In our model, delayed inhibition in synchronized neurons led to a positive net excitation concentrated at the onset of the synaptic input (Fig.

Excitation concentrated at the onset of the synaptic input in synchronized neurons should evoke an onset response, while net excitation spread out over a longer time period in non-synchronized neurons should produce a more sustained response.

Each acoustic pulse was simulated as the summation of 10 excitatory and 10 inhibitory synaptic inputs [24], each temporally jittered (Gaussian distribution, 0 = 1 ms).

Each synaptic input was modeled as a time-varying conductance fit to an alpha function: with a time constant 5 ms and an amplitude determined by the excitatory input parameter of the model (ranging from 0.3 to 6 n8).

A 10 ms delay was added to the synaptic input to simulate the delay between peripheral auditory system and auditory corteX.

Using only synaptic inputs with a short time-constant (5 ms), approximating AMPA and GABA-A receptors, our model was able to simulate the major types of neural representations (synchronized, non-synchronized, and mixed), as well as two atypical types (inhibitory and bandpassed).

Other acoustic stimuli, such as sinusoidal amplitude modulated (sAM) tones, that change their spectral bandwidth and pulse duration with modulation frequency, cannot be represented accurately by our model, however the addition of new parameters that account for the spectrum of the acoustic stimulus [47] and/or the adaptation of synaptic inputs [42] would likely provide further improvements to our description of temporal processing in auditory cortical neurons.

Increasing the temporal jitter of synaptic inputs (S9 Fig) generated non-synchronized responses more typical of real neurons, while maintaining synchronized responses.

While our model’s ability to generate non-synchronized responses required a source of internal noise (or sufficient temporal jitter of synaptic inputs ), other methods of generating internal noise also produced similar results, including injecting noise as a current into the integrate-and-fire model to simulate background synaptic activity [30] (810 Fig) and adding Gaussian noise to the membrane potential spiking threshold [31] (811 Fig).

In our computational model, the time-varying conductance used to simulate the neuron’s synaptic input was simplified to only approximate the AMPA and GABA-A currents evoked by the acoustic pulse train, with a time-constant of 5 ms [24] (see Methods).

Other than increased temporal jitter of synaptic inputs (uniform distribution, 0 = 8.7 ms), no other sources of noise were added to the model.