In one particular type of computerized language-learning simulation, the computer takes on the role of a participant in the target language conversation. This type of simulation has been called a computer-assisted language-learning conversational simulation (CALL CS). We argue that the validity of a CALL CS (from the language teacher's point of view) depends on its structure, whereas its reliability (necessary for the willing suspension of disbelief by the language learner) derives from an absence of highly disruptive discourse disjunctures, or communicative breakdowns, as defined by Winograd and Flores.