When “VR therapy” isn't really VR a commentary.

What is the paper actually arguing?

Perez-Marcos's target is terminology with clinical consequences. Rehabilitation research, he argues, uses “virtual reality” for three different things: genuine virtual reality experiences, the hardware and software systems that deliver them, and off-the-shelf videogames repurposed for therapy. When trials of all three get pooled into one evidence base, the results stop meaning anything precise.

His positive proposal is a definition. A virtual reality experience worth the name has four interrelated components: immersion (the technology surrounds you), interaction (the world responds to you), sensorimotor contingencies (you perceive the world by moving, as in real life), and illusions, above all the embodiment illusion of feeling that the virtual body is yours.

A Wii bowling game on a television has interaction and little else. There's no immersion, no first-person virtual body, no embodiment. Calling it VR is, in the paper's framing, a category error that has leaked into systematic reviews.

Why would embodiment change what therapy can do?

The reason embodiment matters is mechanistic. Watching a screen engages attention; occupying a virtual body engages the brain's representation of your own body, the same machinery that mirror therapy reaches with a $20 mirror. When the virtual arm you feel to be yours moves smoothly and painlessly, that is first-person evidence about your body, not a game score.

The paper collects supporting threads from the literature: functional imaging showing motor-control regions activating during VR rehabilitation with embodied feedback, and the observation that stroke patients are, if anything, more susceptible to embodiment illusions than healthy controls, which suggests the mechanism remains available exactly in the population that needs it.

It also fits the evidence pattern the author highlights: purpose-built VR interventions for upper-limb recovery after stroke tend to show positive effects, while videogame-based interventions tend to show non-inferiority, roughly as good as conventional therapy, but not better. If the active ingredient is embodiment, that is what you'd expect: the games never contained the ingredient.

How much weight should this paper carry?

It's important to be clear about what kind of evidence this is. The paper reports no new experiment, no sample, no statistics. It is a peer-reviewed argument, closer to an editorial with citations than to a trial. Its claims about evidence patterns are the author's synthesis, not a systematic review with pre-registered methods.

That said, conceptual papers do real work when a field's terms are muddy, and this distinction has held up as a useful reading tool. The 2021 meta-analysis of VR for subacute stroke pooled interventions from Kinect games to purpose-built systems and found its headline effect vanished after publication-bias adjustment; Perez-Marcos's lens offers one candidate explanation, that heterogeneous interventions were never measuring one thing.

What does this mean if you're considering VR-based therapy?

It means the label “VR” on a therapy tells you little by itself. The questions that matter are whether the system is immersive, whether you see and control a first-person virtual body, and whether the program is built around graded, embodied movement rather than gamified distraction.

Karuna's approach sits squarely on the embodiment side of this divide: a head-mounted display, a first-person avatar that mirrors your movement, and exercises built on mirror visual feedback and graded motor imagery principles. The rationale, and the limits of the outcome data, are laid out in how it works.

None of this makes screen-based exercise games useless. Anything that makes movement practice engaging has value, and non-inferiority to conventional therapy is not nothing. The point is narrower: the mechanisms are different, so the evidence for one should not be billed to the other.