Neurons are noisy. Even when we view successive presentations of the same stimulus, the responses of a given neuron will vary. Moreover, this noise is correlated between neurons. Because correlated variability cannot be removed by averaging over the population, it can distort population responses to different stimuli, reducing the certainty of what stimulus occurred at any given moment. Researchers typically discard this variability by averaging the responses of neurons to many stimulus presentations. Trial averaging is part of the methodological backbone of attention experiments in both monkey electrophysiology and human neuroimaging. However, our brains do not have this luxury. From one moment to the next, we make sense of our environments in real time. How then does the brain compensate for neuronal variability?
Schmitz and Duncan answer this question by reviewing findings across several different lines of evidence in non-human animals, including multi-electrode electrophysiology and optogenetics research, and then integrating these findings into a unified model of attention. They show that in addition to increasing average firing rates of individual neurons across trials, attention alters the variability of individual neurons’ responses and their shared variability. This is accomplished by the synergistic effects between a neurochemical—acetylcholine—and a specific cortical circuit, which together act to improve how information is coded by populations of neurons.
However, the invasive recording methods used to advance these discoveries are not feasible in humans. Consequently, they argue that the field of attention research has divided into separate largely non-overlapping communities. This is a big problem. Current non-invasive probes of mass population activity in humans, such as functional magnetic resonance imaging (fMRI), almost exclusively rely on averaging methods. Large-scale effects of attention on population noise are therefore neglected entirely. By proposing a unified computational and neurobiological account of attention, this paper in turn lays the groundwork for re-uniting human and non-human animal research on attention.