Food hoarding and memory have primarily been studied in two bird families, the Corvidae (crows, jays, nutcrackers, etc.) and the Paridae (tits, titmice and chickadees). In both families there are species that hoard large quantities of seeds and nuts in the autumn and depend on these stores during the winter. Caches are concealed or highly inconspicuous and the most efficient way to retrieve them is to remember the exact locations. However, a long-term memory for a large number of caches may be physiologically expensive, and especially after long retention intervals, an alternative strategy could be to retrieve caches by cheaper but less efficient methods. Very few studies have been designed to investigate the decay of the memory in birds, but both field observations and experiments point in the same direction: although long-term hoarding corvids seem to possess an accurate long-term memory, long-term hoarding parids do not appear to. I discuss possible reasons for this and suggest that differences between the families in their degree of dependence on stored food or/and size-related limitations of brain capacity may be important.