I consider a general specification of criminals' objective function and argue that, when the general non-expected utility theory is substituted for the traditional expected utility theory, the high-fine-low-probability result (Becker, 1968) only holds under specific and strong restrictions.