Although classification of mental disorders using more than clinical descri
ption would be desirable, there is scant evidence that available laboratory
tests (i.e, biological indices) would provide more valid classifications t
han current diagnostic systems (e,g. DSM-IV). We used cluster analysis of f
our biological variables to classify 163 psychotic patients and 83 nonpsych
iatric comparison subjects, Analyses revealed a three-cluster solution with
the first cluster reflecting electrodermal deviance, the second cluster re
presenting nondeviant biological function, and the third cluster reflecting
increased nailfold plexus visibility and ocular motor dysfunction. To asse
ss the construct validity of proband clusters we examined ocular motor perf
ormance in 156 first-degree relatives as a function of proband cluster memb
ership. First-degree relatives of third cluster probands exhibited worse oc
ular motor performance than relatives of other cluster probands. Additional
ly, better classification sensitivity and specificity were obtained for the
relatives when they were grouped by proband cluster than by proband DSM-IV
diagnosis. When a single proband characteristic (i.e. eyetracking performa
nce) was used to group relatives, classification sensitivity and specificit
y failed to significantly increase over grouping by proband DSM-TV diagnosi
s. Multivariate biologically defined clusters may offer an advantage over D
SM-TV classification when examining nosology and etiology of psychotic diso
rders. (C) 2001 Elsevier Science B,V, All rights reserved.