Samples collected from a lava channel active at Kilauea Volcano during May 1997 are used to constrain rates of lava cooling and crystallization during early stages of flow. Lava erupted at near-liquidus temperatures (similar to 1150 degrees C) cooled and crystallized rapidly in upper parts of the channel. Glass geothermometry indicates cooling by 12-14 degrees C over the first 2 km of transport. At flow velocities of 1-2 m/s, this translates to cooling rates of 22-50 degrees C/h. Cooling rates this high can be explained by radiative cooling of a well-stirred flow, consistent with observations of non-steady flow in proximal regions of the channel. Crystallization of plagioclase and pyroxene microlites occurred in response to cooling, with crystallization rates of 20-50% per hour. Crystallization proceeded primarily by nucleation of new crystals, and nucleation rates of similar to 10(4)/cm(3)s are similar to those measured in the 1984 open channel flow from Mauna Loa Volcano. There is no evidence for the large nucleation delays commonly assumed for plagioclase crystallization in basaltic melts, possibly a reflection of enhanced nucleation due to stirring of the flow. The transition of the flow surface morphology from pahoehoe to 'a'(a) over bar occurred at a distance of 1.9 km from the vent. At this point, the flow was thermally stratified, with an interior temperature of similar to 1137 degrees C and crystallinity of similar to 15%, and a flow surface temperature of similar to 1100 degrees C and crystallinity of similar to 45%. 'A'(a) over bar formation initiated along channel margins, where crust was continuously disrupted, and involved tearing and clotting of the flow surface. Both observations suggest that the transition involved crossing of a rheological threshold. We suggest this threshold to be the development of a lava yield strength sufficient to prevent viscous flow of lava at the channel margin. We use this concept to propose that 'a'(a) over bar formation in open channels requires both sufficiently high strain rates for continued disruption of surface crusts and sufficient groundmass crystallinity to generate a yield strength equivalent to the imposed stress. In Hawai'i, where lava is typically microlite poor on eruption, these combined requirements help to explain two common observations on 'a'(a) over bar formation: (a) 'a'(a) over bar flow fields are generated when effusion rates are high (thus promoting crustal disruption); and (b) under most eruption conditions, lava issues from the vent as pahoehoe and changes to 'a'(a) over bar only after flowing some distance, thus permitting sufficient crystallization.