The primary aim of the present study is to compare two sea surface temperature (SST) climatologies produced from merchant ship observations over the tropical Atlantic Ocean. The first climatology uses measurements taken between 1911 and 1972. The second uses 1964--1984 data. The second climatology is generally warmer by a few tenths of degrees. Part of the difference is due to changes in the temperature measurement techniques with time. The difference also shows, especially along shipping routes, some spatial discrepancies which are not due to instrument changes. It is demonstrated the difference is mostly caused by inhomogeneity in the spatial and temporal density of the available observations used to construct the first climatology. Using a simple technique to eliminate spurious bias, the analysis allows isolation of the long-term trend of SST anomalies in the tropical Atlantic Ocean during this century. South of the thermal equator, a warming trend occurs principally during boreal summers and extends to the present with an increasing range when calculated over the past 26-year period. In the north, such a warming is more questionable between the 1940s and 1970s. It is more evident during the past 26-year period but with a temperature range less than half of that for the southern basin. Thus large abnormal patterns have a tendency to straddle the thermal equator, including a north-south climate dipole. Three spatially limited areas with a cooling trend are suggested. The reliability of SST anomalies along the tropical section of the Europe-South America shipping track in recent years is also discussed. ¿ American Geophysical Union 1990