This paper employs meta-analysis to determine the influence of computer-based scaffolding characteristics and study and test score quality on cognitive outcomes in science, technology, engineering, and mathematics education at the secondary, college, graduate, and adult levels. Results indicate that (a) computer-based scaffolding positively influences learning (g = 0.53), (b) studies with zero threats to internal validity had lower effect sizes than studies with two threats, (c) studies with one threat to external validity had higher effect sizes than studies with zero threats, (d) studies with no fading had higher effect sizes than studies with fixed fading, and (e) students performed better when using conceptual scaffolds than with metacognitive scaffolds. There were no differences based on study design, generic vs. specific, paired intervention, assessment level, or intended learning outcome. Meta-regression indicated that fading or lack thereof explained 30% of the variability in outcomes. The significance of this paper lies in its potential to steer scaffold designers away from fixed fading and metacognitive scaffolds, and toward studying scaffolding in authentic contexts, rather than laboratories. Furthermore, this study indicates that a more comprehensive scaffolding meta-analysis is warranted.