We study a property of procedures for allocating indivisible goods to agents called duplication monotonicity, first proposed by Baumeister et al. (Journal of Autonomous Agents and Multi-Agent Systems 31 (2017) 628–655). An allocation procedure satisfies duplication monotonicity if two agents with identical preferences always receive at least as good a share together than one agent would on her own. We study this property first for rules that take cardinal inputs, i.e., the numerical utility of each item to each agent; and secondly for rules that take ordinal inputs, i.e., a ranking of all the items for each agent. In the first case, the rules are parametrized by a social welfare ordering interpolating between utilitarian and egalitarian approaches. In the second case, the rules are additionally parametrized by a scoring vector. We show that in the ordinal setting, only the rule using utilitarian social welfare satisfies duplication monotonicity. In stark contrast, in the ordinal setting we prove that a form of duplication monotonicity holds under a weak assumption on the social welfare function (satisfied by all our examples), answering a question by Baumeister et al. (Journal of Autonomous Agents and Multi-Agent Systems 31 (2017) 628–655).