next up previous
Next: Orientation-interaction models providing Up: Discussion Previous: Limits of linear

Relevance of Connection Geometries to Functional Properties

An important issue in the understanding of the principles of RF formation concerns the relationships between the morphological structures of connections and their functional implications. To explain the origin of certain features of the RF profile, including orientation selectivity, various authors [79,84,85,97,123] postulated the existence of ``laterally directed'' inhibitory interactions which involve the spatial orientation and elongation of dendrites and axons. Considering long-range interactions, Sillito [102] has also speculated that ``a periodic pattern of inhibitory loci could contribute the predetermined template for all laterally directed inhibitory interactions in the cortex''. Accurate exploration of the morphological character of the intracortical interactions partially support this view by evidencing anisotropies in the axon fields [47,50,78,104,111,112], as well as axon patches running over relatively long distances from the soma [16,44,46,47,78,99]. From a functional point of view, the importance of the spatial organization of intracortical connections for the visual response properties of cortical cells has been proved by realistic modeling and large-scale simulations [41,89,126]. The impact of connection geometries on RF organization can be investigated exploiting the visuotopic organization of cortical areas. In general, a RF, can be thought as a local relational structure, by which a cell analyzes each point of the visual field in terms of how it is related to what is around it. Thanks to retinotopic mapping, asymmetry and anisotropy of axon and dendritic fields manifest themselves also in their visuotopic representations, influencing the relational structures of the cells exposed to the axon field [45,46,47,79,91] even though, as pointed out by [80,133] one must be aware of alternative possibilities such as irregularities in magnification factor. In this perspective, the RF of a cell subject to intracortical inhibition, can be interpreted as the spatial distribution of the signs of all the effects of inhibition on cortical plane. Its topography reflects, therefore, the whole spatial pattern of coupling and not only the direct neuroanatomical connectivity. In Section 3, we observed that the patterns of the resultant interaction fields after inhibition, depend on the spatial organization of the preexisting inhibitory fields. Specifically, inhibitory fields that are centralized around the recipient cell leads to induced couplings that are still concentrated around the same cell. Thus, in this case, their net effect is just a minor variation of the strength of the measurable inhibitory field. Conversely, when inhibition arises from spatially localized clusters, its effect extends over a larger area of cortical plane, involving, even with couplings of opposite sign, cells that are not engaged in direct reciprocal couplings.

A frequently asked question in simple cell RF modeling is how the orientation of a cell's axon field is related to the orientation of its RF. We observed that in our model, asymmetric extension of axon branches, extending for greater distances along one cortical axis than along the orthogonal one, makes the effect of lateral feedback even more specific, compelling recurrent action to express itself mainly along particular visuotopic axes. In the case of a distribution of connections lying along the same axis, with regular distance between clusters, recurrent inhibitory field may yield to a sort of ``spatial resonance'' (cf. sympathetic vibrations in mechanical waves) that leads to an amplification of inhibitory action and to an alternation of positive and negative induced couplings similar to a real oscillation in space. Stereotyped distribution of connections and the periodic and smooth variations of the orientation map are decisive to sustain spatial coherence and thereby to prime the oscillations of the induced couplings. Random arrangements of orientations, as well as irregularities in the anisotropic distribution of clusters and/or in distance among them, result, on the contrary, in a loss of coherence and therefore in the attenuation or even the suppression of the contributions of recurrence.


next up previous
Next: Orientation-interaction models providing Up: Discussion Previous: Limits of linear