Effect of stimulus onset on response variability

Although the source of response variability is not clearly understood, a common idea is that neural response variability originates from the internal activity of neural circuits and declines after stimulus onset [1-3]. By, assessing firing rate (FR) and between trial variability during Churchland et al. [2] showed that regardless of brain state during the task, stimulus onset leads to a significant decline in response variability in different areas of cortex such as V1, V4, MT, the lateral intra-parietal area (LIP), the parietal reach region (PRR), dorsal premotor cortex (PMd), and the orbitofrontal cortex (OFC). This decline can be observed even for non-responsive conditions. This phenomenon represents a general property of the cortex in which the state of the cortex stabilizes with input [2].

By evaluating the activity of the same data we observed that in some cases not only the response variability is not reduced after stimulus onset but it may stay constant or even increase during the evoked period. These results were in agreement with the Larry Abbot’s theoretical work that shows under specific conditions response variability could be unchanged after stimulus onset1. We studied the same datasets and noticed there is a minority of the population that show opposite behavior and be ignored during the averaging process. Here we want to assess the role of this population of neurons and to better understand their role in the neural circuits.

Fano Factor

The Fano factor (FF) is the ratio of the variance of mean firing rate (FR) across different trials to the mean, which is used to measure the across trial variability. We used this metric to quantify the alternations of variability over the course of time. To make the metric Independent from firing rate, we implemented the mean matching method introduced by Churchland et al. In this method, the FF is calculated only for the responses whose mean firing rate stays unchanged before and after stimulus onset.

In order to calculate FF and FR, we used the toolbox provided by Churchland [2]. The figure below illustrates the mean of FF and FR for MT and PMd datasets. In both data sets stimulus, onset appears 400 ms after baseline activity. Normal FF is shown in gray, and black curves show the mean-matched FF. The normal and the mean-matched FF are very similar in all plots, which means the increase in FR at stimulus onset does not affect the reliability of FF.

By analyzing neural data recorded from different layers of the V1 area, we observed that regardless of the type of stimulus, alternations in response variability depend on the neurons type and the layer in which the neurons are located. By classifying cells into simple and complex based on their response patterns, we observed that the neural variability in simple cells is not changed after stimulus onset, whereas in complex cells the FF declines after stimulus onset. In the second step, we classified neurons into inhibitory and excitatory based on the width of their action potential. The results showed a decline in response variability of inhibitory neurons after stimulus onset causes (figure 2).

We then classified cells into sink and source based on their local field potential (LFP) signal polarity to figure out in which layer the neurons are located [4]. We found that the response variability of source neurons is not affected by stimulus onset, whereas a decline in response variability of sink neurons can be observed after stimulus onset (figure 3). From these results, we inferred that decline in response variability after stimulus onset is related to the type of neurons or the layer in which the cells are located.

Supervisor: Dr. Reza Lashgari.