Psychiatric classification remains a major problem because it must still be inferred from semiquantitative ratings of clinical phenomena. Much the same problem confronted biologists attempting to distinguish discrete species on the basis of morphology in the late 19th century, resulting in many inconsistent classifications. Then, in 1894, Karl Pearson developed statistical methods to identify subpopulations from a heterogeneous mixture.1 The existence of two subgroups can be inferred under some conditions when quantitative data have a bimodal distribution, ie, two peaks separated by an intermediate point of rarity. Later, many species were successfully distinguished in this way.2 However, bimodality may also arise from a variety of measurement and selection artifacts, such as the arbitrary exclusion of intermediate cases or the pooling of nonrandom subsamples of extreme cases.3 Accordingly, the inference of multiple discrete disorders from quantitative data requires careful attention to experimental design and statistical analysis.
CLONINGER CR, GUZE SB, CLAYTON PJ. Assessment of Evidence for a Categorical View of Schizophrenia-Reply. Arch Gen Psychiatry. 1986;43(7):713-714. doi:10.1001/archpsyc.1986.01800070103015