The present research presents three experiments that compare performance on a Stroop-like divided attention “detection” task with predictions drawn from models of the Stroop effect, a focused attention task. In three experiments, participants were administered tasks that employed integrated and/or separated word/colour pairs. Experiment 1 compared performance on both Stroop and detection tasks. Experiment 2 measured the impact of Stroop task training on detection task performance. Experiment 3 measured the performance impact of extended training on the detection task. Without practice, participants were equivalently fast detecting words and colours using integrated pairs, and faster detecting colours using separated pairs. After training on either task, participants were faster detecting colours than words with both stimulus types. Overall, results indicated that a model that includes both early and late attentional components best characterizes performance on the detection task.