Stable and unstable person features: A structural account

Deictic paradigms show a hitherto unnoticed asymmetry in their diachronic development. While the deictic categories expressed by pronominal and possessive paradigms are overall stable, demonstrative paradigms can undergo a reorganisation that typically results, from a diachronic viewpoint, in a reduction in the number of deictic oppositions encoded in the system.1 In other words, the resulting demonstrative paradigms display a different organisation of the original deictic content and crucially no comparable reorganisation is attested in personal pronouns and possessives. In this paper, I provide an account for these different diachronic behaviours. I assume, with Harbour (2016) i.a., that demonstrative systems are defined by person features, on a par with pronominal and possessive ones. I also posit that these three classes of deictics show structural differences as to how person features are encoded in their internal structure. Thus, revisiting Polinsky’s (2018), intuition that stability is linked to structural salience, I relate the attested diachronic asymmetry to structural differences across deictic classes: specifically, I argue that person features are only salient, and therefore stable, in personal pronouns and in the indexical part of possessives, but that they are not salient, and therefore unstable, in the indexical part of demonstratives. The empirical domain is restricted to Romance languages (data from contributions in Ledgeway and Maiden 2016 and in Jungbluth and Da Milano 2015), and to the Romancebased creoles reported in the Atlas of Pidgin and Creole Structures (‘APiCS’; Michaelis et al. 2013). However, their validity seems to be wider-ranging. In what follows, only forms with interpretable and valued person features will be considered. Instead, all pronominal