A. R. Haig, E. Gordon, and S. Hook (1997) disputed
G. McCarthy and C. C. Wood's (1985) contention that
scaling should be used when assessing the statistical significance
of between condition (or group) differences in the shapes
of event-related potential (ERP) scalp topographies. Haig
et al. based their contention upon the lack of empirical
realism in McCarthy and Wood's model of within-group
ERP noise, claiming that McCarthy and Wood's results
could not be generalized to realistic ERP data. We argue,
on both empirical and theoretical grounds, that Haig et
al. do not make a compelling case against generalization
of McCarthy and Wood's results. Moreover, Haig et
al.'s conclusion is based upon a misconception of
how scaling should be used. We conclude that when a quantitative
measure of differences between topographic shapes is needed,
scaling is not an option—it is a requirement.