Abstract Taking as our starting point Trivers' (1974) account of parent-offspring conflict, we develop models of the influence of brood size on the optimal level of parental investment (PI) in the whole brood for parent and offspring, and on the magnitude of conflict between them. A modification of Trivers' model is proposed. In general, the benefit of an act of PI to an offspring in a brood of size N is ( N+1)/ N times the benefit to its parent. Therefore as brood size increases, offspring benefit approaches parental benefit, and this is because an increasing proportion of the offspring's benefit is being gained through siblings, to which offspring and parent are equally related. A distinction is drawn between ‘shared’ and ‘unshared’ types of PI. When PI is shared the total benefit accruing is not directly gained by all offspring but is shared amongst them (e.g. food brought to the young). In contrast, unshared PI can simultaneously benefit some or all of the brood (e.g. types of anti-predator defence). For shared investment, PI and conflict are predicted to increase with brood size. Two models of unshared anti-predator defence are described. If the predator characteristically takes the whole brood when it strikes (e.g. altricial nestlings) PI is predicted to increase and conflict decline with brood size, although this effect is inhibited or even reversed for high risk defence tactics because of the higher cost to larger broods if the parent dies. When the predator takes a single offspring (e.g. precocial birds) the parent's optimum PI is independent of brood size, the offspring's optimum PI declines in larger broods and conflict again declines with brood size. The parent is commonly expected to win the conflict over anti-predator care. Predictions concerning PI levels gain support from existing data, largely for birds, but evaluation of those for conflict must await the collection of new data. The distinction between shared and unshared investment is applicable to altruistic behaviour in general.