Tibeto-Burman languages show a drastic difference in the stability of independent and bound 2nd person forms. The 2nd person pronoun * naŋ is the most stable form in the Proto-Trans-Himalayan paradigm, preserved in almost every low-level clade, while 1st person is sometimes replaced. But 2nd person indexation in the verb is highly unstable. There were two distinct forms already in PTH, a suffix # -na , belonging to the original paradigm, and an innovative prefix # tV- , as well as an irrealis form which could be used with 2nd person reference. In daughter languages across the family we find further innovations in this category. These facts pose two questions. The first – why is 2nd person indexation so unstable? – can be interpreted in terms of politeness and face management. But this raises another question – if 2nd person indexation is inherently unstable, why is 2nd person independent pronominal reference not? The difference in stability reflects a difference in function. In Tibeto-Burman languages, with or without argument indexation, independent pronouns are always ‘optional’, i.e. carry some information management function such as contrast. Thus when pronominal reference to the addressee might be awkward, it can always be avoided, so there is no need to innovate face-saving substitutes for it. In contrast, in languages with argument indexation, the verbal index is obligatory, so any desire to avoid direct reference to one’s interlocutor requires adopting an alternative construction which then, over time, may grammaticalize into a new 2nd person index.