Emojis, introduced in the US in 2011 and now ubiquitous, are a set of iconic expressive symbols that are incredibly widespread in computer-mediated communication (CMC), especially among young people. The majority of linguistic research on emojis focuses only on the semantics; however, emerging data suggest that emojis are far more linguistically interesting than merely their semantic contributions to a sentence. Data from Twitter demonstrate that emojis can actually appear as contentful morphological units that behave according to regularly predictable morphosyntactic rules. In this paper I analyze data from several languages including English, German, and Spanish and reach several conclusions about emojis that appear as words. First, these emojis are not merely replacements in text for existing words in a language, but rather they represent the morphosyntactic of a lexically-typed stem, to which inflectional and derivational affixes can be productively added. Second, these emoji stems can undergo morphological changes such lexicalization and grammaticalization. Within pro-speech emojis, I also differentiate two sub-categories of emoji according to how they are interpreted: emojis which have both an iconic and symbolic mechanism of interpreted, and emojis that are interpreted iconically. This work introduces and strengthens the idea that pictures can be part of the morphosyntactic derivation of an utterance and that they can be represented in a language user’s lexicon.