“…The first of these is that non-doubling reinforcement strategies that do not ultimately feed into JC, but that do play (discourse-)specific, and possibly only short-lived and non-generalising, negative reinforcement roles are as likely in Stage III languages as in Stage I or II languages (cf. Kiparsky and Condoravdi 2006, Mosegaard Hansen 2009, Larrivée 2010, and Willis et al 2013 for discussion of doubling reinforcements in Stage I and II languages). The second is that languages which select structurally high elements to serve as reinforcers (Stage II), which then become obligatory concord elements (Stage III), do not seem to proceed to Stage IV as readily as languages in which low reinforcers are reanalysed as concord elements (cf.…”