Quantitative studies of sociolinguistic variation in conversational speech typically pool many instances of a variable without tracking the order in which those instances arose. In this talk, I discuss what we can learn by analyzing ordered sequences of variable tokens, using the familiar ING variable (workin’~ working) as a case study. I propose that looking at ordered sequences of observations can help us disentangle different sources of repetitiveness, such as priming and stylistic covariation, in sociolinguistic production. I present a series of simple simulations that complement the corpus data and suggest that the observed microtemporal clustering (variant repetitiveness) may stem from multiple sources. I argue that these advances require a theoretical orientation toward the question of how speakers make sociolinguistic decisions in real-time speech production.