Article
Katrina MacLeod, Alex Bäcker, Gilles Laurent
California Institute of Technology, Pasadena, CA
3 min readOscillatory and synchronized neural activity has been proposed to play a functional role in feature binding, selective attention, and neural coding. However, a fundamental question remains: which neurons actually read out the temporal information contained in synchronized and oscillatory spike trains? This paper addresses the conditions under which downstream neurons can decode temporal correlations across neural populations, demonstrating that the readout of temporal information depends critically on the membrane time constants and coincidence-detection properties of postsynaptic neurons. We show that neurons with short membrane time constants are selectively tuned to synchronized inputs, while neurons with longer time constants integrate over longer windows and are relatively insensitive to synchrony. These findings provide constraints on the types of neural circuits that can exploit population temporal codes and have broad implications for theories of neural computation and sensory coding in the brain.
The role of neural synchrony in sensory processing and cognition has been debated extensively. Oscillatory and synchronized activity in sensory cortex has been proposed as a mechanism for feature binding and selective attention. Yet a critical and often overlooked question is how downstream neurons could extract and use this temporal information.
Using the olfactory system of the locust as a model, we investigate the computational properties required of a postsynaptic neuron to act as a reader of population-level temporal codes. Our analysis bridges theoretical neuroscience and experimental data to constrain viable mechanisms of neural readout.
Related Reading
Derives the optimal tuning width for neurons encoding a finite set of stimuli — challenging the dominant view that sharp tuning always maximizes information.
Read moreA novel method for estimating the intrinsic dimensionality of high-dimensional data, with applications to neural population data.
Read more