We used the time correlation of shadowgraph images to determine the angle Θ of the horizontal component of the plume velocity above (below) the center of the bottom (top) plate of a cylindrical Rayleigh-Bénard cell of aspect ratio Γ ≡ D/L = 1 (D is the diameter and L ≃ 87 mm the height) in the Rayleigh-number range 7 × 10 7 ≤ R ≤ 3 × 10 9 for a Prandtl number σ = 6. We expect that Θ gives the direction of the large-scale circulation. It oscillates time-periodically. Near the top and bottom plates Θ(t) has the same frequency but is anti-correlated.PACS numbers: 44.25.+f,47.27.Te Turbulent Rayleigh-Bénard convection (RBC) is an important process that occurs in the oceans, the atmosphere, the outer layer of the sun, the Earth's mantle, and in many industrial processes.[1] Despite the seeming simplicity of the idealized laboratory experiment, i.e. a fluid between horizontal parallel plates heated from below and cooled from above, our understanding of the physical mechanism of turbulent RBC remains incomplete. [2,3,4] Much of the heat transport is mediated through the emission of hot (cold) plumes of fluid from a thin boundary layer adjacent to the bottom (top) plate, and these plumes are carried by a large-scale circulation [5,6,7,8,9,10,11] known as the "wind of turbulence". For systems with aspect ratio Γ ≡ D/L = O(1) (D is the diameter and L the height of a cylindrical cell) this wind, when time averaged, takes the form of a single convection roll filling the entire cell. The plume interaction with this circulation is a central component of the turbulent RBC problem, and yet only little is known quantitatively about plume motion and the wind. To a large extent we expect that at least the horizontal motion of the plumes is slaved to the wind velocity, since the only independent force acting on the plumes is the buoyancy force in the vertical direction. Here we present a quantitative study of plume motion and thus, by inference, of the wind direction above (below) the center of the bottom (top) plate. The measured horizontal direction oscillates time periodically, and the oscillations persist with a unique frequency over hundreds of cycles. The oscillations at the top and bottom have the same frequency but a phase which is displaced by half a cycle. We conclude that the wind is a significantly more complicated dynamical system than a simple Rayleigh-Bénard convection cell.Numerous measurements of the speed of the fluid and/or of the temperature at distinct points (i.e. of scalar quantities) were made before and revealed a timeperiodic component. [6,7,8,9,10,11] In some instances these results were interpreted as indicative of a timeperiodic plume emission from the plates. Where direct comparison is possible, we find that the measured frequencies agree quantitatively with our determination of the oscillation frequency of the horizontal wind direction. It has been shown [9] that the corresponding oscillation period is commensurate with the period of circulation of the wind. It had been suggested that periodic plu...