|Abstract:||This paper establishes the claim that geminate sonorants are cross-linguistically marked, and furthermore, that the relative sonority of a geminate positively correlates with its markedness, i.e., the universal ranking *GEMGLIDE >> *GEMLIQUID >> *GEMNASAL holds. This ranking is supported by a cross-linguistic survey of geminate inventories (Podesva 2002; Taylor 1985) as well as by a number of phonological alternations. Second, this paper proposes that this markedness hierarchy derives from the confusability of geminacy contrasts for sonorant segments: the more sonorous a segment is, the more difficult it is to perceive its segmental duration, and hence the less perceptible its geminacy contrasts are. A perceptual experiment on Arabic is reported to support this proposal, which shows that discriminability of geminacy contrasts negatively correlates with relative sonority. The results add to a growing body of literature that claims that languages avoid making a phonological contrast that is not very perceptible, as in Adaptive Dispersion Theory (Flemming 1995; Liljencrants and Lindblom 1972; Padgett 2003) and Licensing-by-Cue (Steriade 1997).