Multiple importance sampling (MIS) is a crucial tool in Monte Carlo light transport simulation. Bidirectional rendering methods in particular rely heavily on MIS to achieve robustness. Many such methods also make use of path correlation to increase efficiency. Photon mapping, path splitting, and path reuse methods gain efficiency by constructing a set of paths that all share a common camera prefix. Unfortunately, this causes correlation that is ignored by the MIS heuristics, which can result in poor technique combination and noisy images. We propose a practical and robust solution to that problem. We incorporate correlation into the balance heuristic, based on quantities that are already required for MIS. Our heuristic achieves considerably lower error than the balance heuristic, while avoiding computational or memory overhead.