Experiments were conducted to define equilibrium combined-flow bed configurations developed under a wide range of oscillatory ( U o ) and unidirectional ( U u ) velocity components, with a constant oscillation period of 8.5 sec and median grain size of 0.09 mm. In the range of flows studied ( U o , 0-0.80 m/sec; U u , 0-0.25 m/sec) the bed phases are: no movement, small 2D ripples, small 3D ripples, large 3D ripples, and plane bed. Small 2D ripples are stable at low U o and U u ; by increasing either velocity component the plan-view bed configuration becomes progressively less regular as small 3D ripples develop. Small 3D ripples are stable at low to moderate U o and a wide range of U u . Under purely oscillatory flow, as U o increases beyond about 0.40 m/sec, large-scale, hummocky bed forms become stable. When a unidirectional component is applied, these bed forms quickly develop a downstream asymmetry and generally move in the direction of U u . The dip angles of the lee flanks increase with increasing U u , although they remain less than the angle of repose. At high oscillatory velocities ( U o , 0.70-0.80 m/sec), plane bed is the stable bed phase when U u > 0.05 m/sec. Much hummocky cross-stratification shows a preferred dip direction of the coset laminae. Our experiments support the idea that such cross-stratification is produced by combined flow: the large 3D ripples produced at high oscillatory velocities and small to moderate unidirectional velocities would presumably generate what might be called anisotropic hummocky cross-stratification (if we had been able to aggrade the bed during the runs). This would be the case even in dominantly oscillatory flow; the experiments show that even a weak unidirectional current of several centimeters per second superimposed upon a strong oscillatory flow results in bed forms that move predominantly in a single direction and that leave sets of laminae that dip predominantly in that direction. In runs at large U o , the purely oscillatory large-scale bed forms were quickly planed off as U u was increased from zero, but a series of almost stationary gentle undulations on the bed surface persisted for indefinitely long times where the original topographic highs were situated, even at the highest unidirectional flow investigated, 0.25 m/sec. This suggests a possible origin for the gentle undulation observed in Cretaceous parallel-laminated shallow-marine sandstones described by Arnott (1987, 1988), with spacings up to a few meters and heights not much more than a centimeter.