Fifty-five college students were assessed on four types of cognitive and metacognitive measures in math, language, and visual/spatial tasks. They were measures on self-reported general metacognitive skills, predicted performance, metacognitive behaviors, and actual performance. The students' self-reported assessments of their ability on a general and specific level were positively correlated with their performance on the problem solving tasks. Students' successful performance were positively correlated with knowledge-based metacognitive behaviors. On predicted performance, students were more accurate in predicting their performance on math and language tasks. On visual/spatial tasks, there was a tendency for the students to overpredict their performance. It was concluded that college students' self-reported assessments of their own ability on a general and specific level in problem solving tasks were good indicators of their level of performance on such tasks. Successful performance was not dependent on the amount of cognitive monitoring per se, but on the type of strategies used. The accuracy of prediction was associated not only with one's ability but it varied with the task as well. (Author)