Emoticons play a key role in digital written interactions. Since the 1980s research has highlighted their growing relevance, as they allow to convey increasingly rich emotional, social, and pragmatic information. This article contributes to this area of research by providing an analysis of emoticons as structural markers in Twitter interactions. Based on a large corpus of Italian tweets, mixed-effect models were used to investigate to what extent and how emoticons are used in this role and what variables most influence their use and their relationship with punctuation marks. Results indicate that emoticons often have the function of clause and sentence boundary marking, either replacing or integrating punctuation marks. Variables affecting their use included user's age, emoticon type, and position within the tweets. In this role their use reveals the pursuit of coherent strategies in discourse organization.