The plate cooling model for the thermal evolution of the oceanic upper mantle has been widely accepted to explain observed variations of depth to oceanic basement and conductive heat flow with the age of the seafloor. Several estimates of “best fitting” plate model parameters, derived from depth, heat flow, and age data, have been proposed, but the viability of the plate model itself has not been rigorously evaluated. We have used published mean depths and depths to basement at Deep Sea Drilling Project/Ocean Drilling Program (DSDP/ODP) drilling sites to test the plate cooling model based on two criteria: First, viable plate models must have coefficients that are consistent with the slope of the corresponding root t line because the half‐space (or root t) subsidence of young seafloor is implicit in the plate model (i.e., the slope of the root t line can be calculated directly from the coefficients of the plate subsidence model). Second, any viable physical model must fit the data with an acceptable degree of systematic misfit; large systematic misfits indicate that the model cannot explain the observations. Fits of half‐space (root t) models to depth versus age data for young (< ∼80 Ma) seafloor indicate basal temperatures in the range 1300 to 1370°C. Based on the age at which the depths deviate from the root t line, the minimum plate thickness that is compatible with the best fitting half‐space models for young seafloor is 120 km. In contrast, all best fitting plate models yield systematically higher temperatures (1450 to 1470°C) and thinner plates (102–118 km). The plate model can explain the depth to basement at DSDP/ODP drill sites with a satisfactory degree of systematic misfit, but we find that there is no plate cooling model that can explain the variation of the mean depths (derived from the DBDB5 database) with age over the entire range of ages (0 to 165 Ma). Models that “best fit” the entire data set have unacceptably large systematic misfits over the entire range of seafloor age, whereas models that minimize the systematic misfit for young seafloor (0 to 81 Ma) fail for older seafloor. The plate model clearly fails to explain the observations. The cooling history of old seafloor is not simply an extension of the cooling history of young seafloor according to a simple plate cooling model. Previous studies have suggested, as an alternative to the plate model, that observed variations of basement depth with age are best explained by the combined effects of a “normal” half‐space cooling process and the dynamic and/or thermal effects of hot spots or mantle plumes. Our results are entirely consistent with the half‐space cooling model, and we find that the best reference model for “normal” subsidence is d(t) = (2600±20)m + (345±3)m (m.y.)−½ t½. We also find that the heat flow predicted by this half‐space model is consistent with the most reliable average heat flow values from the Pacific.