Last year, in a lab at the University of Barcelona, an anonymous woman was fitted with headphones, a microphone, a head-mounted virtual-reality display, a motion-capture suit, and a small remote-triggered vibrator, which was positioned carefully over her larynx. With her display flipped on, the woman saw that she was in a room with hardwood floors, a window, a door, a spindly houseplant, a full-length mirror, and a white loveseat with blue and purple throw pillows that matched the wallpaper. When she gestured toward the rather kitschy oil painting of a nude hanging on the wall above the loveseat, she saw a virtual arm do the same. When she looked in the mirror, she saw a woman—herself, dressed in a pink T-shirt, bluejeans, and ballet flats. She leaned in to get a closer look, twisted sideways. The virtual avatar reproduced her every move in perfect synchrony. Then, still gazing in the mirror, the woman said, “Casa.”
Or perhaps her avatar did—the woman couldn’t be sure. This sense of confusion was precisely what the scientists running the experiment, Mel Slater and Domna Banakou, were trying to elicit. In their previous work together at the university’s Experimental Virtual Environments Lab for Neuroscience and Technology, Slater and Banakou had demonstrated that adult test subjects could be made to perceive a virtual toddler’s body as their own, in a kind of high-tech version of the so-called rubber-hand illusion. Now they were attempting to push the concept a step further, beyond the sense of body ownership and into the realm of agency. They wanted to know whether a subject who believed that a virtual body was hers would believe the same of any words that the avatar spoke.
Slater and Banakou described their experiment in the December 9, 2014, issue of Proceedings of the National Academy of Sciences. To test the voice illusion, they split speech into its component parts—sound, lip movement, and vibration of the vocal chords. At the moment that the avatar mouthed “casa,” the word sounded in the woman’s headphones and the vibrator on her throat buzzed gently. The virtual voice was higher pitched than the woman’s, but this didn’t stop her from reporting, in a subsequent questionnaire, that she believed it to be her own. “Basically, we convinced people that they were saying something that they actually weren’t,” Slater told me. The woman, he noted, was aware of the virtuality of the experience—“On a cognitive level, everybody knows it’s not themselves speaking”—but she also interpreted it as real. In fact, in a speech test at the end of the experiment, she unconsciously squeaked her own voice up several notches, as if to conform to her avatar. Slater and Banakou’s forty-three other test subjects did the same.
Virtual reality is often seen as touristic, a way to visit fantasy worlds without leaving home, but Slater believes that its most alluring application is precisely this: the chance to swap not only surroundings but also selves. “It’s a way of allowing you to step outside of your own body and experience what it’s like to be someone else for a short time,” he said. “What’s remarkable is how plastic the brain is in accepting that illusion.” He described a previous experiment in which both his lab and another group, working independently of each other, showed that, when light-skinned people inhabit a dark-skinned avatar, their implicit racial bias diminishes—although no one knows for how long the effect lasts outside the lab. Another of his illusion studies, published in the open-access journal Plos One in November, helped a group of young women boost their own self-esteem. In the first phase of that experiment, the women—college undergraduates who had scored highly for self-critical attitudes but were not in treatment or seeking therapy—were instructed to say a series of comforting things to a weeping virtual child, who was programmed to respond by looking up and drying her eyes. In the second phase, half of the women were sent back into the virtual space embodied as the child; the adult avatar said the same reassuring things, using the women’s own voices. The participants had been tricked, in other words, into treating themselves with more compassion than they could muster in the real world. And the effects of this self-soothing persisted: immediately following the experiment, the women showed a substantial reduction in self-criticism and an increase in positive affect.
The power of the virtual to mold the real in this way interests Slater, but, as he and Banakou acknowledge in the final lines of their paper, it must be treated with caution. When the illusion of body ownership becomes the illusion of authorship, the implications are potentially serious. Although the avatars in the voice experiment said simple one- or two-syllable words, it’s not hard to imagine a scenario in which they express something more meaningful—“I’m sorry,” for example, or “There is no God.” Would the psychological aftershock of believing that you had voluntarily said these things compel you to call an estranged friend? Would it sow doubt in your believer’s heart?
“Well, yes—it would completely undermine a whole bunch of philosophy, including Descartes,” Matthew Nudds, a professor at the University of Warwick, in England, told me. “If they really are demonstrating a mistaken sense of agency, then, wow, we are quite radically wrong about what we thought about ourselves.” The Cartesian “Cogito, ergo sum,” he said, which is critical to the Western conception of the self, could be destabilized if people began to believe that others’ statements were actually their own thoughts. Fortunately for Descartes, however, Nudds is not convinced that the experiment actually produced such a misattribution. The sticking point, for him, is the difference between a retroactive assumption of responsibility and a concurrent belief in it. The fact that Slater’s participants claimed ownership of the avatar’s voice after the fact does not mean that, in the moment, they believed it to be theirs. (Slater hopes to address this question in another version of the experiment, by using brain imaging to see what neurological processes are triggered when the voice illusion takes effect—a tricky proposition, since most scanning techniques require that the subject be immobile.)
A headset and a motion-capture suit aren’t prerequisites, of course, for changing one’s perspective. As D. Fox Harrell, the director of M.I.T.’s Imagination, Computation, and Expression Laboratory, pointed out to me, art, literature, and the social sciences have long served that purpose quite effectively. Nevertheless, Harrell is convinced of the potential of virtual-reality technology to expand our understanding of the self, particularly if Slater-style illusions can be made more sophisticated and interactive. “What are the new ways of being and seeing yourself when you’re able to blend multiple selves between the virtual and the real?” he asked. “It will be really interesting to see what kind of new social configurations emerge.” Slater’s work points toward a future in which virtual reality will increasingly shape analog reality, in which our illusions will have consequences. Philosophers, ethicists, and even poets will have to adjust accordingly.