Averaging: pulling the signal out of the noise
Evoked potentials are often at the microvolt level and lie beneath background EEG, muscle and mains activity. Averaging uses this difference: the signal arrives at the same moment after each stimulus, whereas noise is random.
When successive recordings are aligned to the stimulus and summed, the signal reinforces and the noise cancels. As a result, the signal-to-noise ratio increases with the square root of the number of repetitions: doubling it requires four times as many repetitions.
Why does stimulus rate matter?
If the rate is an exact multiple or simple ratio of the mains frequency, the interference is captured in the same phase at each stimulus and accumulates instead of cancelling. Non-integer rates such as 5.1 or 11.1 are therefore preferred. Very rapid stimulation can also shrink some components because the system does not recover.
Reproducibility
The way to show a wave is real is that it appears at the same latency in at least two independent averages. Do not interpret a wave from a single average.