Neurons couple up to make decisions


The use of state-of-the-art techniques to study neuronal activity during a navigational task involving sound stimuli broadens our understanding of how neuronal populations produce complex behaviours. See Letter p.92

Ever since neurophysiologists first recorded the activity of two or more neurons simultaneously, they have observed that neurons close to one another often show correlated firing1. This coupling reflects the fact that neurons are wired into densely connected circuits. But although the mechanism by which neuronal coupling arises is simple, its role in brain function has remained unclear. On page 92, Runyan et al.2 provide insight into how neuronal coupling contributes to decision-making processes. Their work also demonstrates how a thoughtful combination of emerging experimental methods can help us to make sense of a challenging problem.

Previous conclusions about the role of neuronal coupling in behaviour seem almost as variable as the number of studies exploring the issue. Patterns of correlation are often subtle, and they vary across brain areas, behavioural states and experimental methods. Some evidence indicates that coupling is a hallmark of essential cognitive processes, binding disparate features into a coherent representation of the world3. Other work points to coupling as a simple artefact of a typical aspect of animal studies — an anaesthetized or inattentive brain4. And, for the past decade, the field of neurophysiology has been dominated by the idea that neuronal coupling is an intrinsic form of 'noise' that interferes with accurate transmission of sensory information through the brain5.

Runyan et al. set out to determine the role of neuronal coupling in two regions of the brain — the auditory cortex, which encodes information about sound stimuli, and the posterior parietal cortex, which is involved in decision-making and spatial navigation. They placed mice in a virtual-reality maze, which the animals navigated using a sequence of auditory cues that indicated the direction they should take when they reached a branch (a noise from the left indicating the left fork, for instance). The authors measured the activity of many single neurons simultaneously during the task, one brain region at a time, using a technique called two-photon calcium imaging. This measures neuronal activity using a genetically engineered protein that fluoresces in response to calcium ions, the level of which increases in neurons as they fire (Fig. 1).

Figure 1: An experimental system for analysing neuronal coupling in mice during behaviour.

Runyan et al.2 studied genetically engineered mice whose cortical neurons fluoresced in response to an influx of calcium ions (Ca2+), which indicates neuronal activity. Mice were placed on a spherical treadmill and navigated a virtual-reality maze projected onto a screen in front of them. Sound stimuli from surrounding speakers indicated the path that the animals should take through the maze. The authors imaged fluorescent signals using a microscope to simultaneously measure the activity of many individual neurons during the task, with the aim of discovering how neuronal populations in two regions — the auditory cortex and posterior parietal cortex — encode information about the stimulus and the subsequent navigational decision.

Next, the researchers performed statistical analyses to assess what their data could tell them about how information on auditory stimuli and impending behavioural choices (the decision to turn left or right) is encoded in each region. They demonstrated that information about the stimulus itself is encoded almost exclusively in the auditory cortex. By contrast, neuronal activity in both brain areas encoded information about upcoming decisions.

Runyan et al. showed that, although individual neurons involved in decision-making were active for only short periods of time, neuronal populations collectively showed correlated activity across longer periods. This coupled activity provided additional information about decisions, above and beyond that encoded by the activity of individual neurons. However, the degree of population coupling depended strongly on brain area. Coupling in the auditory cortex was relatively weak and transient, whereas in the posterior parietal cortex it was strong and long-lasting. Importantly, neuronal coupling in the posterior parietal cortex provided information about an impending decision stably over several hundred milliseconds before the actual response took place. Coupling in this region was stronger during those trials in which the animals chose the correct branch of the maze, suggesting that coupled neuronal activity is causally related to robust behavioural performance.

Together, these results indicate that information about decision-making emerges and is maintained by sequential activation of connected subnetworks of neurons in the posterior parietal cortex. The authors' findings are consistent with a proposed model of working memory in primates6. More broadly, the work supports the classic concept of the 'synfire' chain7 — the idea that a coupled assembly of firing neurons can form a chain that maintains and transmits information through dense thickets of interconnected neurons.

Runyan and colleagues' study involved a combination of several innovative techniques. First, two-photon imaging in animals that were awake allowed the researchers to measure activity from a sufficiently large neuronal population to reliably observe coupling, which occurs sparsely in the cortex. Second, the virtual-reality environment they exposed mice to enabled a naturalistic exploratory behaviour that engaged the whole animal better than standard operant behaviours (in which animals learn to respond to a sensory stimulus to obtain rewards or to escape punishment), which typically minimize variability in sensory inputs or behavioural responses8,9. Third, the authors used a comprehensive computational model to account exhaustively for the many motor and behavioural variables that are part of complex behaviours, which can cause uncontrolled neuronal firing patterns that confound this type of study.

Although Runyan and co-workers' results are compelling, it remains to be proved whether coupling has a causal role in decision-making, or whether the phenomenon is generalizable across brain regions, species or behaviours. Readily available tools to manipulate neuronal activity and record subsequent activity changes in large neuronal populations provide a clear path to resolving these issues.

These results also open up new avenues of research. For instance, it will be interesting to identify the neuronal-circuit architectures that enable long-lasting coupling to emerge. It is also unclear whether the phenomenon is confined to the superficial cortical layers imaged in the current study, or to a particular neuronal population in the local cortical circuitry. Finally, how signals are transformed between the auditory and parietal cortex, and what other cortical areas are involved, will be a key focus of future research.

The authors' sophisticated experimental set-up introduces a new level of complexity to behavioural studies. Obtaining reliable and consistent animal behaviour is vital if any study is to be reproducible, but is challenging for even the simplest behaviours. State-of-the-art task design and analysis are therefore crucial to ensuring that any findings are real phenomena rather than idiosyncratic side effects of a seemingly unimportant task feature — an unintended consequence of, for example, the stimulus chosen, the animal's movements or the type of reward received. Runyan et al. have moved the field forward, establishing an experimental framework that takes these issues into account, and so promises deeper insight into neuronal population coding.Footnote 1


  1. 1.

    See all news & views


  1. 1

    Gerstein, G. L. & Perkel, D. H. Science 164, 828–830 (1969).

  2. 2

    Runyan, C. A., Piasini, E., Panzeri, S. & Harvey, C. D. Nature 548, 92–96 (2017).

  3. 3

    Womelsdorf, T. et al. Science 316, 1609–1612 (2007).

  4. 4

    Ecker, A. S. et al. Science 327, 584–587 (2010).

  5. 5

    Cohen, M. R. & Maunsell, J. H. R. Nature Neurosci. 12, 1594–1600 (2009).

  6. 6

    Goldman, M. S. Neuron 61, 621–634 (2009).

  7. 7

    Abeles, M. Local Cortical Circuits: An Electrophysiological Study (Springer, 1982).

  8. 8

    Moran, J. & Desimone, R. Science 229, 782–784 (1985).

  9. 9

    Fritz, J., Shamma, S., Elhilali, M. & Klein, D. Nature Neurosci. 6, 1216–1223 (2003).

Download references

Author information

Correspondence to Stephen V. David.

Related links

Related links

Related links in Nature Research

Neuroscience: Virtual reality explored

Neuroscience: The symphony of choice

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

David, S. Neurons couple up to make decisions. Nature 548, 35–36 (2017) doi:10.1038/nature23100

Download citation


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.