If the time and amplitude of the peaks are the variables by which you
quantify your data, you should determine these parameters from the
individual trials. To me, this seems a `cleaner' method than to use
time-warping in the averaging process. Afterwards, the parameters determined
from a number of trials can be averaged. Of course, you may end up with
average apmlitude and time parameters that do not correspond to the
average pattern.

--Ton van den Bogert