Ideas at work in Bernard Williams’‘integrity objection’ threaten not only direct act-utilitarianism (act-utilitarianism considered as decision-procedure and as test of rightness) but also indirect act-utilitarianism (act-utilitarianism as test of rightness only). Calculation can decide whether an action is utilitarianly right only if it takes every evaluatively relevant feature of alternatives into account. I assume, following Williams, that utilitarianism is reductionist, i.e., represents every case of an agent's valuing something as a case of having a preference. But thanks to an internal relation between an agent's values and the shape of practical deliberation, they cannot always be so represented, so there are some things agents typically value whose value to them utilitarianism necessarily misrepresents. There are therefore some actions such that calculation cannot decide whether they are utilitarianly right, and utilitarianism is incoherent as a test of rightness.