What are the causes of reduced fit in scaling and clustering studies of semantic proximity data, and how else to measure them: a comment on Storms et al. (2003).

G. Storms, T. Dirikx, J. Saerens, S. Verstraeten, and P. P. De Deyn (2003) noted that the goodness-of-fit indices for scaling and clustering solutions of proximity data from populations with hypothesized semantic deficits tend to be reduced in comparison with control participants, and that this precludes an unambiguous interpretation of such data. However, this effect is not necessarily as inconsistent with the proposal of semantic degradation in these groups, or as consistent with explanations in terms of access or attentional problems, as is suggested by G. Storms et al. Also, although performance on word fluency tasks is certainly constrained by the ability to access stored information, there are ways in which this confounding influence may be controlled for in the study of semantic deficits.