where is the relative potency of the test preparation with respect to the standard one. The equality of the intercepts as and T, constitutes the fundamental assumption of the slope-ratio assay, and the ratio of the slopes provides the relative potency. In standard textbooks, such as in Finney (1952, ch. 7 and 8), or elsewhere in the literature, G (e) is usually assumed to be normal or logistic, and the test for Ho: as = T, and the estimate of = (fT/AIs)1/ are then based on the maximum likelihood estimates of as, T, fls and fi. As has been discussed in Section 1 of Sen (1971), these (so-called parametric) procedures, like their counterparts in parallel line assays, are sensitive to gross-errors or outliers, and may be quite inefficient when the actual G (e) is different from the assumed one; also for G (e) belonging to the class of distributions with heavy tails (such as the Cauchy or the double exponential distribution), they are usually inefficient. In the same spirit as in Sen (1971), some alternative more robust procedures are developed here. In Section 2, we start with robust tests for the validity of the fundamental assumption of the assay. Section 3 is devoted to point estimates of p, while Section 4 deals with the interval estimates of p. Throughout the paper, we take 2 = 1 (see (1.3)); for 2 t 1, we need only to change by p and the consequent changes are not very elaborate. Also, for the convenience of reading, most of the mathematical derivations are given in the mathematical appendix. Finally, in notations and motivations, the paper has a close proximity with its first part (Sen (1971)), which should be read first. In passing, we may remark that though the procedures proposed in this paper (as well as