The Bermuda tide gauge record extends back to the early 1930s. That sea level fluctuations there are highly coherent with dynamic height from hydrographic data has two interesting implications. First, it should contain information about the low-frequency circulation of the Atlantic. Furthermore, because dynamic height contains information on heat storage, it might, on the limited timescales accessible in the record, also contain clues about climate. A simple model of wind forcing of the Atlantic from the African coast to Bermuda uses the Levitus mean density data to estimate the long Rossby wave speed as a function of longitude. Sea level and thermocline variability estimated this way are in remarkably good agreement with observations at periods of more than a few years duration. The peak-to-peak sea level signal is ∼18 cm, which is nearly 25% of the slope across the Gulf Stream at this latitude. The model results suggest that the variability is largest somewhat to the east of Bermuda; fluctuations of ∼10 cm extend as far east as ∼35°W. One surprising result is that at the longest periods in the COADS data, the wind curl has a double-peak structure in longitude. That is, there is a significant amount of power on the eastern side of the ocean as well as near Bermuda. Therefore, it is essential to use the full horizontal resolution of the wind data; using the mean curl across the Atlantic turns out not to be a good way to estimate thermocline variability. One might wonder if the wind data are reliable at these long periods were it not for the good agreement between the results and observed sea level. The power in wind variability increases out to ∼500 months, although with little statistical reliability. Sea level variability however, appears to peak at somewhat shorter periods. Although it is pushing the resolution of the data, this result is a limitation imposed by the basin width scale. The power in the model ocean's response to wind forcing is nearly an order of magnitude larger during the first half of the record (1952–69) than during the second (1970–86). It is likely that significant changes in buoyancy forcing by the atmosphere are coherent with changes in wind. Nevertheless, these results suggest that the variability in sea level—and so in deep temperature—can perhaps be accounted for without invoking changes in stored heat of the deep ocean.