A probed memory task was used to investigate children's short-term visual memory for an object's spatial location or colour. In Experiment 1, 5-yearolds recognised the location of one of three coloured shapes that had appeared in a random spatio-temporal order. Two aspects of the children's performance confirmed their reliance on visual memory. First, performance was impaired when the shapes were visually similar. Secondly, the serial position curve was characteristic of visual memory, with a final-item recency effect and no primacy effect. Experiment 2 assessed 5and 7-year-old children's memory for a shape's colour or its spatial location. Although there was developmental improvement in memory for spatial location, that was confined to pre-recency items, there was no effect of age with regard to the recall of colour. The results go against Hasher and Zacks' (1979) proposal that, in contrast to colour, spatial location would not show developmental improvement because it is remembered automatically. The concept of an object file, that was devised specifically to explain how different visual features of an object are represented (Kahneman & Treisman, 1984), is considered as a potential explanation of both the serial position curve and the distinctive behaviour of different visual features. It is suggested that although 5-year-olds are as adept as 7-year-olds at creating and immediately accessing an object file, they are less able to access information about the visual features of objects whose files are no longer current.