Spatial and temporal variability in sea-surface temperature (SST) is analyzed by the method of principal component analysis (PCA). Variability of satellite-derived SST from the NOAA/NASA Advanced Very High Resolution Radiometer (AVHRR) Pathfinder data over the Arabian Sea is compared to the PCA of the mixed-layer temperature fields from two different Indian Ocean models. Climatological model output is compared to Pathfinder's “averaged” year using data from 1987 to 1990. A 5-year analysis with data and model output (from 1985 to 1989) is also done. The first mode in all the studies accounts for 58.2–95.8% of the SST variability, and is identified with the seasonal warming and cooling associated with the Indian Monsoon. The second mode accounts for 20.6–31.1% and corresponds to the radiative heating of the basin. Time series of the basin's mean SST shows that the models lag Pathfinder SST by approximately one month. The climatological models fail to reproduce the SST variability, in both space and time, of the Arabian Sea. The Luther–Ji model, forced by interannual monthly winds, does much better. The main discrepancies are likely due to the models’ forcing fields underestimating the strength of the monsoon, and the vertical thermal structure not being adequate to represent the real ocean, especially in upwelling areas.