Scalable video coding enables multimedia content providers to cope with the heterogeneity of networks and multimedia devices. In this paper, we introduce the novel concept of Adaptive Residual Interpolation (ARI) as a tool for efficient spatial scalable video coding. This new approach is based on the inter- layer prediction of residual data. We have implemented the ARI tool by making use of the Joint Scalable Video Model software. Tests have shown that our techniques for residual prediction result in bit rate savings of up to 8 % without diminishing visual quality. The ARI tool is especially useful for video sequences with low motion activity or when the residual data in the spatial enhancement layer is higher quantized than the data in the base layer.