Caloric restriction (CR) is the only nongenetic manipulation known to reliably prolong life-span. Modeling suggests that humans would need to restrict their intake for many years to reap any lifespan benefits. The feasibility of such prolonged restriction may hinge on whether hunger diminishes with the time period spent restricted. We used the magnitude of hyperphagia on release from restriction as a bioassay of hunger in restricted mice. During restriction, mice develop a characteristic pattern of neuropeptide signals in the arcuate nucleus that reflect their hunger. This pattern is normalized after the postrestriction hyperphagia, validating hyperphagia as an indicator of the hunger during restriction. Mice were food restricted (80% of ad lib.) for 50 days. They initially lost weight, but then became weight stable and were maintained in CR at this lower level of energy balance for between 0 and 50 days and were then fed ad lib. for 50 days. When released onto ad lib. food, the magnitude of the hyperphagic response was independent of the prior length of CR. Hyperphagia ended when body mass was normalized. Hunger therefore did not diminish even when they were restricted for 100 days, equivalent to about 11 years in humans. The pattern of hyperphagic response suggested that signals coding body mass drive hunger during restriction, and because body mass under restriction remains depressed, this suggests that hunger would never disappear, making restriction to prolong lifespan in humans difficult to accomplish.