Soft Robotics · University of Electro-Communications, Tokyo

Japanese Researchers Built a Soft Robotic Tongue Designed to Mimic the Exact Feel of a Real Lick

It replicates four distinct tongue motions, the softness of living tissue, and the sliminess of saliva. The researchers call it Licker. Their goal is not what you think.

There is a moment in almost every serious robotics research presentation when the room goes quiet in a way that is not entirely comfortable. At SIGGRAPH 2019, that moment arrived when Ryota Shijo and his team at the University of Electro-Communications in Tokyo placed their latest creation in front of an audience of engineers, designers, and computer scientists, and explained, without embarrassment, what it was for.

They had built a robotic tongue. And they had tried, with genuine scientific rigor, to make it feel like a real one.

The robot is called Licker. It is made from compliant elastomer materials, driven by embedded actuation systems, and controlled by biomimetic motion algorithms derived from careful analysis of how a human tongue actually moves. It replicates, according to its creators, four fundamental tongue motions. It is engineered to mimic not just the movement but the tactile sensation, the softness of real tissue and the particular sliminess that saliva produces on skin. The team says demonstrations have confirmed it can present “realistic tactile feeling being licked.”

The internet, when the footage resurfaced in 2025 and 2026, responded more or less as expected. One comment read: “Japan said ‘therapeutic licking robot’ and we all just nodded like it’s normal.” Another called it “next-level foreplay tech.” This reaction, while predictable, misses almost everything interesting about the research.

Why Licking, and Why Now

The stated purpose of Licker is deceptively simple: “to grow social bonding, regardless of species by licking.” That framing is not accidental or whimsical. It is grounded in animal behavior research that the team cites directly in their paper. Dogs lick humans. Cats groom each other. Horses nuzzle. Across the animal world, physical contact of this specific kind is among the most reliable signals of affiliation, trust, and care. The question the Licker team was asking is whether a robot could transmit that signal, if the sensation were real enough.

Japan has particular institutional reasons to take this question seriously. The country’s aging population and the structural loneliness that comes with it have driven decades of investment in socially assistive robotics. Research into robotic pets, therapeutic companion machines, and touch-based interfaces for elderly care has a long and well-funded history there. Licker sits squarely in that tradition, even if it arrived at the problem from an unusual angle.

“At first, we analyzed the human tongue motion and found four basic motions. Based on this result, we developed an originally designed tongue motion robot.”

Ryota Shijo et al., SIGGRAPH 2019 Emerging Technologies

The Engineering Problem Nobody Had Solved Before

The human tongue is, mechanically speaking, one of the most complex organs in the body. It is a muscular hydrostat, meaning it is composed almost entirely of muscle with no rigid skeletal support, which gives it a freedom of movement that is extraordinarily difficult to replicate in hardware. It can extend, retract, flatten, curl, and twist independently along its length in combinations that no conventional actuator can easily reproduce. The tongue’s surface texture changes dynamically with saliva. Its stiffness changes with the underlying muscle state. Replicating any of this at all is hard. Replicating enough of it that a person cannot distinguish it from the real thing is an entirely different challenge.

Shijo’s team approached the problem by decomposing tongue motion into its constituent parts first. They filmed and analyzed human tongue movement and identified four fundamental motion primitives: the building blocks from which all real tongue activity is constructed. They then designed an actuator system capable of executing those four motions in combination, using soft elastomers that approximate the compliance of living tissue. The surface material was engineered to replicate the friction coefficient of a real tongue on human skin, and the team added a saliva-like surface coating to match the particular slick-wet quality that makes a lick feel like a lick rather than a dry brush of silicone.

A separate team at the Tokyo Institute of Technology, led by Yuki Ishikawa and Koichi Suzumori, published complementary work in January 2025 in Frontiers in Robotics and AI, creating a pig tongue soft robot that replicates the intrinsic muscle structure of the tongue itself using thin McKibben artificial muscles embedded in silicone rubber and gel. Where Shijo’s team focused on the output experience, the tactile sensation of being licked, the Tokyo Tech team focused on the mechanical architecture, how a tongue actually generates its movement from the inside. Both lines of research are converging on the same problem from opposite directions.

What It Could Actually Be Used For

The applications that researchers take seriously fall into several distinct areas. The most clinically grounded involves swallowing disorders. Dysphagia, the medical term for difficulty swallowing, affects millions of elderly people and stroke survivors worldwide. Designing food textures and therapeutic interventions for swallowing disorders currently relies heavily on human sensory panels and clinical trials. A robotic tongue that can reliably simulate the tactile experience of a human mouth interacting with food could dramatically accelerate that work, providing a standardized, repeatable, in-vitro model where human subjects are impractical.

Rehabilitation is a second pathway. Touch is among the most therapeutically significant senses for people recovering from neurological events or dealing with conditions that disrupt normal social contact. Robots that can provide meaningful tactile interaction, not just pressure but the specific quality of a living touch, could extend the range of what robotic care assistants can offer.

Robotic pets are the third area, and arguably the most commercially viable in the near term. Japan’s PARO, a therapeutic robotic seal that responds to touch and voice, has been in clinical use in nursing homes since 2003. Licker’s technology could eventually give future robotic companion animals something PARO has never had: the ability to initiate a physical gesture of affection rather than merely receive one.

The honest caveat here: none of these applications has been clinically validated using Licker or related technology. The research is at the proof-of-concept stage. The demonstrations that Shijo’s team describes are informal tactile tests, not peer-reviewed efficacy trials. The gap between a robot that feels realistic and a robot that produces measurable therapeutic benefit is wide, and crossing it will require a different kind of research effort than building the device itself.

The Uncanny Valley of Touch

The uncanny valley is a term from robotics and animation that describes a specific kind of discomfort: the unease that arises when a human-like representation is close enough to real to trigger recognition but wrong enough to feel wrong. Most discussions of the uncanny valley focus on faces and movement. Almost nobody talks about touch.

Licker is, in a sense, an experiment in the tactile uncanny valley. The question it poses is whether the experience of being licked can cross from “clearly artificial” to “indistinguishably real,” and what happens to human psychology at each point along that continuum. A lick that feels almost right might be worse than one that feels obviously mechanical. A lick that feels completely real raises different questions about what we are willing to accept from machines, and what accepting it means for how we think about connection.

These are not questions the engineering paper answers. They are the questions the engineering paper opens.

A lick that feels completely real raises different questions about what we are willing to accept from machines, and what accepting it means for how we think about connection.

Why This Research Is Harder Than It Looks

Touch is the most poorly understood of the human senses in terms of computational and robotic replication. Vision and hearing have been the subjects of enormous investment: cameras exceed human visual acuity in many conditions, microphones can capture frequencies no human ear detects. Touch remains genuinely difficult. The skin contains multiple distinct receptor types responding to pressure, vibration, temperature, pain, and texture, often simultaneously. The tongue adds a further layer of complexity: it is both a touch organ and an active participant in the interaction, not simply a passive receiver.

Shijo’s work is part of a growing subfield of haptic robotics that has largely operated without the public attention that computer vision or language AI attracts. McKibben artificial muscles, the actuator type used in the Tokyo Tech tongue structure paper, have been in development for decades but are only recently becoming precise and miniaturized enough for tasks requiring fine tactile control. Silicone formulations have improved substantially. Embedded sensing, which would allow a robotic tongue to feel what it is touching, not just move in a particular way, is an active area of development that Licker does not yet incorporate.

The research is early. The demonstrations are informal. The clinical validation does not yet exist. And the internet will continue to make jokes. None of that changes the underlying question the work is asking, which is one of the oldest questions in the field: what does it take for a machine to make a person feel less alone? The answer, apparently, is more complicated than anyone expected. It might involve silicone, artificial muscles, and a saliva-like coating. It might involve a gesture that dogs figured out ten thousand years ago.

Ryota Shijo’s team decided the lick was worth understanding. That is, on reflection, not a strange decision at all.

Sources: Ryota Shijo, Mizuki Nagano, Izumi Mizoguchi et al., “Licker: A Tongue Robot for Representing Realistic Tongue Motions,” SIGGRAPH 2019 Emerging Technologies, ACM Digital Library; Yuki Ishikawa, Hiroyuki Nabae, Masaki Gunji, Gen Endo, and Koichi Suzumori, “Pig tongue soft robot mimicking intrinsic tongue muscle structure,” Frontiers in Robotics and AI, Volume 11, January 9, 2025, doi:10.3389/frobt.2024.1511422; UNILAB Tech, “Japanese researchers develop ‘freak bot’ that stunned commenters say is ‘next level,’” March 2026; SIGGRAPH 2019 Emerging Technologies official session listing; ResearchGate, “Licker” paper listing; Marco Marconati et al., “A soft robotic tongue to develop solutions to manage swallowing disorders,” University of Surrey / Nestlé Research Center / INRAE, arXiv, March 2020. Clinical efficacy of Licker in therapeutic settings has not been peer-reviewed or formally validated as of publication.