Water-storage capacity was measured inThuja occidentalis L.,Tsuga canadensis (L.) Carr., andAcer saccharum Marsh. during the dehydration of stem segments 1.5-2.5 cm in diameter. Stem water potential was measured with a temperature-corrected stem hygrometer and cavitations were detected acoustically. Water loss was measured by weight change. Dehydration isotherms consistently displayed three phases. The first phase, from water potential (Ψ) 0 to about -0.2 MPa, had a high capacitance (C>0.4kg water lost· (1 of tissue)(-1)· MPa(-1)) and we have attributed this high C to capillary water as defined by Zimmermann (1983, Xylem structure and the ascent of sap, Springer-Verlag). The second phase from Ψ=-0.5 to about -2.0 had the lowest C values (<0.02 kg·l(-1)·MPa(-1)) and was accompanied by a few cavitation events. This phase may have been a transition zone between capillary storage and water released by cavitation events as well as water drawn from living cells of the bark. The third phase also had a high C (about 0.07-0.22kg·l(-1)·MPa(-1)) and was associated with many cavitation events while Ψ declined below about -2.5 MPa; we presume the high capacitance was the consequence of water released by cavitation events. We discuss the ecological adaptive advantage of these three phases of water-storage in trees. In moist environments, water withdrawn from capillary storage may be an important fraction of transpiration, but may be of little adaptive advantage. For most of the growth season trees draw mainly on elastic storage, but stem elastic storage is less than leaf elastic storage and therefore unlikely to be important. In very dry environments, water relased by cavitation events might be important to the short-term survival of trees.