Flood–frequency estimation is inherently uncertain. The practitioner applies a combination of gauged data, scientific method and hydrological judgement to derive a flood–frequency curve for a particular site. The resulting estimate can be thought fully satisfactory only if it is broadly consistent with all that is reliably known about the flood–frequency behaviour of the river. The paper takes as its main theme the search for information to strengthen a flood–risk estimate made from peak flows alone. Extra information comes in many forms, including documentary and monumental records of historical floods, and palaeological markers. Meteorological information is also useful, although rainfall rarity is difficult to assess objectively and can be a notoriously unreliable indicator of flood rarity. On highly permeable catchments, groundwater levels present additional data. Other types of information are relevant to judging hydrological similarity when the flood–frequency estimate derives from data pooled across several catchments. After highlighting information sources, the paper explores a second theme: that of consistency in flood–risk estimates. Following publication of the Flood estimation handbook, studies of flood risk are now using digital catchment data. Automated calculation methods allow estimates by standard methods to be mapped basin–wide, revealing anomalies at special sites such as river confluences. Such mapping presents collateral information of a new character. Can this be used to achieve flood–risk estimates that are coherent throughout a river basin?