[Eeglablist] Surface Laplacian (CSD) changes baseline period activity?

[Mate Gyurkovics]

Dear all,

I ran a reaction time task in two age groups while recording EEG from 64
channels. I want to look at age related differences in the magnitude of the
N2 component at electrode FCz. I epoched, baseline corrected, then averaged
my data, and the ERPs looked fine - flat baseline periods, then a nice N2
about 200-300 ms after target stimulus presentation. However, when I apply
surface Laplace transformation (or current source density, CSD) to the
data, the waveforms change a little bit. Nothing drastic, but in the
younger age group the baseline period is no longer flat, it appears to be
increasing slightly. Please see the following image:
https://imgur.com/a/nN5s2bA, where the first dashed line indicates the
presentation of the first stimulus on a trial, and time point 0 is the
presentation of the target stimulus, in other words, activity between -200
and 0 makes sense, the gentle upward slope before -200 does not, and it's
not there a) in the untransformed data, b) in adults, c) in all electrodes,
even in the transformed data.

Intertrial intervals were jittered so participants could not predict when
the next trial will come up. Each condition of interest contains over 100
epochs, close to 140 actually.

Does anyone know why this pattern might emerge? Is this cause for concern?
Is it a side effect of the Laplace transform? I want to stress again that
the non-transformed (average referenced) data looks perfectly normal.

(Footnotes about the image: It is actually showing the grand average
waveform across all participants and conditions, but the pattern is
visible. It's a little more pronounced in the young group and not at all
present in adults. Also, I know the units on the Y axis should be
microvolts/cm2, I have accidentally forgotten to change the label.)

Thank you for your help.

Best,
Mate