Online restaurant reviews contain expressions of customer expectations in prose as well as in star ratings that indicate overall customer satisfaction. In prose, one way customers communicate that expectations are or are not met is through a grammatical construction called adversative connectives (ACs) (i.e., constituents such as but, although, however, and even though). In the present study, we examine the relationship between star ratings and customers’ use of ACs by employing a combination of content analysis, mixed-effects models, and thematic analysis in a corpus of nearly 35,000 online reviews for restaurants located in the United States. The results reveal an important way customers communicate their (dis)satisfaction online. Specifically, the statistical modeling indicates that the ACs used and the content they emphasize have a significant relationship with star ratings. Restaurant owners can use these findings to focus on the most important information in customer reviews, especially when they are sifting through many reviews or through reviews for which no summative rating is provided.