The idea that noise has a detrimental effect on signal detection is a commonly held view; as noise increases, the signal-to-noise ratio (SNR) is expected to decrease. But in some systems, noise can actually enhance signal detection, an effect known as stochastic resonance. In these systems, SNR is not a linear function of noise; instead, gradual increases in noise are initially associated with a steep and coincident increase in SNR, before signal detection begins to degrade. Do neurons show stochastic resonance? In other words, can increases in noise lead to an enhanced detection of synaptic inputs? Stacey and Durand have recently provided evidence in support of this intriguing idea.

Using a computer model of hippocampal neurons, the authors have previously shown that the external introduction of physiological levels of noise improve signal detection. They have now gone on to show, in hippocampal slices, that endogenous noise can have a similar effect. Stacey and Durand applied subthreshold currents to the CA3 hippocampal region, a manipulation that enhanced random synaptic activity and increased noise levels recorded from individual CA1 neurons. The authors simultaneously delivered test pulses on an independent pathway to evoke subthreshold synaptic potentials in CA1 neurons, and observed that the triggering of action potentials by the test pulses significantly increased when the noise source was on. So, SNR increased steeply as a function of noise, roughly following the predictions of the stochastic-resonance model — an initial increase and a subsequent reduction of SNR. However, levels of noise high enough to start degrading the signal could not be induced because of experimental constraints, and the predicted decrease in SNR was only observed in a computer simulation. But despite this limitation, the observations of Stacey and Durand indicate that noise can enhance the detection of synaptic activity in the hippocampus. As the random activity used by the authors as the source of noise falls within physiological levels, stochastic resonance might indeed favour the detection of weak or distal synaptic inputs.