SPIKE TRAIN ENCODING BY REGULAR-SPIKING CELLS OF THE VISUAL-CORTEX

1. To study the encoding of input currents into output spike trains by regular-spiking cells, we recorded intracellularly from slices of the guinea pig visual cortex while injecting step, sinusoidal, and broadb and noise currents. 2. When measured with sinusoidal currents, the fre quency tuning of the spike responses was markedly band-pass. The prefe rred frequency was between 8 and 30 Hz, and grew with stimulus amplitu de and mean intensity. 3. Stimulation with broadband noise currents dr amatically enhanced the gain of the spike responses at low and high fr equencies, yielding an essentially flat frequency tuning between 0.1 a nd 130 Hz. 4. The averaged spike responses to sinusoidal currents exhi bited two nonlinearities: rectification and spike synchronization. By contrast, no nonlinearity was evident in the averaged responses to bro adband noise stimuli. 5. These properties of the spike responses were not present in the membrane potential responses. The latter were rough ly linear, and their frequency tuning was low-pass and well fit by a s ingle-compartment passive model of the cell membrane composed of a res istance and a capacitance in parallel (RC circuit). 6. To account for the spike responses, we used a ''sandwich model'' consisting of a low- pass linear filter (the RC circuit), a rectification nonlinearity, and a high-pass linear filter. The model is described by six parameters a nd predicts analog firing rates rather than discrete spikes. It provid ed satisfactory fits to the firing rate responses to steps, sinusoids, and broadband noise currents. 7. The properties of spike encoding are consistent with temporal nonlinearities of the visual responses in V1 , such as the dependence of response frequency tuning and latency on s timulus contrast and bandwidth. We speculate that one of the roles of the high-frequency membrane potential fluctuations observed in vivo co uld be to amplify and linearize the responses to lower, stimulus-relat ed frequencies.