To understand language, we rely on mental representations of words and their meanings. What constitutes these representations? How are they learned? To address these questions, I investigate how children learn and interpret the disjunction word "or". The highly abstract and context-dependent interpretation of or challenges word learning theories and provides an exceptional opportunity to better understand how words are associated with their meanings. "Or" has several interpretations, including exclusive and inclusive disjunction. Inclusive disjunction holds when A is true, B is true, or both. For example, a waiter may ask if you would like something to eat or drink, not excluding the possibility that you choose both. Exclusive disjunction is true when only A is true, or only B is true, but not both. If the waiter later asks whether you would like to see the dessert menu or have the check, his "or" is most likely interpreted as exclusive. He is suggesting that you should choose one or the other. Given these complexities in the interpretation of "or", how do children learn it? A previous study has shown that when parents talk to their children, the majority of "or"-examples they use are exclusive. I present an annotation study on parents' speech to children that replicated this finding. Nevertheless, comprehension studies have found that preschool children understand the inclusive interpretation of disjunction around four years of age. In an experimental study with a novel paradigm, I replicated this finding in simple existential sentences. These two findings lead to a puzzle in the literature: How can children learn the inclusive interpretation of "or" if they rarely hear it? I argue that this puzzle arises in models of word learning that directly map words to their meanings, thereby ignoring accompanying linguistic and conceptual cues. I present an in-depth annotation study demonstrating that exclusive interpretations correlate with contextual cues (such as intonation and the semantic relation of the alternatives "or" combines with) in children's input. Applying supervised learning techniques to the annotated data, I found that a learner who makes use of these contextual cues can learn the inclusive as well as exclusive interpretation of disjunction from the language heard. These findings indicate that the representation of a word like "or" cannot be isolated from the linguistic and conceptual environment in which it appears. The linguistic and conceptual aspects of "or"'s environment can act as cues that aid its acquisition and interpretation. Together, these studies show that learning a function word like "or" requires richer lexical representations than are currently assumed by our theories of word learning.