1. A blind in-patient investigation of the phenomenology and treatment of 137 hospitalized female depressives is described, in which amitriptyline was compared with imipramine. Using discharge without E. C. T. as the criterion of success, amitriptyline, which relieved 81 per cent. of patients, was superior to imipramine, which relieved 54 per cent. (p<.002).

2. Four symptoms increased significantly with age and five increased disproportionately with increasing overall severity.

3. Five sets of prognostic variables were studied for the two drug groups and for twelve comparable sub-groups delineated by age and severity of illness. These variables were: (i) the Hobson scale score; (ii) the initial severity of the depressive symptoms; (iii) the presence of delusions; (iv) the effects of age, menopausal state, type and severity of depression; and (v) the response to one week of treatment.

(i) The Hobson scale score was of no predictive value with either drug.

(ii) Severity of individual symptoms did not affect the outlook with amitriptyline, but severity in a number of symptoms was significantly associated with a low rate of response to imipramine.

(iii) Delusions, when present, lessened the effectiveness of amitriptyline and nullified the effectiveness of imipramine.

(iv) In regard to age and menopausal state, the superiority of amitriptyline was clearly evident in patients over 50 (p=.003) and in those who were post-menopausal (p=.002). Endogenous depressives over 50 did significantly better with amitriptyline (p=.001), though reactive cases did well on both drugs. Not only was amitriptyline significantly more effective in very severe depressives according to the ratings of both physicians (p=.003) and nurses (p=.0008), but imipramine, according to these raters, was significantly less effective in severe cases than it was in milder ones.

Correlations between age, menopausal state, type of depression and severity, were calculated; the first three variables were found to be highly inter-correlated, but none of the three correlated appreciably with the fourth. An analysis of variance carried out on 120 of the 137 patients revealed a highly significant difference between the drugs in favour of amitriptyline, and a significant interaction between drugs and severity of illness.

(v) The response to one week of treatment was found to be significantly correlated, in the case of amitriptyline, with percentage drop in the Hamilton scale score (r=.28, p<.05). Thirty-nine of the 56 patients responding to amitriptyline showed 30 per cent. or greater improvement, was followed by recovery without E.C.T. in patients over 60 and in those over 50 with severe depression. On the Hamilton scale the 56 amitriptyline responders in comparison with the 13 non-responders, were initially less severely ill, though this was not clinically distinguishable; after a week, however, non-responders could readily be differentiated by their failure to improve in three symptoms—anorexia, middle insomnia and reduced work and interests. The early tranquillizing action of amitriptyline was not of value in predicting outcome.

4. Patients successfully treated with amitriptyline spent no longer in hospital than did a matched retrospective control group treated with E.C.T.

5. The practical implications from this study would seem to be:

(i) Depressed patients with unequivocal depressive delusions should be given E.C.T.

(ii) Other depressives should be given amitriptyline. Of any 10 patients, 8 or 9 can be expected to recover in four to six weeks, and 6 or 7 will improve significantly within the first week of treatment. At this time, failure to improve in anorexia, middle insomnia and reduced work and interests, should be taken as an indication for alternative treatment.