Objective: To examine the applicability of a neural network classifica
tion strategy to examine the independent contribution of psychomotor d
isturbance (PMD) and endogeneity symptoms to the DSM-III-H definition
of melancholia. Method: We studied 407 depressed patients with the cli
nical dataset comprising 17 endogeneity symptoms and the 18-item CORE
measure of behaviourally rated PMD. A multilayer perceptron neural net
work was used to fit non-linear models of varying complexity. A linear
discriminant function analysis was also used to generate a model for
comparison with the non-linear models. Results: Models (linear and non
-linear) using PMD items only and endogeneity symptoms only had simila
r rates of successful classification, while non-linear models combinin
g both PMD and symptom scores achieved the best classifications. Concl
usions: Our current non-linear model was superior to a linear analysis
, a finding which may have wider application to psychiatric classifica
tion. Our non-linear analysis of depressive subtypes supports the bina
ry view that melancholic and nonmelancholic depression are separate cl
inical disorders rather than different forms of the same entity. This
study illustrates how non-linear modelling with neural networks is a p
otentially fruitful approach to the study of the diagnostic taxonomy o
f psychiatric disorders and to clinical decision-making.