Information Jitter Derivative Method: A Novel Approach to the Analysis of Multiplexed Neural Codes

How to mathematically separate out the different components of a neural code and to identify the unique contribution of each of these components to sensory coding and behavior is an open question in neuroscience. Here we present a novel approach to decompose the information encoded in the temporal structure of a spike train into the unique, complementary information contained in its different temporal scale components. We do this by analytically inferring the derivative of the information with respect to the precision with which the neural activity is measured. We demonstrate that the negative of this derivative represents the non-redundant information carried by each temporal scale and therefore constitutes an exact breakdown of the total information. The proposed approach, which we called Information Jitter Derivative (IJD) method, uses a jitter procedure to manipulate the precision of the neural activity and allows to precisely identifying the relevant timescales in the encoding of the stimulus information. We validated the IJD method on simulated and real data. In particular, we show that the IJD is able to uncover the different strategies used by the retinal ganglion cells of the axolotl salamander to encode information about different visual features. Importantly, we found that coarse and fine spatial features are encoded into different temporal scales. The Information Jitter Derivative method thus provides a way of studying in detail the information processing capabilities of a multiplexed neural code by breaking down the temporal information contained in the neural activity into its unique, complementary temporal scale components.