Johanna,
You can use the option in 3dDeconvolve for a 1D input "-input1D" and have your SCR waveform be your dependent variable. You can code for the stimfiles as you would with an FMRI timeseries. Depending on how you want to analyze your SCResponses you could look at the coefficient for the peak response or create a general linear test (-glt) to calculate area under the curve (AUC) as change from baseline.
Given the inherent variability present in SCR waveforms you may get some screwy numbers using AUC because the waveform may start off below baseline, then rise above it summing to a zero response. If you see this problem using AUC, then maybe the peak response, average of the two timepoints surrounding the peak response, or the average of the last several timepoints (second interval response) may work better.
Keep in mind that for 3dDeconvolve it calculates a change from baseline where baseline is calculated across the entire time series. Because SCR can vary on a changing (and not necessarily linearly dependent) SC level, I have found it is usually better to use a local baseline for each response(change from the average of the 5-10 sec before the stimulus or from the value at stimulus onset), not a baseline derived from the entire waveform if the waveform was from a long experimental session where baseline variability/instability may have occured.
One thing I have found with SCR data analysis is that you should try many different things because the variability in the waveform and between subjects may show an experimental effect in one analysis, but not the other.
Hope this helps,
Christine Smith