchronized”. Synchronized events may occur singly, or repetitively with a more stochastic or rhythmic character. With respect to the processing of behavioural output, this synchronization range of 2 to 20 ms is rather short and we will therefore use the term “fast” for it. Accordingly, neural signals which are correlated positively in this range will be called “fast synchronized”. In this sense, rhythmic signals with half cycle durations of 5 to 10 ms (100 to 20 Hz) will be termed synchronized fast cortical oscillations (FCOs).
