The traditional approach to light amplification is shown to be based on the idea that photons of the incoming beam act as independent particles. If atomic stimulation is instead attributed to the action of the wave, a different photon distribution is obtained in the final state. The new distribution turns out to be Poissonian, in agreement with some experimental evidence. The expected rate of coincidences for an empty-wave amplification experiment is calculated, and found up to 30% higher than obtained from (...) the traditional approach. (shrink)