Note added in proof

Recent work by Damour, Piazza and Veneziano (2002a, 2002b) suggests that the universality of free fall (hence the equivalence principle) might be violated near =a/a~10¹². The prediction refers to pairs of test masses made of Cu and Be or Pt and Ti. The present experimental limit for Cu and Be is (Be,Cu)=(1.9±2.5)·10¹² (Su et al., 1994). In Baeßler et al. (1999) the differential acceleration between test masses in the gravitational field of the Sun (a0.6 cm s²) has a 1 statistical uncertainty a=5.6·10¹³ cm s², hence a/a~10¹². However, since this experiment compares accelerations of "earth's core" and "moon/mantle" like test bodies, the composition-dependent acceleration a_CD of the Earth and Moon towards the Sun is smaller than a because only a fraction of their mass is contained in their cores and mantles. The authors conclude that a_CD/a=(+0.1±2.7±1.7)·10¹³.