An analysis of global surface temperature data is described. The data originates from the European Centre for Medium Range Weather Forecasting, and covers the period 1986--1991 with a spatial resolution of 1.125 degrees. A global average temperature, Tav, defined as the area weighted average of local temperatures, has been calculated for each day in this time period. In addition to clear seasonal variations, a small monthly oscillation is visible. The data for each year has been fitted to the form Tav=A+Bx+Ccos(Dx+E), where x is the day within the year. Subtracting the fitted seasonal variation from the data isolates the remaining monthly oscillations. The magnitude of this oscillation is about 0.2¿ K. A Fourier analysis shows the time period to be 30¿3 days, and an anti-correlation between southern and northern hemisphere is observed. The oscillation is found to be stronger towards the polar regions and absent in the tropics. Possible physical and model effects are considered. ¿ American Geophysical Union 1994