This paper describes a corpus of task-based conversational speech produced by English and Spanish native talkers speaking English and Spanish as both a first and a second language. For cross-language comparability, speech material was elicited using a picture-based task common to each native language group. The bi-directionality of the corpus, stemming from the use of the same speakers and the same language pairing, makes it possible to separate native language factors from the influence of speaking in a first or second language. The potential for studying first language influences and non-native speech using the corpus is illustrated by means of a series of explorations of acoustic, segmental, suprasegmental and conversational phenomena. These analyses demonstrate the breadth of factors that are amenable to investigation in a conversational corpus, and reveal different types of interactions between the first language, the second language and non-nativeness.