In 1914, Foerster1 summarized the literature on the association of erythema nodosum with tuberculosis, laying special emphasis on the association of the cutaneous picture with miliary tuberculosis in childhood. He mentioned in particular the reports of Pollak,2 Dunlop,3 Abt,4 and Sezary,5 of patients who had died of tuberculous meningitis in close sequence to an attack of erythema nodosum. Such a clinical sequence implies a hematogenous distribution of tuberculosis bacilli; and yet the outstanding fact of the literature has been the failure of the majority of efforts to demonstrate the presence of bacilli by inoculations either from the blood stream or the tissue of the nodose lesions. Of the numerous investigators who have attacked this problem only Brian,6 Hildebrandt,7 and Landouzy8 claim positive results.
Schönfeld's9 study of thirteen cases of tuberculids, undertaken under Zieler's10 direction, secured only negative results. It was