Emerging principles of population coding: in search for the neural code

- Correlations re-distribute the noise in the neural population-response.
- Information is limited when signal and correlated noise overlap.
- Heterogeneity and single-cell variability provide a way to overcome correlated noise.
- Additional constraints are required to further test hypotheses for the neural code.
- Theory should be developed to study the transmission of information.

Population coding theory aims to provide quantitative tests for hypotheses concerning the neural code. Over the last two decades theory has focused on analyzing the ways in which various parameters that characterize neuronal responses to external stimuli affect the information content of these responses. This article reviews and provides an intuitive explanation for the major effects of noise correlations and neuronal heterogeneity, and discusses their implications for our ability to investigate the neural code. It is argued that to test neural code hypotheses further, additional constraints are required, including relating trial-to-trial variation in neuronal population responses to behavioral decisions and specifying how information is decoded by downstream networks.