The brain can accurately predict the forces needed to efficiently manipulate familiar objects in relation to mechanical properties such as weight. These predictions involve memory or some type of central representation, but visual analysis of size also yields accurate predictions of the needed fingertip forces. This raises the issue of which process (weight memory or visual size analysis) is used during everyday life when handling familiar objects. Our aim was to determine if subjects use a sensorimotor memory of weight, or a visual size analysis, to predictively set their vertical lift force when lifting a recently handled object. Two groups of subjects lifted an opaque brown bottle filled with water (470 g) during the first experimental session, and then rested for 15 min in a different room. Both groups were told that they would lift the same bottle in their next session. However, the experimental group returned to lift a slightly smaller bottle filled with water (360 g) that otherwise was identical in appearance to the first bottle. The control group returned to lift the same bottle from the first session, which was only partially filled with water so that it also weighed 360 g. At the end of the second session subjects were asked if they observed any changes between sessions, but no subject indicated awareness of a specific change. An acceleration ratio was computed by dividing the peak vertical acceleration during the first lift of the second session by the average peak acceleration of the last five lifts during the first session. This ratio was >1 for the control subjects 1.30 (SEM 0.08), indicating that they scaled their lift force for the first lift of the second session based on a memory of the (heavier) bottle from the first session. In contrast, the acceleration ratio was 0.94 (0.10) for the experimental group (P < 0.011). We conclude that the experimental group processed visual cues concerning the size of the bottle. These findings raise the possibility that even with familiar objects we predict fingertip forces using an on-line visual analysis of size (along with memory of density), rather than accessing memory related to object weight.