What is happening in the valley?

A change of category. An entity in the uncanny valley is not judged as a robot doing a passable job of pretending to be human; it is judged as an abnormal human doing a bad job of seeming like a normal person. Something sufficiently non-human has its human features noticed and generates empathy; something almost human activates the observer's model of a person and fails it.

What triggers it most reliably?

A mismatch between appearance and movement. Something that looks mechanical and moves in a human way, or something that looks human and moves mechanically, produces the effect in both directions. The finding has been connected to perceptual uncertainty and to predictive coding.

Did Mori get the idea from Freud?

No, and the association came from a translation. Jasia Reichardt rendered bukimi no tani literally as uncanny valley in 1978, and the English phrase then acquired a link to Ernst Jentsch's 1906 essay on the psychology of the uncanny and Freud's 1919 extension of it. Mori was writing about robot design.

Is the mechanism established?

No. Support for each proposed account is described as inconsistent. Mathur and Reichling found that judging a face took longest for those deepest in the valley, which supports the categorisation-difficulty account, and also found that the confusion did not mediate the emotional effect, which argues against it being the mechanism.

Does it apply to things that are not human?

Yes, which is one of the stronger findings. Yamada and colleagues found negative affect at the midpoint of a morphed continuum between a cartoon dog and a real dog, and Ferrey and colleagues found the same maximum at the midpoint for human and non-human entities alike. Schoenherr and Burleigh point to a wider aversion to hybrids, including opposition to genetically modified food.

Does the valley move?

Mori's curve predicts it does. The response climbs back to ordinary human empathy once the resemblance becomes indistinguishable, so the valley is a region rather than a permanent condition. What is uncanny in one decade tends to be unremarkable in the next, and the discomfort measures the remaining gap.

Uncanny Valley

Bukimi no Tani

Masahiro Mori, a robotics professor, introduced bukimi no tani genshou in 1970: the uncanny valley phenomenon. As a robot is made more human in appearance, the emotional response becomes more positive and more empathetic, until the resemblance is almost complete, at which point it drops sharply into revulsion. Push further, past the point where the figure is indistinguishable from a person, and the response climbs back to the level of ordinary human empathy.

Studied Objects, Figures & PatternsVery common

Plotted, that is a steep fall and a steep rise, and the dip between them is the valley.

Why the uncanny valley is about categories

The most useful formulation in the material is about categories rather than about appearance. A robot in the uncanny valley is not judged as a robot doing a passable job of pretending to be human. It is judged as an abnormal human doing a bad job of seeming like a normal person.

Everything follows from that switch. If an entity looks sufficiently non-human, its human characteristics stand out and generate empathy. If it looks almost human, it activates the observer's model of another person, which carries detailed normative expectations, and against those the non-human characteristics stand out instead.

The condensed test is a mismatch: something mechanical that moves in a human way, or something human that moves mechanically. Both produce the effect, and the finding has been connected to perceptual uncertainty and predictive coding.

What the English translation added to bukimi no tani

Jasia Reichardt translated bukimi no tani literally as uncanny valley in the 1978 book Robots: Fact, Fiction, and Prediction. The English phrase then acquired an association Mori had not made, with Ernst Jentsch's psychoanalytic concept of the uncanny from his 1906 essay On the Psychology of the Uncanny, and with Sigmund Freud's 1919 extension of it in The Uncanny.

An unintended association introduced by a translator is worth recording precisely because it now looks like the concept's intellectual ancestry. Mori was writing about robot design.

Why no explanation of the uncanny valley is established

Support for each proposed mechanism is described as inconsistent, and the honest summary is that several plausible accounts exist and none is established.

Category error is one: an entity that is ambiguous is sorted slowly or incorrectly, sometimes called category failure or categorisation difficulty. It has been compared to a sorites paradox, since human and non-human are qualitatively different categories being linked by a quantitative measure of human likeness.

Conflicting perceptual cues is the related account: perceptual tension when cues to category membership disagree, experienced as eeriness in the way cognitive dissonance is experienced as discomfort.

Mathur and Reichling tested it directly and produced a result that cuts both ways. The time subjects took to judge whether a robot face was human or mechanical peaked for the faces deepest in the valley, which supports the difficulty account. But that perceptual confusion did not mediate the valley's effect on social and emotional reactions, which suggests confusion is not the mechanism.

Burleigh and colleagues found that faces midway between human and non-human produced eeriness diverging from an otherwise linear relationship. Yamada and colleagues found cognitive difficulty associated with negative affect at the midpoint of a morphed continuum, and their example, a series between a cartoon dog and a real dog, shows the effect is not confined to humans. Ferrey and colleagues confirmed the same maximum of negative affect at the midpoint for human and non-human entities alike. Schoenherr and Burleigh point to a general aversion to hybrid entities, including opposition to genetically modified organisms. Moore built a Bayesian model founded on categorical perception that accounts for perceptual conflict quantitatively.

Why the uncanny valley might be about death

The fourth account holds that an uncanny robot elicits an innate fear of death, and it is set out in four specific mechanisms rather than as an atmosphere:

a mechanism with a human facade and a mechanical interior plays on a fear that people are themselves soulless machines
androids in states of mutilation, decapitation or disassembly resemble a battlefield after a conflict and act as a reminder of mortality
most androids are copies of actual people, making them doppelgangers, which can elicit a fear of being replaced at work or in a relationship
the jerkiness of an android's movement may unsettle by suggesting a loss of bodily control

A fifth account is mate selection: stimulus-driven appraisal of uncanny stimuli activating an evolved mechanism for avoiding poor mate choices.

Where the uncanny valley is now

The examples given run from robots and animatronics through lifelike dolls to computer animation and, predominantly, images produced by artificial intelligence. Virtual reality, augmented reality and photorealistic animation have raised the stakes by making the almost-human commonplace.

Mori's curve predicts what follows. The valley is a region, not a permanent condition, and the response climbs back once the resemblance becomes indistinguishable. What is uncanny in one decade is unremarkable in the next, and the discomfort measures the gap rather than the technology.

Told apart

Often confused with Uncanny Valley

Common questions

Questions about Bukimi no Tani

What is happening in the valley?: A change of category. An entity in the uncanny valley is not judged as a robot doing a passable job of pretending to be human; it is judged as an abnormal human doing a bad job of seeming like a normal person. Something sufficiently non-human has its human features noticed and generates empathy; something almost human activates the observer's model of a person and fails it.
What triggers it most reliably?: A mismatch between appearance and movement. Something that looks mechanical and moves in a human way, or something that looks human and moves mechanically, produces the effect in both directions. The finding has been connected to perceptual uncertainty and to predictive coding.
Did Mori get the idea from Freud?: No, and the association came from a translation. Jasia Reichardt rendered bukimi no tani literally as uncanny valley in 1978, and the English phrase then acquired a link to Ernst Jentsch's 1906 essay on the psychology of the uncanny and Freud's 1919 extension of it. Mori was writing about robot design.
Is the mechanism established?: No. Support for each proposed account is described as inconsistent. Mathur and Reichling found that judging a face took longest for those deepest in the valley, which supports the categorisation-difficulty account, and also found that the confusion did not mediate the emotional effect, which argues against it being the mechanism.
Does it apply to things that are not human?: Yes, which is one of the stronger findings. Yamada and colleagues found negative affect at the midpoint of a morphed continuum between a cartoon dog and a real dog, and Ferrey and colleagues found the same maximum at the midpoint for human and non-human entities alike. Schoenherr and Burleigh point to a wider aversion to hybrids, including opposition to genetically modified food.
Does the valley move?: Mori's curve predicts it does. The response climbs back to ordinary human empathy once the resemblance becomes indistinguishable, so the valley is a region rather than a permanent condition. What is uncanny in one decade tends to be unremarkable in the next, and the discomfort measures the remaining gap.

Added 2026-08-23 · Revised 2026-08-26

Nearby in the catalogue

All of Objects, Figures & Patterns