Abstract
It is argued, on experimental grounds, that Weinberg's second sum rule must be abandoned for vector-current spectral functions as long as the vector-meson dominance approximation is qualitatively valid.