We describe a new method for measuring skin conductance responses, designed to overcome the problem of overlapping skin conductance responses. The method relies on the assumptions that the underlying sudomotor nerve signal has a shorter time-constant than the skin conductance signal itself, and that the sudomotor bursts arrive as discrete, separated events. By converting the skin conductance signal into a time-series with a shorter time-constant, we are able to extract the separated peaks in the estimated underlying driver signal. The separated driver peaks are then used to re-estimate each individual skin conductance response. The method is automated and applied to a normative database of 735 subjects, for which skin conductance was measured during an auditory oddball paradigm. (c) 2005 Elsevier B.V. All rights reserved.