To scale or not to scale: McCarthy and Wood revisited.

We dispute McCarthy and Wood's (1985) claim that some form of scaling should be applied routinely to ERP data before determining differences in scalp distributions between conditions (or groups). Their simulation study involved assumptions about the nature of the variability within each condition, most significantly that the standard deviations are identical at all electrodes, irrespective of the means. Alternative plausible assumptions may be proposed for which scaling is unnecessary. Furthermore, we show that the two main forms of scaling they proposed may distort or even completely eliminate real differences in scalp topography reflecting genuinely different underlying sources.