GATSBY COMPUTATIONAL NEUROSCIENCE UNIT
UCL Logo

Neuronal variability: noise or incompetence?

Alex Pouget

Computational Cognitive Neuroscience Laboratory, University of Rochester, USA

Neuronal responses exhibit a large amount of variability, whether in spike counts or spike timing. This variability is thought to arise from two main sources: stochastic fluctuations, often called noise, and fluctuations in variables internal to the brain, such as the level of attention. We will argue that there is in fact a third source of variability, which we call incompetence. We use the term 'incompetence' to refer to deterministic computations that are suboptimal due to incomplete knowledge of the statistics of incoming spikes. This source is likely to be an important source of variability, if not the most important one, in the sense that it has a very large impact on behavioral performance. Incompetence also explains several puzzling results, and in particular the fact that variability in the brain depends on task difficulty and does not appear to be uniform throughout cortex.

 

BACK