Summary: Even amid apparent noise and chaos, cortical neurons cooperate to communicate reliably, enabling the brain to extract order from variable activity.
Source: EPFL
Neurons communicate by sending brief electrical pulses called spikes. When an isolated neuron receives the identical input repeatedly, it can reproduce the same spike pattern very reliably. Yet recordings from living cortex show that neuronal activity often looks highly variable. Why this discrepancy?
Two main factors explain the difference between the high reliability seen in isolated neurons and the variable activity observed in intact cortical networks. First, synaptic transmission itself is inherently unreliable: when one neuron signals another, the release of chemical neurotransmitter at the synapse can fail unpredictably. As lead researcher Max Nolte notes, the probability that a chemical signal crosses a synapse between two cortical pyramidal neurons can be surprisingly low, sometimes on the order of 10%. This trial-to-trial uncertainty means each connected neuron effectively hears a different message each time.
Second, cortical networks combine excitatory and inhibitory neurons into recurrent circuits. Small fluctuations and random synaptic failures are amplified by these recurrent connections, producing rapidly diverging activity patterns — a hallmark of chaotic dynamics. In other words, tiny, random differences grow over time across the network, producing large variations in spike timing and firing rates.
These two sources — stochastic synaptic release and network chaos — help explain why single neurons in vivo appear so variable. One implication is that reliable information may not come from a lone neuron but from averaging across many neurons: the brain may read out collective activity, listening to the whole ensemble rather than a single cell.
Using detailed simulations to separate noise sources
Directly testing how specific noise sources shape cortical activity is difficult to impossible with current experimental tools. It is not feasible to monitor or manipulate all of the thousands of inputs to a single neuron in a living brain, nor to selectively switch individual noise sources on and off. To overcome those limits, the researchers turned to a biologically detailed digital reconstruction of rat neocortical microcircuitry developed by the Blue Brain Project (Cell, 2015). That large-scale, data-constrained model incorporates realistic synaptic dynamics, including the probabilistic nature of neurotransmitter release, and thus offered a unique platform to dissect how internal noise and recurrent dynamics interact.
Simulations with this detailed microcircuit revealed that spontaneous activity generated by the recurrent network is indeed highly noisy and chaotic: repeated runs with the same starting conditions produced very different spike times and firing patterns. As Nolte explains, unreliable neurotransmitter release is amplified by the recurrent dynamics, producing a rapidly decaying memory of prior activity and a sea of internal variability.
Reliable spike timing can emerge from chaos
Despite this intrinsic variability, the model also revealed a striking ability of cortical circuits to produce reliable responses when driven by external inputs. Spike timing that was inconsistent during spontaneous activity became consistently reproducible when the circuit received sensory-like inputs. Crucially, this shift to reliable millisecond-precision spiking did not require overwhelmingly strong drive: even weak feed-forward thalamocortical input was sufficient to transiently steer the network into a regime of high reliability.
In that brief reliable window, the same recurrent interactions that normally amplify uncertainty instead amplify reliability, allowing neurons to act together and produce precise output. As Blue Brain Project founder Prof. Henry Markram notes, thalamocortical stimuli can evoke spike times with millisecond precision within an otherwise noisy and chaotic network. The effect depends on neurons cooperating as a team — the collective dynamics of the circuit, not solely the properties of individual cells, determine when precise timing emerges.
The results reconcile a long-standing debate about whether cortex uses a rate code or spike-time code: the study shows that noisy, chaotic recurrent dynamics and stimulus-evoked, millisecond-precision spike timing are not mutually exclusive. Instead, internally generated variability coexists with the circuit’s capacity to produce precise, reliable responses when driven by appropriate inputs. This suggests that the large trial-to-trial fluctuations recorded in vivo can reflect an underlying capability to encode precise information, not simply dysfunctional noise.
Source:
EPFL
Media Contacts:
Kate Mullins – EPFL
Image Source:
The image is credited to Blue Brain Project / EPFL.
Original Research: Open access
“Cortical reliability amid noise and chaos”. Max Nolte, Michael W. Reimann, James G. King, Henry Markram & Eilif B. Muller.
Nature Communications. doi: 10.1038/s41467-019-11633-8
Abstract (summary)
The study quantifies internally generated variability in a biophysical model of rat neocortical microcircuitry that includes realistic noise sources. Stochastic neurotransmitter release emerges as a key contributor to internal variability, producing rapidly diverging, chaotic recurrent dynamics. Remarkably, the same recurrent network dynamics can transiently overcome this chaos in response to weak feed-forward thalamocortical inputs, yielding reliable spike times with millisecond precision. Thus, noisy and chaotic network dynamics coexist with stimulus-evoked, millisecond spike-time reliability, offering a unified explanation for variable spontaneous activity and precise stimulus-evoked responses.