Recent papers show that deep ocean temperatures have increased somewhat since 1950, and that the increase is compatible with the predictions of coupled GCMs. The inference presented is that this degree of compatibility constitutes a significant test of the models. The present paper assumes that the measurements and their analysis are correct, and uses a highly simplified coupled model (i.e., an energy balance model with a mixed layer diffusive ocean) to examine whether deep ocean temperature behavior from 1950 to 2000 actually distinguishes between models of radically different sensitivity to doubled CO2. It is found that whenever models are tuned with additional forcing (for example, from aerosols) so as to replicate observed global mean temperature, the warming of the ocean temperature is largely independent of model sensitivity. There is a modest dependence on ocean diffusivity with models' behavior characteristic of large diffusivity while observations are more characteristic of low diffusivity. However, the distinctions appear to lie within observational uncertainty. There is little reason to assume that more realistic models will behave very differently in this regard. Therefore, we conclude that the behavior of deep ocean temperatures is not a test of model sensitivity, but rather a consequence of having the correct global mean surface temperature time history.