The impact of virtual experiments on students’ scientific inquiry skills: Evidence from a three-level meta-analysis

Imagine you are a detective trying to solve a mystery that has been puzzling science teachers for decades: How do we best teach students to think like scientists? The goal isn't just to memorize facts, but to master "scientific inquiry skills"—the ability to ask good questions, build testable guesses, design experiments, crunch the numbers, and draw solid conclusions. For years, educators have been stuck in a debate between two camps. One camp champions Physical Experiments (PE), where students get their hands dirty with real beakers, wires, and microscopes. They argue that touching and feeling the real world is the only way to truly learn. The other camp pushes for Virtual Experiments (VE), where students use computers and simulations to run experiments in a digital world. They argue that digital tools are safer, cheaper, and let students focus on the big ideas without getting bogged down by messy equipment.

The big question is: Does the digital world actually help students learn to think like scientists better than the real world, or is it just a fancy substitute? This isn't just a theoretical squabble; it affects how schools spend their money and how teachers plan their lessons. If virtual labs are just as good, schools can save money and teach dangerous or expensive experiments safely. If physical labs are superior, then schools need to keep investing in real equipment. To settle this, researchers needed to look at the whole picture, not just one or two studies, because individual experiments often tell conflicting stories.

The Great Lab Showdown: Virtual vs. Physical

In this study, two researchers from Northwest Normal University decided to act as the ultimate referees. Instead of running a new experiment themselves, they gathered and analyzed data from 21 different studies published between 2000 and 2025. These studies involved 3,931 students ranging from elementary school to university. They used a super-powered statistical tool called a three-level meta-analysis. Think of this like a "meta-detective" that doesn't just look at the clues (the individual studies) but also checks if the clues were grouped together in a way that might skew the results. This method is crucial because many of the original studies had multiple groups of students or tested different skills, creating a "nested" structure that simpler math tools often mess up.

The Verdict:
After crunching all the numbers, the researchers found that Virtual Experiments (VE) show a significant small-to-moderate positive effect on students' scientific inquiry skills compared to Physical Experiments (PE) when looking at the data as a whole. In the language of statistics, they calculated a score called Hedges' g = 0.454. To put that in perspective, in the world of education, this is considered a solid, meaningful improvement—roughly equivalent to the average effect of a good educational intervention.

What the Paper Clarifies (and What It Doesn't):
The study confirms that, on average, virtual labs provide a measurable boost. However, the paper is careful to say that this overall advantage does not mean virtual labs are universally superior in every single situation. The results are "heterogeneous," meaning the effectiveness varies wildly depending on the context. The researchers emphasize that there is no single consensus that applies to every scenario; rather, the effectiveness depends heavily on what you are teaching and who you are teaching. While the overall trend favors virtual labs, the study found that specific factors like grade level or subject didn't statistically explain why the results varied across the board, though some specific groups did show stronger results than others.

The "Sweet Spots" for Virtual Labs:
The study discovered that virtual experiments shine brightest in specific scenarios:

• High School Physics: When high schoolers are learning physics, virtual labs showed a strong advantage (Hedges' g = 0.603). Physics is full of abstract concepts (like invisible forces) that are hard to see. Virtual simulations make these invisible things visible, helping students focus on the logic without getting distracted by fiddly equipment.
• The "Goldilocks" Duration: The best results happened when the virtual lessons lasted between 4 and 8 weeks. Shorter bursts didn't give students enough time to build skills, and longer stretches didn't seem to add extra value.
• The "Design" Problem: Here is the twist. When it came to the specific skill of designing controlled experiments, virtual labs barely made a difference (Hedges' g = 0.056). The researchers suggest this is because most current virtual labs are like "video games with preset levels"—they guide students through fixed steps. They don't let students build their own experiments from scratch. Physical labs, where students have to actually assemble wires and choose their own tools, might still be the king of teaching how to design an experiment.

Why the Confusion? The "Cognitive-Embodied" Balance
The authors explain this using a clever balance scale.

• Virtual Experiments are great at reducing "cognitive load." Imagine your brain is a computer with limited RAM. Real labs can be distracting (spilling water, broken wires, waiting for equipment). Virtual labs strip away the clutter, letting the brain focus entirely on the logic and the "big ideas."
• Physical Experiments are great at "embodied experience." This is the feeling of holding a tool, seeing the real reaction, and fixing a mistake with your hands. This is crucial for skills like troubleshooting and design.

The study suggests that the "winner" depends on the task. If the task is about abstract thinking (like understanding a physics formula), the virtual lab's ability to clear the mental clutter wins. If the task is about hands-on design (like figuring out how to build a circuit), the physical lab's real-world feedback wins.

How Sure Are They?
The researchers are very confident in their main finding: Virtual labs generally help, especially in high school physics. They used a rigorous method (the three-level model) that corrected for statistical errors that other studies missed, making their conclusion more reliable than previous attempts. However, they are less sure about the smaller details. For example, they found that the "subject type" (like chemistry vs. biology) might matter, but the evidence wasn't strong enough to say for sure (the statistical "p-value" was 0.079, which is close to the 0.05 threshold for certainty but didn't quite cross it). They also noted that their study didn't include enough data on university students or chemistry classes to make a definitive call for those groups.

The Bottom Line
This paper doesn't declare a total victory for computers over beakers. Instead, it suggests a partnership. Virtual experiments are a powerful tool for helping students grasp abstract concepts and logical reasoning, particularly in high school physics. But they aren't a magic replacement for everything. To truly master the art of scientific inquiry, students likely need a mix: the clarity of the virtual world for the "thinking" parts, and the messy, hands-on reality of physical labs for the "designing and building" parts. The future of science education, according to this study, isn't choosing one over the other, but knowing exactly when to use which one.