Two Wrongs Make One Right

TL;DR

The paper analyzes why science is empirically successful through the realism-antirealism debate, using Drude's classical theory of metals and the Wiedemann-Franz law as a case study. It argues that Drude's correct prediction of the Lorenz number arises not from an approximately true description of electrons but from a fortuitous cancellation of large errors in velocity and heat capacity, challenging the view that success tracks truth. It draws three lessons about coincidences, the non-monotonic relationship between truth and predictive success, and potential synergies with antirealist accounts such as predictive similarity. The discussion extends to the Sommerfeld puzzle, evaluating structural realism as a possible defense and concluding that multiple explanatory routes—coincidence, predictive similarity, and selective realism—may jointly account for scientific success without requiring universal truth-claims about all theories.

Abstract

An influential argument for scientific realism posits that, if scientific theories were not true, their empirical success would be a coincidence. Here, I show that the false Drude's theory succeeds in explaining the Wiedemann-Franz law by coincidence--a fortuitous compensation of errors.

Figures (1)

• Figure 1: a, Experimental values of the Lorenz number ($L$) for various metals measured at a temperature of 273 K (red dots), compared with the predictions of Drude's theory (blue line) and Sommerfeld's theory (green line); experimental values are from Ashcroft1976 (p. 21). b, Illustration of Drude's theory, where conduction electrons (small yellow circle) are pictured as classical particles that collide with immobile positively charged ions (large circles). c, Comparison of the Lorenz number ($L$) of silver (Ag) as determined from experimental measurements (red dot), Drude's theory (blue line), Sommerfeld's theory (green line), and the two fictitious versions of Drude's theory that include either the Sommerfeld electron velocity or Sommerfeld heat capacity (yellow lines); notice that $L$ is given on a logarithmic scale.