A detailed suite of seafloor heat flow measurements and seismic reflection profiles has been completed in a young (circa 1 Ma) area on the eastern flank of the Juan de Fuca Ridge that is characterized by unusually smooth basement topography and uniform sediment cover. Measurements spaced nominally 100 m apart along one 5‐km‐long line segment define a coherent pattern of heat flow variation. The profile exhibits a series of four maxima and minima with an average half‐wavelength of 600 m and an amplitude of variation of 35 mW m−2, roughly 15% of the average background heat flow of 270 mW m−2. The heat flow variations are uncorrelated with local basement topography or sediment thickness variations and may reflect cellular convection in the extrusive layer of the igneous oceanic crust. This layer, which is bounded above by low‐permeability sediments and below by low‐permeability intrusive rocks, is imaged locally along a multichannel seismic reflection profile and estimated to be about 600 m thick. Temperatures estimated at the top of this layer by extrapolation of the seafloor heat flow measurements average roughly 40°C and vary between adjacent heat flow maxima and minima by only about 7 K. These observations, together with a series of numerical simulations of hydrothermal circulation in a confined, permeable upper crustal layer, provide quantitative insights into the convection process. Values of upper crustal permeability used in the simulations ranged from below the critical value required for convective instability to roughly 2 orders of magnitude above. Results show the amplitudes of lateral seafloor heat flow variability, upper basement temperature variability, and fluid‐pressure variability to be strongly dependent on permeability. For example, the amplitude of heat flow variations is predicted to increase rapidly above critical conditions, from zero at a subcritical permeability (k) of 2 × 10−14 m2 to 75 mW m−2 at k = 5 × 10−14 m2 and to a maximum of 90 mW m−2 at k = 8 × l0−14 m2. Predicted variability then falls roughly as k−1/2, reaching the level observed in the Juan de Fuca Ridge flank study area at a permeability of 2 × 10−12 m2. Although a value at near‐critical conditions is also allowed by the results, the higher value is probably correct, for it is known from other observations that hydrothermal circulation persists in crust of much greater age, despite the effect of chemical alteration, which reduces permeability, and thermal aging, which reduces buoyancy driving forces. The value of average permeability thus estimated for the upper oceanic crust is more than 1 order of magnitude greater than values determined in deep‐ocean boreholes. The borehole values may be lower because they are representative of older and/or more highly altered crust or because they do not correctly represent the permeability at the full scale of the convective system.