The generation and growth of small water waves by a turbulent wind has been investigated in a laboratory channel. The evolution of these oscillations with fetch was traced from their inception with amplitudes in the micron range under conditions of steady air flow. The experiments revealed that the waves are generated at all air velocities in small bursts consisting of groups of waves of nearly constant frequency. After traveling for some distance downstream, these wavelets attain sufficient amplitude to become visible. For this condition a wind speed critical to raise waves is well defined. After the first wavelets appear, two new stages of growth are identified at longer fetches if the airspeed remains unchanged. In the second stage, the component associated with the dominant frequency of the wave spectrum initially grows most rapidly with fetch until it attains an upper limit of amplitude consistent with the well known equilibrium range, which appears to be universal for wind waves on any body of water. The frequency of this dominant wave in the second stage remains constant with fetch up to equilibrium, but tends to decrease with increasing wind shear on the water. In the third stage of growth, only wave components whose energy is lower than the equilibrium limit tend to increase in amplitude so that the wave spectrum is maintained at equilibrium in the high frequency range of the spectrum. We found no features of the mean air flow or its turbulence structure as characterized by the distribution of longitudinal intensity and energy spectra that could be attributed to disturbances by the first ripples. Under the shearing action of the wind, the first waves were found to grow exponentially. The growth rate agreed with the estimated from the viscous shearing mechanism of Miles (1962a) to a fractional error of 61% or less. Slightly better agreement was obtained with the viscous theory of Drake (1967) in which Miles' model is extended to include the effect of the drift current in the water induced by the wind. But for the magnitude of the currents observed in the tunnel, this improvement is believed to be insignificant.