Interaction data, which consist of events involving pairs of actors over time, are now commonly available at the finest of temporal resolutions. Examples of such data include cell phone call logs, twitter replies and retweets, and animal proximity logs. Existing statistical methods for modeling these data are based on point processes and directly model interaction "contagion," whereby one interaction increases the propensity of future interactions among actors. In this talk, we present an alternative approach to using temporal-relational point process models for continuous-time event data, which is able to capture a wide array of interaction patterns. Our model characterizes interactions between a pair of actors as either spurious or as that resulting from an underlying, persistent connection in a latent social network. We argue that consistent deviations from "typical" interaction behavior, rather than solely a high frequency of interactions, are crucial for identifying well-established underlying social relationships. We illustrate our methodology's ability to infer latent network structure using Bluetooth proximity data from college students during an academic school year.