Neuronal firing is determined largely by incoming barrages of excitatory postsynaptic potentials (EPSPs), each of which produce a transient increase in firing probability. To measure the effects of weak transient inputs on firing probability of cortical neurons, we compute Phase-Response Curves (PRC). PRCs, whose shape can be related to the dynamics of spike generation, document the changes in timing of spikes caused by an EPSP in a repetitively firing neuron as a function of when it arrives in the interspike interval (ISI). The PRC can be exactly related to the post-stimulus time histogram (PSTH) so that knowledge of one uniquely determines the other. Typically, PRCs have zero values at the start and end of the ISI, where EPSPS have minimal effects and a peak in the middle. Where the peak occurs depends in part on the firing properties of neurons. The PRC can have regions of positivity (negativity) corresponding respectively to speeding up (slowing down) the time of the next spike. A simple canonical model for spike generation is introduced which shows how both the background firing rate and the degree of post-spike afterhyperpolarization contribute to the shape of the PRC and thus to the PSTH. PRCs in strongly adapting neurons are highly skewed to the right (indicating a higher change in probability when the EPSPs appear late in the ISI) and can have negative regions (indicating a decrease in firing probability) early in the ISI. The PRC becomes more skewed to the right as the firing rate decreases. Thus, at low firing rates, the spikes are triggered preferentially by inputs that occur only during a small time interval late in the ISI. This implies that the neuron is more of a coincidence detector at low firing frequencies and more of an integrator at high frequencies. The steady-state theory is shown to also hold for slowly varying inputs.