Abstract The high frequency of mandibular retromolar spaces among the Neandertals is often cited as a derived character relative to Early and Middle Pleistocene members of the genus Homo . Four explanations have been put forward relating variation in Neandertal dentofacial variables to the high frequency of retromolar spaces: (1) an anterior migration of the dental arcade (Coon, 1962; Howells, 1975; Wolpoff, 1980); (2) a posterior retreat of the zygomatic and anterior ramal regions relative to a fixed molar position (Trinkaus, 1987); (3) a shortening of the dental arcade due to mesiodistal molar diminution (Rak, 1986); and (4) a shortening of the dental arcade due to a combination of anterior migration of the postcanine dentition and posterior migration of the anterior dentition (Spencer & Demes, 1993). Analysis of the temporal and geographical distributions of retromolar space frequencies for available fossil mandibles ( n =77) reinforces the interpretation of a high frequency of retromolar spaces being derived in later Pleistocene Neandertal clade specimens relative to Early and Middle Pleistocene Homo . Yet, the contrast between these late archaic humans and early modern humans (especially from the Near East) is less pronounced, removing this feature from the list of possible Neandertal autapomorphies. Logistic regression analysis of a sample subset ( n =22) for which mandibular length, dental arcade length and ramus breadth could be measured demonstrates that retromolar spaces are best seen as a combined result of reduced dental arcade lengths (i.e. mesiodistally reduced postcanine dimensions) and reduced ramus breadths, in the context of little or no reduction in overall mandibular length. While the proximate spatial mechanisms accounting for higher frequencies of retromolar spaces and related facial features among the Neandertals are straightforward, ultimate causal mechanisms are more elusive. Posterior dental reduction and the associated reduction in zygomatic and ramal dimensions can be explained, at least in part, as an aspect of masticatory gracilization, probably a result of technological shifts in food processing. The maintenance of total facial prognathism, and hence mandibular length, has proven more difficult to understand.