The intensifier system is well known for its perpetual recycling of fresh innovations; however, neither qualitative nor quantitative analyses have offered a consensus on which social factors are involved in the increased use of one variant at the expense of another, nor do we know much about sites of innovation. In this paper, we delve deep into the intensifier system by considering the distinction between narrative and non-narrative discourse contexts (Labov and Waletsky 1967) and using a “small-within-large” methodology wherein a subset of data from a broad sociolinguistic study is our foundation (Tagliamonte 2008). Our results reveal that narratives have significantly higher intensification rates than non-narratives, which we interpret as a linguistic resource to increase affective meaning when performing the identity work inherent in storytelling (Schiffrin 1996). Further, the statistically significant predictors for intensifier use in narratives are predominantly semantic, involving adjective type and emotional value with no significant social factors. Yet in non-narrative discourse, syntactic factors predominate and both gender and age are statistically significant effects. Partitioning the data by discourse context uncovers additional sociolinguistic bifurcation. Indeed, a more detailed examination of the interaction of speaker age and gender reveals how critical the narrative/non-narrative contrast is in the ebb and flow of changes within this system. While younger speakers of both genders show an increase in really in narratives, in non-narratives younger women exhibit a heightened usage compared to older women (4% vs. 21%). The results for very are equally suggestive: younger women use less very in both registers but there is a sharp decline in non-narratives in particular. This suggests that innovations rise first in narratives for all speakers and then diffuse to non-narratives lead by younger women. Taken together, the findings from this study support earlier observations that greater care should be placed on the discourse embedding linguistic variation and change (see e.g., Cheshire 2005 et seq). We have demonstrated that language change actually begins and ends in stories.