A relatively recent development in the generative framework is the hypothesis that there exist in syntax silent elements (SEs) that have a semantic content that is recovered by accessing their phonologically overt counterparts (cf. Kayne 2005, 2012 and Her and Tsai 2015, among others). In this paper, we provide a careful assessment of the two SEs that have been argued by O’Neill (2011) and Homer (2015) to be present in the French (ne)…que exceptive construction; namely silent rien ‘nothing’ and silent autre ‘other’. In doing so, we take to heart one of the main points made by Her and Tsai (2015) in relation to their criticism of Kayne (2012); namely, that for a proposed SE to be learnable, there cannot be any deviation in meaning from its overt counterpart. That is, the recoverability constraint assumed in the generative framework to be at work in, say, PF-deletion ellipsis, applies to all phonologically silent categories, including SEs. Additionally, as Her and Tsai argue, if semantic deviance between SEs and their phonologically overt counterparts were allowed, SEs would become ‘empirically intractable’. We argue that while positing a silent n-word in (ne)…que is faithful to the recoverability constraint on silent categories, the alleged second SE, namely, silent autre ‘other’, is not semantically equivalent to its phonologically overt counterpart in several respects. As we demonstrate, however, if one assumes instead that its overt counterpart is plus ‘more’, the recoverability requirement is restored.