Analyses of five transatlantic hydrographic sections suggest that the annual average Atlantic meridional overturning circulation was 30% weaker at the beginning of the 21st century compared to the late 1950s. We examine the simulation of the Atlantic meridional overturning in the climate model HadCM3. In both the control experiment and simulations of the 20th century, there is no significant trend in the overturning on the timescales of the observations and a change of 30% in the annual average overturning is very unusual. We use HadCM3 to highlight uncertainties associated with the method used to derive annual average overturning from the five observations. Assuming hydrographic sections to be representative of their respective years, inferring the flow field from density via geostrophy and using constant Ekman and Gulf Stream transport, all introduce uncertainties in both the mean and variability of the reconstructed overturning. Using monthly hydrographic fields rather than annual mean fields increases the variability of the estimated model meridional overturning circulation. Continuous monitoring of the overturning is needed to enable the interannual and seasonal variability to be understood better and to allow detection and attribution studies.