Although a great deal of research has been done to probe the effects of task complexity variables on the specific features of L2 learners’ output along the resource-directing dimension of the Cognition Hypothesis (Robinson, 2001a, 2003, 2005), only a few studies (e.g. Gilabert, 2007 Robinson, 2001a, 2001b Yuan & Ellis, 2003) have explored the effects of the resource-dispersing variables of task complexity on L2 output. Neither is there a rich literature on the effects of mutual interaction of these variables and task condition variables on the output. In addition, few studies have directly involved learners in oral tasks on the contrary, most of the previous studies have focused on written tasks and the oral production resulting from the performances of those tasks. This study investigated the effects of resource-dispersing variables and task condition variables on the complexity of L2 output. To this end, Preliminary English Test (PET) and an interview were administered to 20 EFL learners. After ranking the scores from the highest to the lowest, two expert/expert pairs and two novice/novice pairs were chosen to perform four tasks. The tasks were sequenced from the least to the most complex and the pairs were required to perform each task at a session, one pair after another. Their performances were tape-recorded and transcribed, and the data were subjected to statistical analysis. The results of the study indicated that, no matter whether the pairs were novices or experts, their output became more and more complex as the tasks increased in complexity. This is incompatible with the claim made by the Cognition Hypothesis that task complexity along the resource dispersing variables does not lead to the complexity of the output (Robinson, 2001a, 2005).