1

Feeling Uncanny

IT USED TO be the case that the first point of reference for evoking and discussing the uncanny feeling was ghosts or hauntings. Nowadays, we are more likely to talk about humanoid robots. The reason is that the uncanny has been adopted as a technical concept by roboticists. In 1970, robotics professor Masahiro Mori wrote a paper describing what he calls “the uncanny valley.”1 The uncanny valley is the sudden and dramatic dip in our felt affinity for an object as its resemblance to human beings increases. Rather than showing up as familiar and comforting, the object shows up as eerie or strange: uncanny. Mori populates this uncanny valley with corpses, zombies, and prosthetic hands, and he warns roboticists that they need to seriously confront and understand this phenomenon if they are going to build humanoid robots.
Mori’s brief description of the uncanny valley raises a number of important initial questions about how the uncanny feeling arises and the conceptual space it occupies. First, it is not obvious why affinity should drop as similarity increases. Why does affinity not increase straightforwardly? In a footnote, Mori suggests that the valley may be based on a survival instinct “that protects us from proximal, rather than distal, sources of danger.”2 Such proximal sources of danger include “corpses, members of different species, and other entities we can closely approach.”3 The uncanny affect thus resembles the feeling of disgust and is strongly associated with death and disease.
But Mori’s primary intuition seems to be that affinity drops as we revise our opinion about where an object falls on the similarity axis: “once we realize that the hand that looked real at first sight is actually artificial, we experience an eerie sensation. [. . .] When this happens, we lose our sense of affinity, and the hand becomes uncanny.”4 The story, I take it, is this: we believe that the hand is a human hand but then realize that it is artificial. Upon this realization, our sense of affinity drains away and we feel uncanny. But this story raises a number of further questions. First, it seems that what we are doing here is correcting ourselves. To correct ourselves is to reposition the object on the similarity and affinity axes. How does this repositioning bear at all on whether affinity drops as similarity increases? Further, it is not obvious that such a correction would produce an uncanny feeling. It might, of course, produce a sense of shock, surprise, or fear—heightened, of course, in any non-ideal epistemic environment, such as darkness: “If someone wearing the hand in a dark place shook a [person’s] hand with it, the [person] would assuredly shriek!”5 But why should a shock or/and a negative sense of affinity involve an eerie feeling? Why is an eerie feeling produced by making a mistake—and a mistake about the degree of human likeness in particular? Why do we not just correct the mistake and move on? It is not obvious that what we have here is either uncanny or a valley.
Even if we could explain how such mistakes produce a dip in affinity, and one which is uncanny, it is not clear why the dip should be a valley rather than an uncanny cliff. The other side of the curve (which makes the valley) signals increased human similarity and positive affinity. Mori situates ordinary dolls here. But we might have expected that with increased similarity comes an increase in mistakes and so a deepening rather than lessening of the uncanniness. Further, such mistakes should be a temporary phenomenon and need not concern robotics indefinitely. The mistaken identification, and so the uncanniness, should pass with time, insofar as time and exposure increase our familiarity and our sensitivity to relevant distinguishing features. Over time, we should find it harder to confuse robots and humans. The uncanny valley should vanish.
All this suggests that tying uncanniness to mistaken identification is extremely problematic. Mori also appeals to movement, saying that when something is in motion it heightens the severity of the valley. Why should animateness accomplish this, as opposed to merely increasing the degree of human likeness? Does animateness, as a fundamental category, somehow figure in the explanation of the eeriness of negative affinity? Does it distinguish those classes of mistakes that produce the uncanny feeling from those that do not? The idea of machines come to life does indeed seem to get at something of uncanniness, since such coming to life is associated with horror, which in turn is associated with uncanniness. (Mori: “If the mannequins started to move, it would be like a horror story.”)6
Mori’s essay and the subsequent discussions of the uncanny in robotics serve to raise the question of what the uncanny feeling is and how it arises. None answers the question. We know that the uncanny feeling is related to horror, surprise, fear, and disgust. It seems to arise when we make mistakes—especially in the dark, and especially about what is living and what is dead. In this chapter, I want to clarify the domain of the uncanny feeling. However, my aim will not be to identify a single set of conditions that make for uncanniness. Rather, I want to draw out and loosen up our sense of the semantic field surrounding the uncanny so that we may be sufficiently limber to adapt ourselves to Heidegger’s novel and robustly ontological use of the term. I begin where Mori’s essay leaves off, at the beginning of a philosophical exploration of the uncanny feeling.

Jentsch’s Uncanny

In his 1906 essay “On the Psychology of the Uncanny,” Ernst Jentsch identifies the psychical condition of possibility of the uncanny affect as a lack of orientation or a psychical uncertainty.7 This uncertainty is a difficulty incorporating something into your ‘ideational sphere’: not being able to fit that thing into the world that you know and the life that you lead. ‘Lack of orientation’ suggests that this uncertainty is more basic than a lack of conceptual knowledge; it points to experiences where we just don’t get what to do with something or how to stand with respect to it. Jentsch speaks as if this uncertainty causes the uncanny affect, although he leaves the precise relation unclear. In any case, his claim is that in situations of uncertainty there will generally follow a feeling of the uncanny. The conditions of the uncanny, then, are more or less the conditions of uncertainty. By considering Jentsch’s examples, however, we will be able to specify this further: the uncertainty that produces the uncanny feeling must be an irresolvable uncertainty.
Jentsch splits the world into things old, known, and familiar and things new, foreign, and hostile. Uncertainty emerges ‘naturally’ in relation to the latter, since we have to work to figure out what kind of thing this novel object is and so how to fit it into our ‘ideational sphere.’ Jentsch suggests that literature often makes use of this kind of uncertainty in evoking the uncanny affect, since such uncertainty relies on ignorance about what the entity is, and an author can reliably produce and manipulate our ignorance. The uncertainty is more pronounced the more it has to do with the most basic distinctions that we make in organizing our world—for instance, that between the animate and inanimate. Thus the uncanny feeling is associated with apparently inanimate objects becoming animated and vice versa. The uncertainty is dispelled when we determine what kind of thing we are dealing with and thus incorporate it into the realm of the familiar. With the novel rendered familiar, the uncanny feeling is replaced by whatever affect is appropriate to the actual state of affairs.
Jentsch gives the example of a traveler: “someone sat down in an ancient forest on a tree trunk” (presumably a fallen tree trunk, or a root or branch) and “to the horror of the traveler, this trunk suddenly began to move and showed itself to be a giant snake.”8 At first, the traveler sees only what she expects to see; she locates the ‘tree trunk’ in her ideational sphere quickly and without difficulty. In doing so, the traveler fails to acknowledge those features of the entity that would allow her to recognize it as a snake. Suddenly, one of these—motility—bursts into her experience and the traveler is forced to revise her identification of the entity. In the gap before she does so, the traveler experiences the inadequacy of her initial discovery and so finds herself suspended in doubt, uncertain as to whether the movement she has perceived is the movement of the tree or that of some living entity. She feels uncanny. As soon as she grasps the entity as a snake, the traveler has made the entity familiar again and experiences the appropriate affect: fear.
Jentsch’s way of thinking about the uncanny is obviously similar to Mori’s, which also traffics in surprises about animateness. But I disagree with both, for I do not think that Jentsch’s example, as described, is a case of the uncanny affect—precisely because it is a story about a mistake. (Mori’s examples may yet be uncanny, but they will be so for a different reason.) On Jentsch’s story, the traveler mistook a snake for a fallen tree trunk. What makes this kind of error uncanny for Jentsch, given that not all errors are? Jentsch implies that we might attribute the specifically uncanny quality of this experience to an ambiguity of in/animateness, arguing that this is the kind of uncertainty that is most often uncanny. But he does not explain why this is so, or how such an uncertainty works to produce the uncanny feeling in this example. At first, we might think that what is significant about an ambiguity of in/animateness is that it involves an uncertainty regarding categories that are eminently significant for living and acting in the world. The traveler herself is at stake in her grasp of the entity as either animate or inanimate; mistaking the snake as a tree puts her in real danger (presumably in a way that misidentifying the snake as some harmless animal would not). When an entity shifts between our most basic organizing categories, we find ourselves at risk in an uncertain world. The more central an entity or way of understanding entities is to our lives, the more its instability affects us.
But is this what makes us feel uncanny? If identifying entities as either animate or inanimate is crucial to leading a human life, then having to reassess such an identification is important; it matters to us that we get it right. If we get it wrong, we must revise—and this revision will be a significant event. In the tree/snake case, we can no longer be complacent and comfortable but must flee or fight. In this transition, we may be at a loss as to how to behave towards the entity, but the affect that we experience is not the uncanny feeling. It is surprise or confusion. This may seem like the uncanny feeling because, in this case, it transitions to fear—of which Freud holds the uncanny to be a species and which Jentsch makes no effort to distinguish from the uncanny feeling.9 (Indeed, he describes the traveler’s affective condition as ‘terror,’ which is plausibly fear mingled with surprise.)10 Either such a case of mistaken identification does not involve the uncanny feeling or the uncanny feeling is simply fear mixed with surprise and confusion.
Jentsch’s other examples of the uncanny ambiguity of in/animateness are significantly different from the tree/snake case and suggest that the uncanny affect is something other than the affect experienced by the traveler. Jentsch mentions automata and epileptics during a seizure.11 Both of these present an ambiguity of in/animateness—but, importantly, one that cannot be resolved (even if it becomes familiar and so ceases to be uncanny; more on this situation in the next section).12 Even if an automaton is known to be mechanical, and the epileptic seizure to be a manifestation of ‘mechanical’ bodily processes, both remain disquieting. Our doubts persist and no amount of evidence or argument can do away with them. (Jentsch brings his tree/snake example closer to this kind of case when he suggests that it is more uncanny when the doubt is obscure and so harder to recognize and resolve.)13 Jentsch hypothesizes that the persistence of doubt and of the feeling of uncanniness is due to either “semi-conscious secondary doubts” or a “lively recollection of the first awkward impression lingering in one’s mind.”14 But here he seems to miss the point. If we have doubts, they persist because the ‘resolution’ that assigns the entity to the category of either animate or inanimate is never final. This, in turn, is because the entity in question does not fit cleanly into these categories. For many of us, the automaton or the epileptic having a seizure has features that cannot fully be accounted for by the category of ‘inanimate’ or ‘mechanical.’
I suggest that an ambiguity of in/animateness is uncanny when this ambiguity calls the very distinction between animate and inanimate into question. Consider ghosts, which do the same for the distinction between living and dead (or past and present). A ghost is neither living nor dead, neither past nor present—and so both living and dead, past and present. Ghosts are uncanny because they ambiguously span different categories, belonging to both and neither. Experiences of such anomalous phenomena reveal that the ‘ideational sphere’ or the set of categories in terms of which we make things familiar is not adequate to some entities. The very category of ‘ghost’ marks this failure, since it is a category for a certain kind of anomaly. (Thus the vocabulary of ghosts and haunting is often used to evoke the structure of uncanniness, especially by Derrida.)15 Ghosts, epileptic seizures, and automata straddle the gaps between the joints in our ways of making sense of the world and in doing so ...