Is there madness in the method? A comment on Storms et al. (2003).

G. Storms, T. Dirikx, J. Saerens, S. Verstraeten, and P. De Deyn (2003) offer a critique of multidimensional scaling studies purporting to reveal systematic semantic deficits in Alzheimer's disease and other syndromes which, in contrast to prior claims, demonstrates that patient performance is indiscriminable from random data. The argument is an indictment against the conclusions drawn by previous researchers; but a consideration of the materials used in these studies suggests that the problem may not lie in the method per se, but in poor stimulus design. In all of the similarity judgment tasks reviewed, stimulus items were differentiated solely on the basis of subtle conceptual distinctions of the sort known to be most vulnerable to semantic impairment. Thus, one would not expect systematic responding from semantically impaired patients to begin with. An alternative design is suggested.