This article is based on a study which attempted to find evidence of factors influencing the kind of interaction found in small group work in language learning among young adult English teacher trainees in a developing country. The particular factors studied were the degree of ‘tightness’ or ‘looseness’ of the tasks, the degree of ‘openness’ or ‘closedness’ of the tasks, and the degree to which the tasks could be described as ‘procedural’, meaning that they led to discussions about what decisions to make, or ‘interpretive’, meaning that they led to the participants having to interpret data according to their understanding and experience. The study follows earlier studies by Barnes and Todd (1977), Long and Porter (1985), Pica and Doughty (1985), Swain (1985), Doughty and Pica (1986), and Pica (1987), and examines the data, using mostly similar categories but adding two new categories—instructional input and hypothesizing—in an attempt to characterize features of learner output. The data itself consists of task-based, small-group discussions set as part of their normal work to three small groups of trainees with the purpose of developing their language ability. The task types differ in objective and demand and the study tries to see how these differing task types may influence the kind of interaction that results. The study found no significant differences in the level of modification occurring in the three task types but found significant differences in the levels of hypothesizing and of instructional input between the interpretive tasks and the task requiring decisions about procedures. The study suggests that the level of challenge of a task, measured by its procedural or interpretive nature, may be an important variable in ensuring that the learners are pushed into framing their ideas in more novel language and thus have opportunities to ‘learn’ and not only to ‘practice’.