¿ <i>Andé o anduve</i> ? Regularization of preterit <i>andar</i> in Spain

Abstract Usage-based approaches to language variation and change can explain how token and type frequencies interact such that lower token frequency word-forms might be regularized and adapted to higher type frequency patterns. This quantitative variationist study explores the regularization of the preterit andar ‘walk’ in Spain, a phenomenon that has received little empirical attention except for one recent analysis of bogotano Spanish in a controlled production task. We advance research on this understudied topic by considering more open-ended production data, a different set of regional varieties (i.e., European Spanish), and five independent linguistic variables, the latter four of which are new to this topic: person/number, meaning/function of andar , polarity, formality, and structural priming. Data were analyzed from two corpora which offered comparable written data (e.g., websites, newspapers, blogs), with a focus on European subsets of data: El Corpus del Español (ECDE) and Sketch Engine. Mixed-effects regression revealed that numerous predictors favored andar regularization: negation, informality, regularized primed tokens, 2PL and 3SG subjects, and the meaning/function of ‘walk’. We provide a rare systematic analysis of andar regularization in European Spanish in written data and reveal that regularization is indeed constrained by a range of linguistic predictors.