Despite decades of research, we still know less than we would like about the association between joint attention (JA) and language acquisition. One reason for this is that we still have not agreed how to define, operationalize, and measure joint attention. The goal of this study is to examine the impact of applying two different joint attention operationalization schemes—reflecting distinct theoretical perspectives—to the same dataset of video-recordings of semi-naturalistic toy-play interactions between 12-month-old children and their caregivers (N = 39). We identified joint attention around relevant naming events to determine how these choices affect interpretations of the role of joint attention in vocabulary acquisition. We compared a gaze-based coding scheme, consistent with associative accounts of joint attention, with a socially coordinated joint attention coding scheme, based on social-pragmatic theories that require, in addition to gaze overlap, evidence of shared awareness. We then extracted two measures from each scheme: average joint attention event duration and the temporal overlap between joint attention events and naming events (JA overlap). We found that while measures of joint attention were predictive of later expressive vocabulary above and beyond language-based measures in both coding schemes, model comparison based on AIC/BIC indicated that joint attention defined as coordinated JA was preferred over joint attention defined as gaze overlap. Furthermore, the best fitting model predicting later vocabulary favored predictors based on the coordinated JA scheme. Our results suggest that a social operationalization of joint attention leads to better predictors of later vocabulary size than a gaze-based operationalization of joint attention. In addition, the current study emphasizes the critical role of methodological choices in understanding how and why joint attention is associated with vocabulary size.