This study focused on the effects of using metacognitive strategy cuing integrated into problem-solving activities in a web-based learning environment. Purposes of the study were to investigate: (a) differences in posttest achievement between students who received metacognitive strategies embedded as cues in engineering problem sets and students who did not receive the treatment; (b) differences in perceptions of problem-solving skills between students who received metacognitive strategies embedded as cues in engineering problem set and students who did not receive the treatment; (c) differences in thermodynamics knowledge; (d) problem-solving steps students reported using across problem sets; (e) characteristics of sampled students, and (f) students' perceptions of web-based problem sets. The sample consisted of 81 students enrolled in an undergraduate thermodynamics course. In-class lectures were scheduled twice weekly, and web-based problem sets were assigned as homework. Two groups, the treatment group using embedded metacognitive cuing and the control group not using the embedded metacognitive cuing, practiced with problem-solving activities over a fifteen-week-semester. Two-thirds through the semester, comprehensive posttest achievement scores were compared between groups. Analyses showed no significant differences between groups when metacognitive strategies were incorporated into web-based problem sets. An instrument was developed and validated to measure students' perceptions of their abilities to plan, monitor, and evaluate problems. Pre- and post testing of students' self-reported perceptions were measured. The results indicated no significant differences between groups. When differences in thermodynamics knowledge and skills between students were measured, pretest to posttest results showed equal improvement for both groups, contradicting the hypothesis those students in the treatment group would improve in skills and knowledge more than the control group. A frequency analysis revealed differences in the amount of times students' reported using engineering problem-solving steps while working through exercises. Most frequently chosen was Step Two - List Variables (91 %) and Step Seven - Solved Equations (91%).. The least chosen response was Step Four -- Made/stated Assumptions which was selected only three percent of the time. Implications from this investigation, along with previous research, facilitate definition of boundary conditions when employing metacognitive cuing in web-based learning.