The study is concerned with the application of Multiple Discriminant Analysis in order to identify the different variables which explain the phenomenon of sickness of industrial units, their relative contribution in determining the propensity of sickness, as also to suggest a possible cost-off point which may separate the units belonging to the sickness class from whose belonging to the healthy ones. For purposes of the present study, out of a total of 91 industrial units listed as sick by the IDBI in its portfolio, a sample of 20 units belonging to six different industry classes has been drawn. Likewise a list of 26 healthy industrial units belonging to the same set of industry classes has been drawn so as to develop the discriminant function separating sick units from the healthy ones. Two different models have been tested for the years 1975, '76, '77, and '78. Both the models give us more or less identical results. Either of these may therefore be used for discriminating sick units from the healthy ones.