How is tactile information affected by parameters of the population such as non-uniform fiber sensitivity, innervation geometry and response variability?

Abstract

Analysis of population responses in the tactile system requires a step beyond the isomorphic representations that are commonly presented. Using a simple model based on our data for spheres contacting the fingerpad, we illustrate how the parameters of the population itself have a profound effect on the fidelity of neural representations or codes. The effects of these parameters, such as innervation density, variability of sensitivity, type and covariance of noise are not apparent from single unit responses and, at least at present, require a theoretical or modeling approach of some sort.

Introduction

Defining the response properties of isolated single neurons is only the first step in understanding how such neurons process information. The next step, which is essential, is an examination of the characteristics of groups or populations of neurons.

Many years ago, Mountcastle and others introduced the concept of ‘reciprocal interpretation’ in the somatosensory system. The essential principle is that if the responses of a single cutaneous neuron are plotted over the receptive field, then the resulting receptive field profile can also be viewed as the idealized population response for the neurons from that region of skin. This is still the most commonly used approach to population reconstructions. For primary afferent fibers, in many cases such ideal population responses are more or less isomorphic representations of the stimulus.

For example, when a sphere contacts the fingerpad, the receptive field profiles of the SAIs and hence, by ‘reciprocal interpretation’, the ideal population response (Fig. 1) reflects the shape, position and contact force of the stimulus [7]. A cylinder contacting the finger results in a cylinder-like response in the SAI population, which changes its orientation as the orientation of the cylinder changes [5]. LaMotte and colleagues [12], [13] have shown near isomorphic population responses for ellipsoids and wavy surfaces scanned over the skin, and Johnson and colleagues [15], [1] have shown isomorphic representations of letters and patterns of raised dots scanned over the finger. Isomorphic representations have also been demonstrated in the somatosensory cortex [4].

But real primary afferent fiber population responses differ substantially from these idealized images in at least three ways. Firstly, not all afferents are identical; in particular, sensitivities vary widely among fibers of the same type so that the images are distorted [11]. Secondly, noise of some form or another produces random fluctuations in the response patterns [8]. Thirdly, the innervation density is not infinite so that such images are not continuous but are sampled, in fact fairly sparsely and perhaps non-uniformly [9]. The question that we wish to address is: what impact do these factors have on the population response and on its ability to represent or encode the various features of the stimulus?