Auditory steady-state responses (SSRs) are believed to result from the superimposition of the middle latency responses (MLRs) evoked by individual stimuli. Our recent studies challenge this hypothesis in several regards. Surface-electrical recordings performed in 16 normal subjects showed that the prediction curves obtained by MLR linear addition failed to predict SSRs at rates other than 40 Hz. Amplitude and phase differences between actual and predicted responses point to the intervention of phenomena related to the recovery cycle of the neural networks underlying the SSR generation. A click train paradigm at a 40 Hz rate was utilized and an approximation to the response evoked by the last stimulus was obtained by an analytical handling. The most relevant feature of this response in comparison to the MLR was the appearance of an additional activity which could be related to the fast oscillations of auditory cortical neurons. Our findings suggest that the responses evoked by individual stimuli during steady-state stimulation change by increasing the repetition rate, thus contradicting the most widely accepted hypothesis of the MLR linear addition in the SSR generation.