Electric fields carry unique information beyond spiking activity during working memory across species

Modern theories of neural representation identify spiking activity and synaptic interactions as the main informational substrate, despite increasing evidence that population-level dynamics and endogenous electric fields play a key role in cognition and may constitute more stable computational variables. For example, it has been shown that whereas representations at the neuronal level can be quite unstable over long periods of time - a phenomenon known as representational drift -, electric field patterns seem to maintain higher stability across trials. In addition, multiple distinct microscopic circuit configurations can give rise to similar low-dimensional dynamics and nearly identical macroscopic field patterns, suggesting a form of representational invariance beyond single-neuron activity. This raises the possibility that electric fields encode stimulus-relevant information that is partially decoupled from their synaptic constituents. Here we test the hypothesis that endogenous electric fields carry stable and potentially unique representational content during stimulus encoding and working memory maintenance. We re-analyze electrophysiological datasets (local field potentials and spiking activity) recorded during analogous delayed-match-to-sample tasks across multiple animal species. We focus on whether macroscopic field dynamics exhibit greater trial-to-trial stability and cross-condition generalization than the underlying spiking activity, particularly during delay periods. We further investigate whether persistent field structure can be observed even in cases where single-neuron activity is highly variable or weakly correlated across trials, consistent with the existence of wave-like or low-dimensional field dynamics sustained by irregular neuronal spiking. Finally, we test whether electric field representations persist beyond stimulus offset in regimes where classical mechanisms of working memory—such as persistent spiking or short-term synaptic plasticity—might be weak or absent. Our framework provides an empirical test of whether macroscopic electric fields constitute a representational substrate in their own right, supporting stable encoding and maintenance of information across species and neural recording scales, thus opening the door to uncover previously-unknown mechanisms and representational formats for information encoding and maintenance in neural systems.