That's only if you prize one set of observations over another. In this case, Gould was observing inconsistencies in practice and using the basic scientific practices of today to interrogate the ways in which another era's less technologically-advanced scientists used social bias in order to fill the gaps and make big leaps towards questionable conclusions.

The evolution of the method of measuring intelligence was itself the subject, not whether the data could be confirmed in the end. Disagree all you like about Gould's possible misunderstandings of older science data. He was after methodological inconsistency as a window into the social construct that supported those methods. Those observations held up well because, for the most part, the traces of bias and race in their conclusions are made clear by the fact that much of the hard data with which these scientists were working were warped by the way they handled their cadavers, their specious reasoning regarding the correlation between mental health and physical markers, because they could not actually sample blood and decode DNA and take into account social conditions that may have affected their specimens.