Many people think negatively about computer mediated communication (CMC), pointing specifically to the hostile and aggressive language that has become commonplace in many online spaces. As CMC increases, it is important to better understand where, when, and why hostile content is produced. In what virtual venues is it more or less common? How does site design encourage or discourage it? Which users are more likely to produce hostile content? These are some of the questions my research addresses. The sample is composed of user-generated content (UGC) produced on six political news sites. The sample consists of 1200 comments and includes other forms of UGC that accompany the comments such as user names, profile pictures, and profile quotes. The content was analyzed both quantitatively and qualitatively to identify patterns that can help to better explain and predict the production of hostile and aggressive content.