This article illustrates the use of spoken corpora for a contrastive study of casual conversation in English and Spanish. It models an eclectic methodology for cross-linguistic comparison at the level of discourse, specifically of exchange structures, by drawing upon analytic resources from corpus linguistics (CL), conversation analysis (CA) and discourse analysis (DA). This combination of perspectives presents challenges and limitations which will be discussed and exemplified through a case study that explores agreement and disagreement sequences. English data have been retrieved from the Santa Barbara Corpus of Spoken American English (SBCSAE; cf. Du Bois et al. 2000, 2003) and Spanish data from Corpus Oral de Referencia del Español Contemporáneo (CORLEC). The case study reveals the need for spoken corpora to include complete conversations, discourse annotation, sound files and detailed contextual information. This means a step forward from corpora of spoken language to discourse corpora and a challenge for CL, CA and DA in the near future.