Virtual Reality vs. Real Life: How Brain Neurons Light Up
Space-mapping brain neurons do not “light up” in scans, when exposed to a virtual reality (VR) similar to that used in kids’ video games, the way they do in the “real world.” Only 50 percent of those neurons—which are found in the hippocampus—become active in the virtual world.
This was according to a paper published in a recent issue of Nature Neuroscience by the team of neurobiologist Mayank Mehta, University of California Los Angeles (UCLA) professor of physics and neurology. In the experiments decribed in that paper, rats were exposed to VR while their bodies were partially immobilized, and their heads were free to move.
“I think this paper makes it clear that spatial perception depends on a synthesis of information from multiple sources, including all the senses,” Loren Frank told Bioscience Technology. Frank, a UCSF neuroscientist, was not involved in the study. “In a way, that is what makes the hippocampus remarkable: it is able to create a coherent representation of space from many different sorts of information.”
Several neuroscientists contacted by Bioscience Technology agreed the work solidifies the notion that the less mobile the head during games and experiments, the more the brain experiences these as less than life-like.
“This paper did not surprise me based on the views we expressed in our commentary paper that was published in the Journal of Cognitive Neuroscience last year,” said Dartmouth College neuroscientist Jeffrey Taube, also uninvolved with the study. “Our paper discusses the importance of movement cues being available (proprioceptive, motor, vestibular) for normal spatial cognition/perception. Without these cues available, it is not surprising that the place cell signal in the hippocampus may not be as robust or as well-defined as usual. It's like playing football with half a team. You can do it, but you won't do it as well as when you have a full team (analogous to a full constellation of available cues).”
He concluded, “This paper shows the importance of these cues for normal spatial processing.”
Mirroring the world of teen games
From the start, Mehta set out to replicate the experience of teens playing video games. “Without doubt, it is the system we have used that best approximates the conditions used in humans,” Mehta told Bioscience Technology. “In fact, we devised our VR system to mimic not only video games used for recreation, but far more importantly, the video games used in clinics to diagnose memory damage in patients. We emphasized this quite a bit in the manuscript.”
Mehta’s team focused on the hippocampus, a brain region critical for memory and making spatial “maps.” When people explore any space in real life, their hippocampal neurons generate “cognitive maps.” Neuroscientists believe that the hippocampus does this by analyzing distances between visual landmarks. Other senses—like smell and hearing—could play a role as well, but these senses were not thought to be crucial.
To test whether the hippocampus can create spatial maps using only “seen” landmarks—sans all other sensory triggers—Mehta’s team created a noninvasive VR environment and watched the response of hundreds of hippocampal neurons in rats. Rats were harnessed and placed on a spherical treadmill surrounded by a “virtual world” on video screens on all sides— a virtual environment even more immersive than IMAX for humans—in a dark isolated room. Mehta’s team analyzed the behavior of the rats and their neurons both there and in a “real” room designed to look like the virtual room.
Results were dramatically different. In the virtual world, hippocampal neurons fired randomly, as if the neurons didn’t know the rats’ locations at all. This was despite the fact that the rats behaved normally in the virtual world, and showed that they knew where they were in the virtual space.
Effectively, the cognitive map vanished in the VR environment, and more than half of the neurons that were highly active in the “real” environment turned off.
As Mehta noted, the virtual world in the experiment was very similar to virtual reality used by humans.
Neural music interrupted
Mehta’s team also looked at groups of neurons. Research has shown that neuron groups generate electrical activity in fascinatingly complex patterns key for learning. Such neurons communicate using two different languages—at once. One language is based on rhythm; the other, on intensity. Astonishingly, all hippocampal neurons speak both languages simultaneously. Mehta likens this to complementary melodies in a Bach fugue. In the virtual and real worlds, the language based on rhythm possesses similar structures, even though each is conveying different messages. It is the language based on intensity that is utterly disrupted in the virtual world.
When people engage in memory recall—or when they simply move—hippocampal activity gets rhythmic. The rhythms enable memory formation and recollection. These neurons interact with other neurons like musicians in an orchestra, Mehta likes to say. Perfect synchronization is hugely important. Learning and memory disorders may be the result when those rhythms break down.
Mehta has shown before that brain rhythms are critical for hippocampal connections to grow and strengthen with learning.
Two fundamental messages
Gyorgy Buzsaki, a New York University Neurosciences Institute professor expert in self-organizing brain rhythms, was also uninvolved with Mehta's study. He told Bioscience Technology the study offered two “fundamental” take-home messages.
“First is the reliance on internally generated patterns when the environmental and body cues are diminished (i.e., VR). It is as if the brain makes up for the missing ingredients and constructs a plan by some expectation. But its performance is inferior compared to times when the expectations are constantly verified by feedback.”
Mehta’s team, he said, emphasized this with one-to-two second-long “motifs” that neuroscientists call the "'life time of cell assemblies. This is a general rule, present not only in the hippocampus under most circumstances, such as navigation, memory and even REM sleep, but also in the neocortex, where it has been shown to be under the control of short term plasticity of the pyramidal cell-interneuron synapse. This is a fundamental principle of self-organized or internally generated brain dynamics.”
Buzsaki said the second important take-home message from the Mehta paper is the idea that “constraining the animal’s behavior introduces hard-to-interpret problems in brain activity. In the Mehta study, the body was fixed, but the head could still be moved and exploit feedback from visual parallax, vestibular, neck muscle and other information. In another, almost identical task by Dimitry Aronov and David Tank in Neuron, somewhat more freedom of movement was provided to the rat, and the similarities to RW were much higher. By extrapolation, when the head is fixed, as is the routine in many recent VR tasks, the discrepancy between real world and VR is expected to be higher than shown here in the Mehta study. Head-fixed preparation is convenient for the experimenter but should not be a goal (unless specific questions are to be addressed). Efforts should be made to build technologies that allow sophisticated measurements in unrestrained, freely behaving animals.”
Tank, head of the Princeton University Neuroscience Institute, told Bioscience Technology that he agreed. “Yes, the method in our recent Neuron paper allows the subject’s head (and body) to turn naturally in order to change directions in the virtual world; that is different than the method used in [Mehta’s] Nature Neuroscience paper. Rotation is sensed by the vestibular system, so our method retains more natural vestibular system signals during the navigation. One hypothesis is that it is this difference that results in the firing of neurons recorded in our virtual navigation to be more similar to what is observed during completely unconstrained ‘real world’ conditions.”
Tank, also uninvolved with Mehta’s work, says a “direct” comparison can’t be made between teens and video games, and the rats in Mehta’s virtual world. “But the addition of appropriate vestibular input may help explain why people find the new Oculus VR systems—which change the virtual reality display as it monitors your head rotation-- very realistic.”
Facebook acquired the Oculus VR company in 2014 for $2 billion.
"The lack of vestibular inputs, or the full set of body motions, is one possible explanation for why the brain maps broke down," Mehta concluded to Bioscience Technology. "However, another possibility is that when the rats fully rotate in the virtual world, they are simultaneously rotating in the real world, too, so they are getting spatial information, not just from the virtual visual cues, but from the non-specific cues—such as sounds and smells—of the real world surrounding the virtual world. Hence, our laboratory is doing additional studies to disentangle these possibilities, and determine how the brain puts together information from multiple senses—sounds, sights, smells, etc.—to generate the perception of reality, a fundamental question that has puzzled scientists and philosophers alike."
(This story has been updated.)