As the role of learning in complex intelligent systems has become more prominent in Artificial Intelligence, researchers have become more interested in the topic of lexical learning. Their concerns have included: how to recognize new words in speech recognition tasks (Bazzi and Glass 2000), how to extract lexical semantics from corpora (Boguraev and Pustejovsky 1996), how to link lexical semantics to constrained semantic representations (Thompson and Mooney 2003; Siskind 1996) and how to ground new lexical items in the agent’s environment (Steels 1996; Roy 2003). Psychologists have also been concerned with this issue, from a number of different viewpoints. Developmental psycholinguists have asked questions about word segmentation (Cairns et al. 1997; Christiansen, Allen, and Seidenberg 1998), influences on and from (proto)syntax (Pinker 1987), innate biases on semantics (Imai and Gentner 1997), the rapid pace of word learning in children (Carey and Bartlett 1978), and other topics. Psycholinguists studying adults have also investigated word learning, particularly in the domain of second-language learning (Mitchell and Myles 1998). There are likely a number of ways in which wisdom developed in each of these fields could be beneficially transfered. Here, we will briefly review some recent important findings in the literature of lexical acquisition, then outline an experiment being conducted that will investigate the role of statistical pattern extraction, attention, and reference on lexical representations, and conclude by discussing the results that we may find and their relevance to natural language in AI and NLP.