Abstract Australian children aged 6–9 years (N = 120, 71 females; data collected in 2021-2022) were tasked with remembering the locations of 1, 3, 5, and 7 targets hidden under 25 cups on different trials. In the critical test phase, children were provided with a limited number of tokens to allocate across trials, which they could use to mark target locations and assist future memory performance. Following the search period, children were invited to adjust their previous token allocation. Although 8- to 9-year-olds prospectively allocated proportionately more tokens to more difficult trials, 6- to 7-year-olds did so only in retrospect. During middle childhood, humans become increasingly adept at weighing up when to rely on their unaided capacities and when to offload cognitive demand.