I present a theoretical effort to develop tools and statistical analysis of neural responses to repeated presentations of identical time varying stimuli. Such experiments may produce responses characterized by regions of elevated firing separated by longer periods of low firing rate referred to as “events.” Unlike previous methods, which find events based on a firing rate threshold, I present a four parameter, reproducible method which first discovers “spike patterns” (subsets of trials with similar spike timing) using unsupervised clustering and a spike train metric, followed by the use of an interspike interval threshold to detect events. I present results from in vitro data showing that the precision of the resulting events is higher than that estimated using a firing rate threshold technique; events within a single spike pattern may be very precise, even if they overlap in time across spike patterns. This analysis provides a model of neural activity which preserves information about spike patterns which can be used to generalize single unit recordings to multi-unit activity. I also present a statistical test to characterize whether events are correlated with one another. A new insight provided is that the choice of time scale for the metric space analysis should maximize the information in the distances between spike trains, regardless of specific timescales in the data. The event finding procedure works well for data sets with more than 20 trials, and with events which are well separated within spike patterns.

I also present a comparison of two methods for selecting one of two stimuli present in the receptive field of a single cortical neuron. In the first, clustering of excitatory and inhibitory synaptic input in the dendrites of a model layer 2/3 pyramidal cell is demonstrated to be insignificant in the selection of signals by unbalancing inhibition. In the second case, phase locking of excitatory inputs at two different phases of a local γ oscillation is demonstrated to produce statistically significant stimulus selection in output firing rate.

Together these results contribute to the analysis, representation and understanding of neural responses to time-varying signals.