Time and distance estimates were elicited with either unit-based (e.g., "How many days until...") or end-based (e.g., "On what date...") questions. For intervals of uncertain extent, unit-based estimates were consistently lower than were the corresponding end-based estimates. The observed patterns are consistent with an anchoring and adjustment process: When people generate unit-based estimates of uncertain dates or distances, they may anchor on the "here" or "now" and adjust incrementally by the unit; such adjustment, however, is often insufficient and yields systematic underestimation. Although this anchoring and adjustment cannot be directly observed, consistent with the hypothesized process, reliance on larger units yielded higher estimates and warning about insufficient adjustment reduced the effect. Implications for research on anchoring, the planning fallacy, and everyday judgment are discussed.