In pragmatics, "Horn implicatures" are drawn when a speaker makes an utterance more complex in order to signal a less likely meaning. For instance, if Mary asks, "Could Tim go fetch the car?", and Sam replies "Tim has the capacity to do that", instead of just saying "Tim can do that", Sam is implying that it is somehow unlikely that Tim will actually go fetch the car, even though he is capable of doing so. Such implicatures have been extensively discussed, and recently modeled within the Rational Speech Act model. This project will investigate Horn implicatures experimentally, based on an already established paradigm which has been shown to work in French-speaking adults. Using referential communication tasks with pseudo-words, it can be shown that participants' intuitions about the meaning of words follow Horn implicatures. However, we still don't know whether these intuitions rest on pragmatic inferences, or to expectations about the structure of the lexicon, since words in French follow Zipf's law of abbreviation (frequent words tend to be used less frequently). We plan to run a simple online experiment which will adjudicate between these two interpretations of our effect.