In this paper we present some suggestive--although not conclusive--evidence that deixis plays a distinct role in grammaticalization, first suggesting through the French and Buryat data that deictics themselves may grammaticalize differently than non-deictics, then suggesting through examination of case-marking that they may trigger different grammaticalization processes in affixing elements. Using evidence from Mara and Mangarayi case-marking, we showed that deictic and optionally deictic elements can act as a group, triggering grammaticalization of a genitive case-marker different from that found on other nominals. Then, using evidence from Yakulta and Konkani, we showed that certain grammaticalization processes seem to have affected unambiguous deictics more completely than ambiguously deictic forms. French encliticization also offered evidence for this kind of deictic hierarchy, whereby processes which are obligatory on unambiguous deictics are optional on ambiguous ones.
By this means we suggest that the neglect of deixis in grammaticalization discussions, although understandable because of the difficulty of establishing a deictic function for a reconstructed form, may in fact be a mistake. After all, we know that perceptual categories such as animate/non-animate have an effect on grammaticalization, resulting in splits in grammatical case-marking which have been extensively discussed (Silverstein 1976). It is not ipso facto unreasonable that equally salient perceptual categories such as deictic/non-deictic might also have left marking splits. We believe this possibility should be further explored.