This chapter reviews the growing body of research focused on understanding the relationship between emotion and alcohol vulnerability in animal models of alcohol use disorders. Emotional status is directly modeled in rodents through analyses of ultrasonic vocalizations (USVs) with 22-kHz calls signaling negative affect and 50-kHz calls signaling positive affect. We review work with selectively bred high-alcohol drinking “P” and “HAD-1” rat lines that reveals high spontaneous rates of negative-affect USV emissions. We make the case that a focus on USV counts underutilizes the rich multidimensional acoustic properties associated with USVs. We argue that algorithms such as our WAAVES algorithm make the application of advanced statistics to the multidimensional USV acoustic properties feasible, and will allow more accurate characterization of emotional phenotypes. We briefly review some of our unpublished data that takes this approach. We show that advanced statistical techniques such as linear mixed modeling and linear discriminant function analysis reveal important properties of the data previously unexamined using call counts. For example, across three data sets (alcohol-naïve male P vs. NP, alcohol-naïve female vs. male HAD-1, and alcohol-naïve male P vs. Long Evans rats) we show that individual acoustic characteristics as well as the full complement of acoustic characteristics taken together differ across groups when the focus is on negative-affect calls but not positive-affect calls. The implications of this work for human alcohol research are discussed.