Glacial sea surface temperatures (SSTs) simulated by an atmosphere-mixed layer ocean model are compared with those reconstructed by the Climate: Long-Range Investigation, Mapping, and Prediction (CLIMAP) Project using planktonic microfossils. Two methods of comparison are employed. The first is global and uses the subjectively analyzed (i.e., hand contoured, then digitized) data set published by CLIMAP. The second is restricted to only those discrete locations where the CLIMAP sediment cores were taken. Both methods indicate that many aspects of the reconstructed glacial SST changes are simulated reasonably well by the model, although there are areas of disagreement. The extent of the disagreement appears smaller when the SSTs are sampled at discrete locations, because the largest discrepancies occur in the subtropical Pacific where data are sparse. When examined separately for each ocean, the magnitude of the disagreement in low latitudes roughly corresponds to the magnitude of the uncertainties in SST estimation using planktonic microfossils, being largest in the Pacific and smallest in the Atlantic. Because the largest discrepancies occur where uncertainties in estimation are large, no clear determination of whether the climate model exaggerates or underestimates low-latitude climate sensitivity appears possible. Nonetheless, sampling at discrete locations may be the best procedure for evaluating climate model performance, because errors associated with extending analyses to data-void areas can be avoided and uncertainties associated with inadequate spatial sampling made more evident. ¿ American Geophysical Union 1996