Spandrels and trait delimitation: No such thing as "architectural constraint"

Forty years ago, Gould and Lewontin used the metaphor of a building's "spandrels" to highlight that organismal traits could be the inevitable consequence of organismal construction, with no alternative configurations possible. Because adaptation by natural selection requires variation, regarding a trait incapable of variation as an adaptation could be a serious error. Gould and Lewontin's exhortation spurred biologists' efforts to investigate biases and limitations in development in their studies of adaptation, a major methodological advance. But in terms of the metaphor itself, over the past 40 years there are virtually no examples of "spandrels" in the primary literature. Moreover, multiple serious confusions in the metaphor have been identified and clarified, for example, that the "spandrels" of San Marco are pendentives, and pendentives are perfect examples of adaptation. I look back over the sparse empirical fruits of the "spandrels" metaphor, and ask what the clarifications of the past 40 years mean for biological theory and practice. I conclude that if there is anything to be rescued from the clarified spandrels metaphor, it is not "constraint" at all. Instead, it is the still-unresolved issue of trait delimitation, which is how to parse organisms into subsets that are tractable and biologically appropriate for study.