[Eeglablist] P300 Window

[Zahra Rezazadeh]

Dear EEG/ERP Researchers,
I have done a visual P300 matrix speller experiment (with some modifications) on 6 participants during 5 sessions.
I had four different conditions, which all share the same matrix but have varying stimulus intensification intensities.
One of my hypotheses was about the effects of training throughout these 5 sessions.
I know that the manipulations that I have done result in higher P300 latency and lower P300 amplitude.

My question is about deciding the exact P300 window that I would take mean target amplitudes and mean difference between targets and non-targets from.
The P300 peaks in my single-subject and also my grand average plots for my different conditions and different participants happen with different latencies. 
Using a broad range of 200-1000ms (stated as the recommended window in general) would include so much irrelevant data into my processing.
Using smaller windows would include P300 in some but not all cases.

I have read papers about it. I have not found any rules to decide about the window.
It is only stated that the latency would be higher when tasks are more difficult.

Questions:
1. Do you know any way to firmly decide on one window for all participants and all conditions (without neglecting good data and without including irrelevant data)?
2. Does it make sense to take ERPs for each condition/participant from their own window? (some conditions are harder, some participants are slower)

Thanks in advance for your support.
Bests,
Zahra