- The uncanny valley is the hypothesized relationship between the extent to which a humanoid resembles an actual human and the negative emotional response such a humanoid evokes.
- In 1970, the Japanese professor of robotics Masahiro Mori identified this phenomenon as bukimi no tani genshō, which was subsequently translated into English as the ‘uncanny valley’ by Jasia Reichardt.
- The violation of human norms, the salience of mortality, the avoidance of pathogens, the challenge to human identity and discord between perceptual cues may explain the uncanny valley.
- Research has been conducted on the uncanny valley in various contexts, and the means developed to overcome the phenomenon include both the increase and the decrease of humanoids’ anthropomorphic attributes.
- The role of heterogeneous phenomena, cultural constructs, and humans’ greater natural familiarity with fellow humans seem to challenge the scientific semblance of the uncanny valley theory.
In This Article
The uncanny valley can be described as the hypothesized relationship between the extent to which a humanoid entity resembles an actual human being and the emotional response such an entity evokes.
The theory holds that humanoid entities which closely resemble actual humans can provoke strangely familiar or ‘uncanny’ feelings of revulsion or eeriness in the onlookers.
The ‘valley’ herein signifies a drop in the affinity of the observer for the humanoid object; this affinity would otherwise rise in proportion to the object’s human likeness (MacDorman & Chattopadhyay, 2016).
The humanoid objects may include a variety of entities, including 3D animations, virtual reality, photorealistic animation, robotics, and lifelike dolls (MacDorman & Chattopadhyay, 2017).
As the object’s appearance gradually becomes indistinguishable from reality, the observers may feel a sense of creepiness, unease, or even disgust. The uncanny valley has manifold implications for artificial intelligence, robotics, and devices that are designed to serve and assist people.
Origin and History
In an article published in 1970, the Japanese professor of robotics Masahiro Mori identified this phenomenon as bukimi no tani genshō (Mori, 2012). Subsequently, this Japanese term was translated into English as the ‘uncanny valley’ by Jasia Reichardt in her book ‘Robots: Fact, Fiction, and Prediction’ in 1978 (Kageki, 2012).
In his original hypothesis, Mori held that the more a robot resembles an actual human, the more empathetic and positive the emotional response of the observers becomes—until the resemblance reaches a certain point (Mori, 1970).
At this point, the positive and empathetic emotional response quickly turned into intense revulsion. However, as the robot’s appearance continues to become even more human, the positive emotions return, and the level of empathy approaches the level seen in interactions between human beings.
The emotional revulsion takes place in the uncanny valley between the barely human appearance and the fully human appearance. At this stage, the robot, who looks almost like a human, constitutes a strange spectacle to the observer, who feels a sense of uneasiness or revulsion.
It must be noted that while the uncanny valley may have originated as a concept in the work of Mori, Mori was not the first person who discerned its existence.
Following the observation of a trigonocephalous viper’s face, Charles Darwin wrote in The Voyage of the Beagle: “I imagine this repulsive aspect originates from the features being placed in positions, with respect to each other, somewhat proportional to the human face; and thus, we obtain a scale of hideousness” (Darwin, 1839).
In light of Mori’s description of the uncanny valley, Darwin’s experience, as we will discuss later, seems to unveil certain cognitive processes possibly underlying this phenomenon.
The uncanny valley has been observed in a variety of spheres. Its salience in movies, however, merit special attention.
The short, animated film, Tin Toy, produced by Pixar in 1988, seems to be the first instance of the uncanny valley associated with movies. Billy, the silly infant character in the film, which was animated using the PhotoRealistic RenderMan software, elicited negative reactions from the audience (Capps, 2009).
Despite the success of the film, the issues associated with the animation of Billy would subsequently lead the film industry to take the uncanny effect seriously.
Final Fantasy: The Spirits Within
Final Fantasy: The Spirits Within was a computer-animated science fiction movie directed by Hironobu Sakaguchi. The almost natural yet imperfect depictions of humans in this photorealistic film provoked eerie reactions from viewers.
Observing the movie’s characters, Peter Travers wrote in Rolling Stone, “you notice coldness in the eyes, a mechanical quality in the movements” (Travers, 2001).
Furthermore, in The Guardian, Peter Bradshaw commented that the faces of the characters look “shriekingly phony precisely because they’re almost there but not quite” (Bradshaw, 2001).
A Christmas Carol
The 2009 computer-animated film directed by Robert Zemeckis was an adaptation of Charles Dickens’s A Christmas Carol. The reviewers described the movie’s animation as creepy.
The New York Daily News’ Joe Neumaier wrote that “the animated eyes never seem to focus” and that “for all the photorealism, when characters get wiggly-limbed and bouncy as in standard Disney cartoons, it’s off-putting” (Neumaier, 2009).
The Adventures of Tintin: The Secret of the Unicorn
The 2011 3D action-adventure film based on Herge’s The Adventures of Tintin, too, seemed to produce an uncanny effect among some viewers. The Economist’ N.B. noted that the characters’ “features are those of flesh-and-blood people” and “yet they still have the sausage fingers and distended noses of comic-strip characters” (N.B., 2011).
Moreover, The Atlantic’s Daniel D. Snyder wrote that Tintin’s original face is “now outfitted with an alien and unfamiliar visage” with “his plastic skin dotted with pores and subtle wrinkles” (Snyder, 2011).
Several theories have been proposed to account for the cognitive mechanism responsible for the uncanny valley.
The Violation of Human Norms
It is possible that the uncanny valley results from a non-human entity’s failure to measure up to the standards of an actual human being (Saygin, 2011) (MacDorman & Ishiguro, 2006).
When an object appears to be sufficiently nonhuman, its human features become more salient and tend to elicit empathy. Nonetheless, when an object looks almost human, its nonhuman features become more noticeable.
The failure of an object of this ilk to measure up to the normative expectations established for a human generates in the observer a feeling of strangeness.
This means that a robot in the uncanny valley is judged not by the standards set for a robot expected to perform various human activities but by the standards set for an actual human being expected to work like a normal person.
Hence, according to this explanation, the humanlike robot’s inability to fully resemble human norms causes the uncanny valley.
The Salience of Mortality
Another explanation holds that the uncanny valley results from an inborn fear of death coupled with culturally accepted mechanisms for coping with the inevitability of death (MacDorman & Ishiguro, 2006).
According to this theory, androids evoke our subconscious fears of replacement, reduction, or annihilation. For instance, when androids resemble actual people, they may be construed as doppelgängers.
Consequently, an observer could be afflicted with the fear of being replaced in a certain sphere of life such as in a relationship or on the job. Moreover, androids that are partially disassembled and are depicted in a state of decapitation or mutilation may evoke in the observer pictures of a battlefield in the aftermath of a conflict.
Hence, such scenes can be reminiscent of human mortality. Additionally, the mechanical interior of an almost humanlike robot can evoke the thought that human beings, too, are merely soulless machines.
Furthermore, the mechanical and jerky movements of such an android may elicit the fear of losing control over one’s own body.
The Avoidance of Pathogens
This theory holds that the uncanny valley might be activating the cognitive mechanism which had originally evolved to help humans avoid sources of pathogens (Rhodes & Zebrowitz, 2002) (Moosa & Ud-Dean, 2010) (Roberts, 2012).
According to this proposition, robots and androids in the uncanny valley may resemble human organisms with defects. Since the presence of defects implies disease, a feeling of aversion may be induced in the observers.
We know that the more a particular organism resembles a human, the more closely related genetically that organism is likely to be to humans. Moreover, greater genetic similarity is associated with a higher probability of contracting pathogenic viruses, bacteria and other parasites.
Therefore, the visual stimuli of the uncanny valley may elicit the same reactions such pathogens do. Robots and androids, for these reasons, can engender the feelings of revulsion or alarm that diseased humans and dead corpses produce.
The Challenge to Human Identity
Research shows that an increasingly anthropomorphic appearance can enhance a perceived threat to human identity and distinctiveness (Kaplan, 2004).
Consequently, the more an object resembles an actual human being, the more it seems to challenge the social identity of humans. Therefore, the perceived threat to human uniqueness could be construed as a push to redefine humanness.
This attempt to blur the categorical distinction between humans and non-humans may elicit negative reactions and feelings of unease (MacDorman & Entezari, 2015; Ferrari & Paladino, Jetten, 2016).
The Discord between Perceptual Cues
This theory posits that the visual stimuli of the uncanny valley activate contradictory cognitive representations (Elliot & Devine, 1994) (Ferrey, Burleigh & Fenske, 2015).
For instance, a humanlike figure’s possession of robotic features may induce perceptual tension, whereby an observer would be receiving contradictory cues concerning category membership. This cognitive tension constitutes a psychological discomfort similar to cognitive dissonance.
Research suggests that the deeper a robot’s face is in the uncanny valley, the longer an observer takes to gauge whether the face is actually human or not, and therefore, the greater cognitive challenge the observer encounters in categorizing the visual stimulus (Mathur & Reichling, 2016).
Studies also suggest that this cognitive challenge is associated with the negative emotions of the uncanny valley (Yamada, Kawabe & Ihaya, 2013).
Perceptual mismatch and categorization difficulty, according to this explanation, seem to be the primary causes of revulsion or eeriness (Kätsyri, Förger, Mäkäräinen & Takala, 2015).
A study was conducted in 2009 using five monkeys to evaluate the evolutionary mechanism causing the negative reactions in the uncanny valley (MacPherson, 2009).
Herein, the monkeys were given three pictures: one realistic 3D monkey face, one unrealistic 3D monkey face, and one real photo of a monkey face. The eye gaze of each monkey subject was construed as the proxy for either aversion or preference.
The monkeys looked less at the realistic 3D picture than either the unrealistic 3D picture or the real photo. This outcome was interpreted as a negative emotional reaction of the monkeys toward the realistic 3D picture.
As is the case with the uncanny valley, close yet imperfect resemblance seemed to induce aversion. The study suggested that neither human culture nor cognitive processes unique to humans could fully explain the feelings of unease in the uncanny valley. The aversion herein, therefore, seems to be evolutionary in origin.
A study that examined the relationship between the eeriness produced by a virtual character and the perception of its anthropomorphic sound and motion seemed to indicate that the anthropomorphic features of sound and motion could exaggerate the uncanniness of the character (Tinwell, Grimshaw & Williams, 2010).
It also seemed that the uncanniness arose with an increase in the perceived deficiency of human likeness in the virtual character’s facial expression, voice, and movement of the mouth during speech.
Another similar study was conducted to assess the uncanniness produced by humanlike virtual characters whose upper faces showed a perceived lack of expression for various emotions (Tinwell, Grimshaw, Williams & Nabi, 2011).
The investigation controlled individual parameters for the facial muscles for six different emotions: disgust, happiness, surprise, anger, fear, and sadness. The results of this study seemed to indicate that humanlike, animated, talking-head, and high-fidelity virtual characters would be rated as uncanny.
Notably, the same virtual characters would be rated significantly uncannier when their emotional expressivity and movement were limited to the upper face. Moreover, the level of this heightened uncanniness seemed to depend on the type of emotion being conveyed.
While sadness, surprise, fear, and disgust seemed to elicit more uncanniness, happiness and anger seemed to be associated with relatively less uncanniness.
A study that experimentally investigated whether the effects of the uncanny valley exist for images of static robotic faces asked subjects to rate the likability of two sets of faces: first, 80 robotic faces from the internet, and then, a graphically and morphometrically controlled set of faces (Mathur & Reichling, 2016).
These two sets spanned from extremely humanlike to very mechanical. To gauge the level of trust toward each face, the subjects engaged in an investment game that indicated how much they would ‘wager’ on a robot’s trustworthiness.
While the explicit rating of likability indicated a robust effect of the uncanny valley, the implicit rating of trust (via the investment game) showed an uncanny valley effect that seemed to be more dependent on context.
This outcome seemed to imply that while category confusion is associated with the uncanny valley, it does not mediate the impact on emotional reactions.
The Neuroscience behind the Uncanny Valley
A study employing functional Magnetic Resonance Imaging (fMRI) repetition suppression examined the selectivity of the human action perception system, which comprises parietal, temporal, and frontal areas for motion or/and appearance of the perceived agent (Saygin, 2011).
The subjects of the study observed the body motions of a human, a robot, or an android. The android had the biological appearance of an actual human but the movement of a mechanical robot.
While the extrastriate body area indicated greater suppression for appearance resembling humans, the action perception system was not necessarily selective for motion or appearance per se.
Conversely, different responses were found to the discrepancy between motion and appearance. In the bilateral anterior intraparietal sulcus, a key node in the action-perception system, the suppression effects for the android were stronger than those for either the human or the robot.
This outcome seemed to unveil a heightened error of prediction stemming from the brain’s negotiation of an agent that looks like a human but moves like a robot.
Avoiding the Uncanny Valley
Avoiding the uncanny valley requires an acquaintance with the causes of the valley and distinct approaches suiting the nature of each cause (Saygin, Chaminade & Ishiguro, 2010) (Saygin, 2011).
For instance, while a human being with a human voice and a robot with a synthetic voice will not evoke eeriness, a robot with a human voice would be in the uncanny valley.
In such an instance, amending the robot’s appearance to resemble that of a human might counter and remove the uncanny effect. Alternatively, replacing the robot’s human voice with a synthetic voice too would accomplish the same objective.
Moreover, an animated virtual character that looks like an actual human being but moves in a nonhuman and mechanical fashion may produce an uncanny effect (Goetz, Kiesler & Powers, 2003)(Saygin, 2011).
However, matching motion and appearance by infusing the character with either a less humanlike appearance or a more humanlike movement may level the uncanny valley.
Furthermore, because appearance by itself can cause the uncanny valley, amending appearance in certain instances would be the only way to eliminate the uncanny valley (MacDorman, Green, Ho, & Koch, 2009) (Saygin, 2011).
For example, the use of anomalous facial proportions by artists to enhance the attractiveness of computer-generated characters may produce an uncanny valley effect.
The unrealistically flawless facial features may induce eeriness in the observers. However, the uncanny valley herein may be avoided by striving for more realism in the appearance of such virtual characters.
Several objections have been raised that challenge the scientific semblance of the uncanny valley theory.
First, the uncanny valley may comprise manifold heterogenous phenomena stemming from multiple and overlapping causes (MacDorman, Green, Ho & Koch, 2009). These may involve learned or evolved circuits for face perception.
Moreover, the role played by distinct sense modalities and psychological constructs cannot be discounted (MacDorman & Ishiguro, 2006). For instance, an observer’s cultural background may significantly influence the perception of an android and, therefore, the dynamics of the valley in a particular situation.
This suggests that younger people who are more familiar with robots and androids would be less likely to be impacted by the hypothesized uncanny effect.
A second criticism holds that the uncanny valley is merely a result of humans’ greater familiarity with fellow humans and that it does not indicate a unique phenomenon (Cheetham, Suter & Jäncke, 2011).
According to this view, the uncanny valley is simply a form of information processing such as frequency-based effects or categorization. It can be argued that a certain category boundary defines ‘the valley’ in a categorization process.
Therefore, the effects of the uncanny valley may be divided between those attributable to individual exemplar frequency and category boundary (Burleigh & Schoenherr, 2014).
For instance, the negative emotional reactions that certain visual stimuli elicit could be attributed to the frequency of exposure. Research suggests that altering the frequency of practicing items shows a dissociation between the uncertainty related to category boundaries and the uncertainty based on the frequency of the exemplars.
The possibility of an uncanny valley for all degrees of human likeness too seems to challenge the theory (Hanson, Olney, Pereira & Zielke, 2005). This proposition holds that the uncanny valley can occur anywhere on a broad spectrum that ranges from the abstract to the perfectly human.
For instance, Capgras delusion, which is a delusional misidentification syndrome, can cause its sufferers to believe that people they associate with have been replaced with duplicates.
Though the sufferer may rationally acknowledge that the duplicate looks identical to the actual person, the irrational belief that the actual person has been replaced persists. In some cases of Capgras delusion, the duplicate is perceived as a robot.
Michael Lewis and Hadyn Ellis contend that the simultaneous presence of identifiability and perceived unfamiliarity herein stems from the coupling of an intact mechanism for over-recognition with a damaged mechanism for covert recognition (Ellis & Lewis, 2001).
Hence, it is possible that the uncanny valley results from problems associated with categorical perception, which is particular to how the brain processes information.
“An Uncanny Mind: Masahiro Mori on the Uncanny Valley and Beyond”. IEEE Spectrum. 12 June 2012.
Bradshaw, Peter (August 2, 2001). “Final Fantasy: The Spirits Within – Film – The Guardian”. The Guardian.
Burleigh, T. J.; Schoenherr, J. R. (2014). “A reappraisal of the uncanny valley: categorical perception or frequency-based sensitization?”. Frontiers in Psychology. 5: 1488. doi:10.3389/fpsyg.2014.01488
Capps, Chris (September 29, 2009). “Cattle Rustlin” in the Uncanny Valley”
Charles Darwin. The Voyage of the Beagle . New York: Modern Library. 2001. p. 87.
Cheetham, M.; Suter, P.; Jäncke, L. (2011). “The human likeness dimension of the “uncanny valley hypothesis”: behavioral and functional MRI findings”. Frontiers in Human Neuroscience. 5: 126. doi:10.3389/fnhum.2011.00126
Elliot, A. J.; Devine, P. G. (1994). “On the motivational nature of cognitive dissonance: Dissonance as psychological discomfort”. Journal of Personality and Social Psychology. 67 (3): 382–394. doi:10.1037/0022-35126.96.36.1992
Ellis, H.; Lewis, M. (2001). “Capgras delusion: A window on face recognition”. Trends in Cognitive Sciences. 5 (4): 149–156. doi:10.1016/s1364-6613(00)01620-x
Ferrari, F.; Paladino, M.P.; Jetten, J. (2016). “Blurring Human–Machine Distinctions: Anthropomorphic Appearance in Social Robots as a Threat to Human Distinctiveness”. International Journal of Social Robotics. 8 (2): 287–302. doi:10.1007/s12369-016-0338-y
Ferrey, A. E.; Burleigh, T. J.; Fenske, M. J. (2015). “Stimulus-category competition, inhibition, and affective devaluation: a novel account of the uncanny valley”. Frontiers in Psychology. 6: 249. doi:10.3389/fpsyg.2015.00249
Hanson, David; Olney, Andrew; Pereira, Ismar A.; Zielke, Marge (2005). “Upending the Uncanny Valley”. Proceedings of the National Conference on Artificial Intelligence. 20: 1728–1729.
J. Goetz, S. Kiesler and A. Powers (2003) Matching robot appearance and behavior to tasks to improve human-robot cooperation, The 12th IEEE International Workshop on Robot and Human Interactive Communication, 2003. Proceedings. ROMAN 2003., Millbrae, CA, USA, 2003, pp. 55-60, doi: 10.1109/ROMAN.2003.1251796.
Kaplan, F. (2004). “Who is afraid of the humanoid? Investigating cultural differences in the acceptance of robots”. International Journal of Humanoid Robotics. 1 (3): 465–480. CiteSeerX 10.1.1.78.5490
Kätsyri, J.; Förger, K.; Mäkäräinen, M.; Takala, T. (2015). “A review of empirical evidence on different uncanny valley hypotheses: support for perceptual mismatch as one road to the valley of eeriness”. Frontiers in Psychology. 6: 390. doi:10.3389/fpsyg.2015.00390
Kitta MacPherson (2009-10-13). “Monkey visual behavior falls into the uncanny valley”. Proceedings of the National Academy of Sciences. Princeton University. 106 (43): 18362–18366. Bibcode:2009PNAS..10618362S. doi:10.1073/pnas.0910063106
MacDorman, K. F., Green, R. D., Ho, C. C., & Koch, C. T. (2009). Too real for comfort? Uncanny responses to computer generated faces. Computers in human behavior, 25(3), 695–710. https://doi.org/10.1016/j.chb.2008.12.026
MacDorman, K. F.; Chattopadhyay, D. (2016). “Reducing consistency in human realism increases the uncanny valley effect; increasing category uncertainty does not”. Cognition. 146: 190–205. doi:10.1016/j.cognition.2015.09.019
MacDorman, K. F.; Ishiguro, H. (2006). “The uncanny advantage of using androids in social and cognitive science research” (PDF). Interaction Studies. 7 (3): 297–337. doi:10.1075/is.7.3.03mac
MacDorman, K.F.; Entezari, S.O. (2015). “Individual differences predict sensitivity to the uncanny valley”. Interaction Studies. 16 (2): 141–172. doi:10.1075/is.16.2.01mac
MacDorman, Karl F.; Chattopadhyay, Debaleena (2017). “Categorization-based stranger avoidance does not explain the uncanny valley effect”. Cognition. 161: 132–135. doi:10.1016/j.cognition.2017.01.009
Mathur, Maya B.; Reichling, David B. (2016). “Navigating a social world with robot partners: a quantitative cartography of the Uncanny Valley”. Cognition. 146: 22–32. doi:10.1016/j.cognition.2015.09.008
Moosa, Mahdi Muhammad; Ud-Dean, S. M. Minhaz (March 2010). “Danger Avoidance: An Evolutionary Explanation of Uncanny Valley”. Biological Theory. 5 (1): 12–14. doi:10.1162/BIOT_a_00016
Mori, M. (2012). Translated by MacDorman, K. F.; Kageki, Norri. “The uncanny valley”. IEEE Robotics and Automation. 19 (2): 98–100. doi:10.1109/MRA.2012.2192811
N.B. (October 31, 2011). “Tintin and the dead-eyed zombies”. The Economist.
Neumaier, Joe (November 5, 2009). “Blah, humbug! “A Christmas Carol’s 3-D spin on Dickens well done in parts but lacks spirit”. New York Daily News.
Rhodes, G. & Zebrowitz, L. A. (eds) (2002). Facial Attractiveness: Evolutionary, Cognitive, and Social Perspectives, Ablex Publishing.
Roberts, S. Craig (2012). Applied Evolutionary Psychology. Oxford University Press. p. 423. ISBN 9780199586073.
Saygin, A.P. (2011). “The Thing That Should Not Be: Predictive Coding and the Uncanny Valley in Perceiving Human and Humanoid Robot Actions”. Social Cognitive and Affective Neuroscience. 7 (4): 413–22. doi:10.1093/scan/nsr025
Saygin, A.P., Chaminade, T., Ishiguro, H. (2010) The Perception of Humans and Robots: Uncanny Hills in Parietal Cortex. Proceedings of the 32nd Annual Conference of the Cognitive Science Society (pp. 2716-2720)
Snyder, Daniel D. (December 26, 2011). “”Tintin” and the Curious Case of the Dead Eyes”. The Atlantic.
Tinwell, A.; et al. (2010). “Uncanny Behaviour in Survival Horror Games”. Journal of Gaming and Virtual Worlds. 2: 3–25. doi:10.1386/jgvw.2.1.3_1
Tinwell, A.; et al. (2011). “Facial expression of emotion and perception of the Uncanny Valley in virtual characters”. Computers in Human Behavior. 27 (2): 741–749. doi:10.1016/j.chb.2010.10.018
Travers, Peter (July 6, 2001). “Final Fantasy – Rolling Stone”. Rolling Stone
Yamada, Y.; Kawabe, T.; Ihaya, K. (2013). “Categorization difficulty is associated with negative evaluation in the “uncanny valley” phenomenon”. Japanese Psychological Research. 55 (1): 20–32. doi:10.1111/j.1468-5884.2012.00538.x
Mori, M., MacDorman, K. F., & Kageki, N. (2012). The uncanny valley [from the field]. IEEE Robotics & Automation Magazine, 19(2), 98-100.
Moosa, M. M., & Ud-Dean, S. M. (2010). Danger avoidance: An evolutionary explanation of uncanny valley. Biological Theory, 5(1), 12-14.
Brenton, H., Gillies, M., Ballin, D., & Chatting, D. (2005). The Uncanny Valley: does it exist and is it related to presence. Presence connect, 8(1).