Bruno Gonçalves and David Sánchez's Crowdsourcing dialect characterization through Twitter subjected 50 million geotagged tweets to lexical analysis (beginning with stuff like auto/carro/coche/concho/movi) to come to this conclusion. Neither author belongs to the academic linguistic establishment, and they challenge the traditional view, which in Spain at least has defined linguistic variation to a considerable extent on a neo-feudal, territorial basis - in X we speak Xish, because that's what the regional government pays for in order to maximise central government grants and minimise audits. There are obviously grave problems with sampling (who uses Twitter? can tweets really be regarded as representative utterances? can such bold claims be based solely on word use?), but the mere fact that the dataset is pretty large and un-self-conscious gives it a huge edge over traditionalist surveys like Al meu poble en diem, and it's the most interesting (albeit incredibly short) study of modern Spanish I've seen for a while.