Attention to a task, and the language it requires to be performed, can be described in relation to two theoretical models which have prompted research into the effects of task demands on learning and performance outside the field of second language acquisition (SLA). These are the SEEV (selection, effort, expectancy and value) model of selective attention, and the Multiple Resources Model of divided attention to task demands. This paper describes these models, and some of their implications for research into the effects of second language task demands on speech performance and SLA. (Contains 2 figures.)