Previous studies on the effectiveness of virtual laboratories for learning have shown inconsistent results over the past decade. The purpose of this research was to explore the effects of a virtual laboratory and meta-cognitive scaffolding on students' data modeling competences. A quasi-experimental design was used. Three classes of eighth graders from southern Taiwan participated in this research and were assigned to the Experimental Group ? (EG ?), the Experimental Group ? (EG ?), and the Control Group (CG). EG ? (n=25) received the virtual laboratory and meta-cognitive scaffolding in the teaching and learning. EG ? (n=28) received the virtual laboratory only in the teaching and learning. The CG (n=27) received the lecture with the cookbook laboratory. The teaching unit was Heat and Specific Heat, and the teaching time for the three groups was six lessons (of 45 minutes each). The Data Modeling Competences Test (DMCT) designed by the research team was used as the data collection instrument. The results showed that the virtual laboratory and meta-cognitive scaffolding had effects on students' data modeling competences. This research shows the importance of the meta-cognitive scaffolding strategy for virtual laboratories when conducting data modeling teaching.