I wish to define neuronal activities by analyzing the peaks of each channel in EEG signals. Some papers use the negative peaks as a measure of neuronal activity and others deploy the positive peaks. Which peak polarity should I use?
The peak polarity of an EEG is arbitrary.
Positive and negative in an EEG measure is arbitrary. If you measure the EEG between two electrodes and you flip the wires, the polarity of the signal will reverse too.
Only when standard EEG montages are used, in which the active and reference electrodes are strictly defined, then signal polarity makes sense, but only because a certain convention is followed.
If you are interested in peak amplitudes - they are independent of polarity, as long as the peak latencies are defined and absolute values are reported.
If you are interested in peak latencies and wish to compare those with others, then you do need to be aware what polarity your signal has. If you find out the polarity is reversed due to whatever reason - multiply every sample simply with -1. Problem solved.