In the brain, the only constant is change. A neuron’s firing rate—the average number of spikes it produces per second—can fluctuate wildly depending on whether an animal is awake or asleep, how bright a light is, or simply because of "representational drift" over several weeks. If the brain relied solely on these fluctuating rates to represent the world, our perception would be hopelessly unstable.
Recent research suggests that the brain solves this problem by anchoring information in the precise temporal structure of neural activity. While the volume of neural activity may rise and fall, the relative timing between spikes remains remarkably consistent, providing a stable scaffold for sensory perception and memory.
The Invariance of Relative Timing
One of the most significant challenges for a neural code is intensity invariance. For example, an odor should be recognizable whether it is faint or strong. Schaefer et al. and Margrie propose that the olfactory system achieves this by using a "latency code." When an odor concentration increases, neurons fire earlier in absolute time. However, the relative timing between different neurons stays the same. By focusing on these relative latencies, the brain can identify a stimulus regardless of its intensity (Psychophysical properties of odor processing can be quantitatively described by relative action potential latency patterns in mitral and tufted cells).
This principle extends to vision. The authors of Computing Complex Visual Features with Retinal Spike Times demonstrate that the retina encodes complex features in the relative latency of the very first spikes sent to the brain. Because global changes in contrast shift all spike times simultaneously, the relative differences between them remain stable, allowing for robust feature detection across varying lighting conditions.
Stability Across Brain States
Temporal fidelity also explains how the brain maintains a consistent view of the world across different states of consciousness. It was long thought that the thalamus "gated" or blocked sensory information during sleep. However, Issa et al. and Wang found that the auditory cortex remains highly responsive during sleep. Crucially, the temporal patterns of these responses are nearly identical to those in the awake state, suggesting that the structural representation of sound is preserved even when we are unconscious (Sensory Responses during Sleep in Primate Primary and Secondary Auditory Cortex).
This stability is rooted in the way neural populations organize their activity. Luczak et al. and Barthó show that cortical activity is organized into "packets" of sequential firing. The internal ordering of neurons within these packets is preserved whether the brain is in a synchronized (sleep-like) or desynchronized (awake) state (Gating of Sensory Input by Spontaneous Cortical Activity). Similarly, Lee et al. demonstrate that the statistical "shape" of inter-spike intervals remains stable across wakefulness and sleep, reflecting intrinsic biophysical properties that transcend behavioral states (Computational Classification Approach to Profile Neuron Subtypes from Brain Activity Mapping Data).
The Temporal Scaffold Against Drift
Perhaps the most striking evidence for temporal fidelity is its resilience to time itself. Over weeks, the firing rates of individual neurons in the visual cortex can change significantly—a phenomenon known as representational drift. Yet, Zhu et al. demonstrate that while rates drift, the millisecond-scale timing of spikes remains highly consistent (Temporal coding carries more stable cortical visual representations than firing rate over time).
Xia et al. propose that this stability emerges because the population maintains a "temporal scaffold." While the magnitude of neural bursts varies (gain variability), their precise timing provides a reliable basis for encoding sequences (Stable representation of a naturalistic movie emerges from episodic activity with gain variability). This is echoed in the prefrontal cortex, where Li et al. found that stable sequential firing patterns—rather than trial-specific signals—are what allow rats to estimate time intervals consistently over long periods (Stable sequential dynamics in prefrontal cortex represents subjective estimation of time).
Mathematical and Hardware Robustness
The power of temporal coding is not just a biological observation but a mathematical necessity for robust systems. Etesami et al. and Trouleau show that certain statistical measures of event-driven data (cumulants) are mathematically invariant to random delays, allowing for the recovery of causal structures even in noisy environments (Cumulants of Hawkes Processes are Robust to Observation Noise).
In the realm of neuromorphic engineering, Arthur et al. and Boahen used these principles to build silicon chips that overcome hardware variability. By using plasticity to synchronize neural timing, they created systems where timing precision is independent of the input firing rate, ensuring consistent performance across different stimulus intensities (Learning in Silicon: Timing is Everything).
By prioritizing the "when" over the "how much," the brain creates a representational space that is robust to noise, state changes, and the passage of time.
Key Foundational Work
Sensory Responses during Sleep in Primate Primary and Secondary Auditory Cortex (2008) — contributed 2 concept elements to this direction.
Learning in Silicon: Timing is Everything (2005) — contributed 1 concept elements to this direction.