A "half-pulse" slice selection approach is used in the ultrashort echo time pulse sequence and is required to give minimal transverse relaxation in a two-dimensional acquisition. This method splits the normal excitation radiofrequency pulse in half and acquires a pair of images, each using one of these half-pulses. These half-pulses are used without a refocusing gradient since summing the pair of images yields images with accurate slice selection. When the radiofrequency pulse duration is similar to the sample T(2), characteristics such as the effective echo time and choice of radiofrequency pulse require careful evaluation as some of the approximations in conventional slice selection do not apply. We derive a theory that includes relaxation during excitation using Pauly's excitation k-space formalism. Further, this theory is tested on phantoms with a range of values of T(2) demonstrating the effect on the slice profile. We conclude that relaxation during excitation is significant and should be included in our estimate of the T(2) weighting of the sequence. In general, the T(2) weighting should be measured from the time of the centroid of the excitation pulse.