An analysis of 38 years of wind data from three sites along the South African east and south coasts is made to determine interannual variability. Different period bands are investigated, and it is found that the northernmost site (Durban, at about 30øS) differs markedly from the two southern sites (Port Elizabeth and Cape Town, at about 34øS). At the latter sites interannual trends exist in the yearly averaged wind directions, while for Port Elizabeth the principal axis orientations including winds lying within the "weather band" also show such a trend. It appears that a major readjustment occurred at the two southern sites during the very strong E1 Nifio Southern Oscillation (ENSO) event in 1982/1983, with abrupt changes of up to 30 ø in the wind directions being registered. On the other hand, there are no clearly identifiable trends in the wind speeds, and longer time series will be needed to establish correlations with weaker ENSO events. INTRODUCTION Interannual variability involves both long-term trends, and events which take place at intervals of years. Various parameters can be used to analyze this variability, and in particular air temperature records have proved very valuable [Hansen and Lebedeff, 1987]; on the other hand, the E1 Nifio Southern Oscillation (ENSO) was identified from a difference in air pressures across the Pacific Ocean [Philander, 1983]. Wind measurements are more complex, involving both speed and direction, and by comparison with air temperature there are limited data available. While many weather and climate parameters are closely correlated, it is nonetheless true that wind data can give a different perspective on climate variability. The ENSO events are associated with changes in global wind belts, but cause concomitant changes in precipitation and temperatures [Quinn et al., 1987; EnfieM, 1989]; it should be noted that there is not unanimity in deciding whether and when some ENSO events occurred, and how severe they were. The core ENSO interactions occur along the equator, and it is important to identify teleconnections with areas remote from the dominant effects in the Pacific region. Shannon et al. [1986] failed to find conclusive evidence for an ENSO effect off the southwest coast of southern Africa, and instead postulated that less frequent "Benguela Nifios" occur. In contrast, van Heerden et al. [1988] and Lindesay [1990] found positive correlations between South African summer rainfall and ENSO events, while Walker [1990] identified "Agulhas warm events" outside of the ENSO years.