Reconstructions of Proto-Turkic must account for the correspondence between minimal pairs distinguished by vowel length contrasts in some modern varieties and their cognate pairs distinguished by laryngeal features of stem-final consonants in others (Poppe 1965). The accounts given in various reconstructions generally hold that, at some stage, primary vowel length gave rise to the development of lenis (potentially voiced) stops in certain contexts (Tekin 1975, Johanson 1986, Kabak 2004, Başdaş 2007, Semedli 2012). Variation exists across the Turkic languages with regard to patterns of both contrastive vowel length (Anderson 1998) and laryngeal contrasts in post-vocalic obstruents (Dwyer 2000), necessarily complicating reconstruction. Here I propose an alternative view with regard to these correspondences, wherein vowel length contrasts occurring before stem-final obstruents reflect an earlier voicing contrast in final obstruents. This proposal is grounded in two phonetically-based, cross-linguistically common phenomena: final devoicing (Blevins 2006), and the ‘voicing effect’ (Chen 1970), whereby vowels are longer before voiced consonants.