On the basis of data from a towed thermistor chain collected near 140 ¿W on the equator during April 1987, the zonal wavenumber and vertical structure of internal gravity waves were observed to vary significantly between wave events. Our hypothesis is that this variability is due to changes in the vertical structure of mean horizontal velocity and density. Assuming that the observed waves were the fastest growing modes for shear instability, we solve the Taylor-Goldstein equation, using different analytical basic states, including a zonal and meridional flow, to simulate the different conditions during 4 nights of intense internal wave activity. We find that while the observed waves are of finite amplitude, linear shear instability is sufficient to explain the wavelength and vertical structure of vertical displacement for most of the waves. The fastest growing, unstable, mode-one solutions have e-folding growth times of less than 10 min. These solutions show wave phase speeds and vertical structures to be highly dependent upon the velocity structure of the uppermost 40 m. Near the base of the mixed layer at a flow inflection point the kinetic energy of the mean flow is extracted for wave growth. Wave vertical displacement is maximum near this inflection point. Zonal phase speeds range from -0.8 to -0.1 m/s. The propagation direction of waves with growth rates of 75% of the maximum growth rate can range from about 45¿ north to 45¿ south of the zonal direction. The vertical wave-induced Reynolds stress divergence could explain a discrepancy in zonal momentum budgets of the upper 90 m of this region. Estimates of this stress divergence show that only about 100 days of intense internal wave activity is needed per year for these internal waves to explain estimated residuals of the mean zonal momentum budgets of this region at 50- to 100-m depth.¿ 1997 American Geophysical Union