Neuronal variability: noise or part of the signal?

• Published in: Nature reviews. Neuroscience Vol. 6; no. 5; pp. 389 - 397
• Main Authors: Stein, Richard B., Gossen, E. Roderich, Jones, Kelvin E.
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 01.05.2005; Nature Publishing Group
• Subjects: Action Potentials - physiology; Animal Genetics and Genomics; Animals; Behavioral Sciences; Biological and medical sciences; Biological Techniques; Biomedical and Life Sciences; Biomedicine; Central nervous system; Electricity; Electrophysiology; Fundamental and applied biological sciences. Psychology; General aspects. Models. Methods; Humans; Movement - physiology; Nerve Net - cytology; Nerve Net - physiology; Nervous system; Neurobiology; Neurons; Neurons - cytology; Neurons - physiology; Neurosciences; review-article; Vertebrates: nervous system and sense organs; Spike; Sensory neuron; Motor neuron; Review; Noise; Variability
• ISSN: 1471-003X; 1471-0048; 1471-0048; 1469-3178
• DOI: 10.1038/nrn1668

Key Points Traditionally, the rate of nerve impulses (spikes) over time was considered to be the main carrier of information in the nervous system. Therefore, any variability in the rate of response to a steady stimulus would reduce the information conveyed by a nerve cell. Many nerve cells fire with considerable variability, which would limit their ability to carry information to 2 or 3 bits in 1 s. With time-varying inputs containing the range of frequencies that the neuron responds to, values of information transmission of approximately 1 bit per spike have been calculated. For a neuron that fires tens or hundreds of spikes per second, much higher bit rates are possible than with steady inputs. Variability might also offer distinct advantages in preventing the entrainment of neurons to high-frequency signals. Enhanced sensitivity to weak signals has been proposed, which is known as 'stochastic resonance', as well as a role of variability in the method of Bayesian inference. Recent work on various sensory systems has emphasized the importance of timing, particularly that of first spikes, rather than the rate of firing over time. Rate coding might be more important in the motor system than precise timing. The variability in rate fluctuates with the mean rate (signal-dependent noise). The variability in the motor output in the presence of this noise can be minimized using optimal control theory. Optimal control theory predicts the form of many movements if a specific rule is assumed that relates the standard deviation in rate to the mean rate. This rule is not observed experimentally for either motor neurons or the motor cortex. However, the relationship between the standard deviation in muscle force and the mean force obeys the rule. The reason for the difference between the neural responses and the force output arises from the Henneman size principle. This states that the first recruited motor units are small and, hence, produce minor variations in force. Later motor units are larger and produce greater variations with the magnitude required by the optimal control theory. In the central nervous system, large excitatory postsynaptic potentials (EPSPs) can cause the near synchronous firing of groups of cells that might be important in attention, as well as learning and memory. Interactions in some areas, such as the hippocampus, between ongoing oscillations and spike activity might be used by 'place neurons' to locate the position of the body in external space. Therefore, variability in the firing rate of individual neurons is not simply noise, but might have a range of functions in neurons throughout the nervous system. Sensory, motor and cortical neurons fire impulses or spikes at a regular, but slowly declining, rate in response to a constant current stimulus. Yet, the intervals between spikes often vary randomly during behaviour. Is this variation an unavoidable effect of generating spikes by sensory or synaptic processes ('neural noise') or is it an important part of the 'signal' that is transmitted to other neurons? Here, we mainly discuss this question in relation to sensory and motor processes, as the signals are best identified in such systems, although we also touch on central processes.