Perception of tactile vibrations and a putative neuronal code

We devised a delayed comparison task, appropriate for human and rats, in which subjects discriminate between pairs of vibration delivered either to their whiskers, in rats, or fingertips, in humans, with a delay inserted between the two stimuli. Stimuli were composed of a random time series of velocity values (“noise”) taken from a Gaussian distribution with 0 mean and standard deviation referred to as σ1 for the first stimulus and σ2 for the second stimulus. The subject must select a response depending on the two vibrations’ relative standard deviations, σ1>σ2 or σ1<σ2. In the standard condition, the base and comparison stimuli both had duration of 400 ms and they were separated by a 800 ms pause. In this condition, humans had better performance than did rats on average, yet the best rats were better than the worst humans. To learn how signals are integrated over time, we varied the duration of the second stimulus. In rats, the performance was progressively improved when the comparison stimulus duration increased from 200 to 400 and then to 600 ms. In humans, the effect of comparison stimulus duration was different: an increase in duration did not improve their performance but biased their choice. Stimuli of longer duration were perceived as having a larger value of σ. We employed a novel psychophysical reverse correlation method to find out which kinematic features of the stochastic stimulus influenced the choices of the subjects. This analysis revealed that rats rely principally on features related to velocity and speed values normalized by stimulus duration – that is, the rate of velocity and speed features per unit time. In contrast, while human subjects used velocity- and speed-related features, they tended to be influenced by the summated values of those features over time. The summation strategy in humans versus the rate strategy in rats accounts for both (i) the lack of improvement in humans for greater stimulus durations and (ii) the bias by which they judged longer stimuli as having a greater value of σ. Next, we focused on the capacity of rats to accomplish a task of parametric working memory, a capacity until now not found in rodents. For delays between the base and comparison stimuli of up to 6-10 seconds, humans and rats showed similar performance. However when the difference in σ was small, the rats’ performance began to decay over long inter-stimulus delays more markedly than did the humans’ performance. The next chapter reports the analyses of the activity of barrel cortex neurons during the vibration comparison task. 35% of sampled neuron clusters showed a significant change in firing rate as σ varied, and the change was positive in every case – the slope of firing rate versus σ was positive. We used methods related to signal detection theory to estimate the behavioral performance that could be supported by single neuron clusters and found that the resulting “neurometric” curve was much less steep performance than the psychometric curve (the performance of the whole rat). This led to the notion that stimuli are encoded by larger populations. A general linear model (GLM) that combined multiple simultaneously recorded 2 clusters performed much better than single clusters and began to approach animal performance. We conclude that a potential code for the stimulus is the variation in firing rate according to σ, distributed across large populations.In conclusion, this thesis characterizes the perceptual capacities of humans and rats in a novel working memory task. Both humans and rats can extract the statistical structure of a “noisy” tactile vibration, but seem to integrate signals by different operations. A major finding is that rats are endowed with a capacity to hold stimulus parameters in working memory with a proficiency that, until now, could be ascribed only to primates. The statistical properties of the stimulus appear to be encoded by a distributed population.