In an earlier paper (Raymond C. Battalio, John H. Kagel, and Don N. Mac Donald, 1985), we reported Allais-type violations of the independence axiom of expected utility theory with rats choosing over positively valued payoffs (food rewards). This note extends this research, examining animals' choices over losses, testing for (1) standard Allais-type common ratio effect violations of expected utility theory and (2) fanning out of indifference curves for random prospects, tests of Mark J. Machina's (1982, 1987) hypothesis II (hereafter H2), over previously unexplored areas of the unit probability triangle. Results from a parallel series of experiments using human subjects choosing over real losses are also reported. For both rats and people, we find standard Allais-type violations of expected utility theory and a systematic failure of the fanning out hypothesis in the southeast corner of the unit probability triangle, in the case of losses. Thus, the fanning out hypothesis (Machina 1982, 1987) cannot provide a satisfactory explanation for behavioral deviations from expected utility theory.