Moored current measurements of 11-month duration were carried out in the boundary currents east of Madagascar, near 12°S at Cape Amber where the mean current flows northwestward and near 23°S where the mean current flows approximately southward. Transports derived from the moored current measurements in the depth range 150–1100 m compare reasonably well with those derived from ship sections by Swallow et al. (this issue). At 12°S, very energetic boundary current transport variations occur in the 40- to 55-day-period band, contributing about 40% to the total transport variance, while at 23°S the 40- to 55-day-period band fluctuations contribute only 15% to the total transport variance. The fluctuations near 12°S do not seem to be caused by local wind forcing, which does not show an energy peak in this period band. A significant annual cycle cannot be detected in the moored current and transport time series despite significant variation of wind forcing over the subtropical Indian Ocean. A comparison of the observations is carried out with two different numerical Indian Ocean models, both forced by the seasonally varying winds of Hellerman and Rosenstein (1983). A reduced-gravity model gives mean boundary current transports which compare well with the observations and also shows a negligible seasonal cycle. The multilayer Geophysical Fluid Dynamics Laboratory model also shows a small seasonal cycle. The observational evidence from the western subtropical Indian Ocean appears to be similar to that from the subtropical North Atlantic east of the Bahamas-Antilles arc where also no significant seasonal boundary current response was detected, despite large annual variation of wind forcing over the ocean. The two observational situations and numerical model results for both oceans are compared.