Discovering and mapping predicted but undocumented morphosyntactic variation through Twitter: evidence from Spanish bare count singulars with ir a

Abstract This paper investigates – to the best of our knowledge, for the first time – whether social media platforms can be used to detect and geographically map theoretically predicted but previously undocumented morphosyntactic variants. Specifically, it examines Spanish bare singular count nouns with ir a (lit. ‘to go to + [noun]’) across different national varieties of Spanish, as opposed to Standard Spanish ir al + [noun] (lit. ‘to go to the + [noun]’). Based on Twitter data ( n = 6,206), we show that such bare singular count nouns exist and are particularly prevalent in Colombian Spanish, a fact not previously reported in the literature. However, a follow-up acceptability judgment task with native Spanish speakers ( n = 226) suggests that social media data may produce false positives, especially for low-frequency phenomena. As a result, we argue that such studies should always be complemented by other methods, e.g., from experimental linguistics.