This study discusses findings that replicate and extend the original work of Burns and Vollmeyer (2002), which showed that performance in problem-solving tasks was more accurate when people were engaged in a non-specific goal than in a specific goal. The main innovation here was to examine the goal specificity effect under both observation-based and conventional action-based learning conditions. The findings show that goal specificity affects the accuracy of problem solving in the same way when the learning stage of the task is observation-based as when it is action-based. In addition, the findings show that, when instructions do not promote goal specificity, observation-based problem solving is as effective as action-based problem solving. (Contains 5 figures and 1 note.)