Marine Protected Areas (MPAs) can theoretically achieve two Goals: protection of biodiversity, and replenishment of populations both inside and far outside the MPA boundary. The second is supposed to result primarily from larval export from the MPA. Although there is evidence that 'no take' MPAs protect biodiversity and have higher stocks of larger, older, more fecund fishes, there is scant empirical evidence to support the notion that MPAs actually do replenish unprotected areas, or if they do, over what spatial scale. This notion of replenishment over large scales is largely based on theoretical considerations of larval dispersal and larval biology. Recent research shows that at least fish larvae do not conform to traditional theory: they may have much more control over where they disperse than previously thought. This has important implications for the design and implementation of MPAs, and what we can expect from them as conservation tools. This paper reviews recent advances in understanding larval fish biology and behavioural capabilities and how these impact on the efficacy and design of MPAs. If larvae are as good at resisting dispersal as their behavioural capabilities suggest, then replenishment in ecologically meaningful quantities probably takes place over much smaller scales than previously thought, and MPAs will have to be designed accordingly. These scales, however, are likely to differ spatially, temporally and among species.