Phase-of-firing code

Anna Cattani, Gaute T. Einevoll, Stefano PanzeriarXiv 2015 · DOI · arXiv

Abstract

The phase-of-firing code is a neural coding scheme whereby neurons encode information using the time at which they fire spikes within a cycle of the ongoing oscillatory pattern of network activity. This coding scheme may allow neurons to use their temporal pattern of spikes to encode information that is not encoded in their firing rate.

Introduction

For decades, the "rate code" hypothesis dominated neuroscience, suggesting that information is primarily transmitted through the frequency of action potentials (spikes). However, rate coding faces significant limitations, such as struggling with scale invariance and failing to capture the full complexity of sensory experiences. The phase-of-firing code offers a compelling alternative, revealing that when a neuron fires relative to background brain oscillations carries substantial hidden information. This multiplexed coding scheme allows the brain to represent complex variables—like spatial heading or natural sounds—with high precision and remarkable robustness to noise.

Methodological Framework

Research into phase-of-firing codes synthesizes theoretical models with experimental findings. Key frameworks include:

  • Information Theory: Utilizing "Mutual Information" to quantify the additional data provided by spike timing compared to spike rate alone.
  • Local Field Potentials (LFPs): Analyzing low-pass filtered extracellular potentials to identify the "phase" (the specific part of a rhythmic wave) of network activity, such as the 7-12 Hz theta rhythm.
  • Hopfield’s Analogue Model: A theoretical framework demonstrating how neurons receiving stimulus-specific input and common subthreshold oscillations can achieve scale-invariant recognition through logarithmic transformations.

These methods have been applied to diverse datasets, including recordings from the rat hippocampus (place and grid cells), macaque primary visual cortex (V1) during naturalistic movie viewing, and macaque auditory cortex (A1) during exposure to natural sounds in noisy environments.

Key Insights and Results

The phase-of-firing code provides several advantages over traditional rate-based models, particularly in how it handles complex information streams.

Phase Precession and Spatial Navigation

In the rat hippocampus, as an animal traverses a "place field," neurons exhibit "phase precession": they fire at progressively earlier phases of the theta cycle. This phase-of-firing is often a more accurate predictor of location than the firing rate itself. Furthermore, different phases can carry qualitatively different information; for instance, the descending phase of the theta cycle may predict current position, while the ascending slope encodes heading direction.

Information Multiplexing and Disambiguation

The brain appears to multiplex information across multiple timescales: millisecond-scale spike patterns, tens-to-hundreds of milliseconds firing rates, and Hertz-scale spike timing relative to oscillations. In the macaque visual cortex, incorporating the phase-of-firing increased the total information extracted from neural responses by over 50% compared to firing rate alone. This coding scheme is particularly effective at disambiguating stimuli that elicit similar firing rates.

Comparison of Rate Code and Phase-of-Firing Code for two different stimuli.

Robustness to Environmental Noise

In auditory experiments, while information held in firing rates dropped significantly as environmental noise increased, the information within the phase-of-firing remained relatively stable. This suggests that oscillations provide a reliable internal reference frame for decoding in unpredictable or "unreliable" environments.

Limitations and Future Directions

A critical caveat in this field is that encoding does not necessarily imply decoding. While researchers can extract information from neural phases using computational tools, it remains to be definitively proven that the brain's downstream "readout" mechanisms actually utilize this information to drive behavior.

Future research should focus on:

  1. Behavioral Impact: Directly linking phase-of-firing variations to an animal's actual decisions.
  2. Model Verification: Experimentally confirming theoretical models like Hopfield’s in systems such as the olfactory bulb.
  3. Causal Manipulation: Using tools like optogenetics to shift spike phases and observe resulting changes in perception or navigation.

Reference

Cattani, A., Einevoll, G. T., & Panzeri, S. (2015). Phase-of-firing code. Encyclopedia of Computational Neuroscience.

Summary updated

Abstract

Definition. The phase-of-firing code is a neural coding scheme whereby neurons encode information using the time at which they fire spikes within a cycle of the ongoing oscillatory pattern of network activity. This coding scheme may allow neurons to use their temporal pattern of spikes to encode information that is not encoded in their firing rate.
BibTeX
@misc{cattani2015phase,
  author = {Anna Cattani and Gaute T. Einevoll and Stefano Panzeri},
  title = {{Phase-of-firing code}},
  year = {2015},
  eprint = {1504.03954},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1504.03954},
}
Figure