# A reappraisal of the uncanny valley: categorical perception or frequency-based sensitization?

_Tyler J. Burleigh_ [1]\*, _Jordan R. Schoenherr_ [2]

1. Department of Psychology, University of Guelph, Guelph, ON, Canada  
2. Department of Psychology, Carleton University, Ottawa, ON, Canada

**Correspondence:**  
Tyler J. Burleigh, Department of Psychology, University of Guelph, 50 Stone Road East, Guelph, ON N1G 2W1, Canada  
[e-mail: tburleig@uoguelph.ca](mailto:tburleig@uoguelph.ca)

---

The uncanny valley (UCV) hypothesis describes a non-linear relationship between perceived human-likeness and affective response. The “uncanny valley” refers to an intermediate level of human-likeness that is associated with strong negative affect. Recent studies have suggested that the uncanny valley might result from the categorical perception of human-like stimuli during identification. When presented with stimuli sharing human-like traits, participants attempt to segment the continuum in “human” and “non-human” categories. Due to the ambiguity of stimuli located at a category boundary, categorization difficulty gives rise to a strong, negative affective response. Importantly, researchers who have studied the UCV in terms of categorical perception have focused on categorization responses rather than affective ratings. In the present study, we examined whether the negative affect associated with the UCV might be explained in terms of an individual’s degree of exposure to stimuli. In two experiments, we tested a frequency-based model against a categorical perception model using a category-learning paradigm. We manipulated the frequency of exemplars that were presented to participants from two categories during a training phase. We then examined categorization and affective responses functions, as well as the relationship between categorization and affective responses. Supporting previous findings, categorization responses suggested that participants acquired novel category structures that reflected a category boundary. These category structures appeared to influence affective ratings of eeriness. Crucially, participants’ ratings of eeriness were additionally affected by exemplar frequency. Taken together, these findings suggest that the UCV is determined by both categorical properties as well as the frequency of individual exemplars retained in memory.

**Keywords:** uncanny valley, categorical perception, category learning, categorization, exemplar theory, exemplar-based, frequency-based, affect

---

## INTRODUCTION

Categorization is a critical determinant of human survival. In the absence of categories, humans would be required to learn whether each stimulus that we encountered was desirable or noxious as well as whether the conspecifics that we encountered were kin or competitors. The variability in cross-cultural folktaxonomies (Medin and Atran, 1999), color classification (Regier and Kay, 2009), and speech perception (Pisoni et al., 1982) demonstrates that while humans might have prepotent responses to ranges of stimuli, many of these distinctions can be modified or must be learned. When available within the classification system of society, these categories can be associated with strong, negative affect responses (Schoenherr and Burleigh, 2014). Thus, categories both reflect and determine one’s experience of the world.

Group membership and identity form an especially relevant class of categories for humans (for a review, see Fiske and Taylor, 2013). In the social context, repeated exposure to individuals within a group can increase affiliation and conformity (for review, see Bond and Smith, 1996) among group members while also leading to negative affective responses toward out-group members (for review, see Cialdini and Goldstein, 2004). This suggests the possibility that mixing features that have strong associations with members of contrasting categories will either lead to a reduction in positive affect or an increase in negative affect (Burleigh et al., 2013). In contrast to categorical perception, sub-categorical properties such as exposure to individual exemplars has long been considered an important determinant of affective responses (e.g., Fechner, 1876; Maslow, 1937; Zajonc, 1968). The present study considers how the comparatively low frequency of exposure to stimuli selected from a region of a continuum can lead to negative affective responses. We examine this in the context of negative affective responses to stimuli containing features from contrasting categories.

In the context of human factors, Mori’s (1970) Uncanny Valley Hypothesis (UVH) suggests that human-like objects in our environment might come to be associated with negative affect if they possess a certain degree of human-likeness. Recently, a number of authors have suggested potential explanations of the UVH that are either explicitly or implicitly based on categorical perception (Cheetham et al., 2011, 2013; Moore, 2012; Burleigh et al., 2013; Yamada et al., 2013; Ferrey et al., submitted). While these studies have made important theoretical contributions, the implications of different learned category representations on the UCV phenomenon have not been directly tested. In the present study, we sought to address this gap by using a category-learning paradigm in which groups of participants received different sets of training stimuli consisting of computer-generated creatures. We examine participants responses to creatures following training, specify the conditions in which affective minima associated with the UCV would be observed, the properties of category learning that would determine the location of affective minima, and what underlying representation of category structures would best fit the response patterns that we observed.

## THE UNCANNY VALLEY

While the essential phenomenon of the uncanny valley has a number of cultural antecedents (Schoenherr and Burleigh, 2014), the Uncanny Valley Hypothesis (UVH) remains underspecified. Mori (1970) initially proposed that human-like stimuli can elicit positive or negative feelings depending on their degree of similarity to humans. In contrast to the linear relationship between familiarity and positive affect for human and human-like faces (see Experiment 1, Burleigh et al., 2013), the UVH predicts a non-linear relationship. Mori’s account assumes that as stimuli become defined by more human-like features, they will elicit greater positive affect. But importantly, his account also assumes that there is a critical region of intermediate human-likeness where a sharp decrease in positive feelings are observed. As illustrated in **Figure 1**, the proposed relationship resembles a cubic function, and the global minimum is referred to as the “valley.”

It is important to distinguish between the classic and generalized forms of the UVH. The classic account of the UVH provided by Mori (1970) is defined as a non-linear function. In his account, the x-axis of this function is defined as human-likeness, and it is anchored by a non-human or minimally human-like entity at one end (e.g., a robot) and a real human at the other end. One reason to question this account is that it was informed by anecdotal evidence in the context of the human-like design of machines and other artifacts. The basic premise that negative affect could be a consequence of mismatch between features associated with contrasting classes would explain many cross-cultural phenomena (Schoenherr and Burleigh, 2014). In contrast to this, the UVH could be taken as assuming that the non-linear response function observed with human-likeness is a special case of more general cognitive and affective processes associated with stimulus frequency and categorical perception. Thus, it follows that similar non-linear phenomena should be observed in response to perceptual continua that represent non-human anchors with similar properties.

Evidence that has been interpreted as supporting the classic UVH has been obtained from studies using a variety of stimuli selected from a number of ontological categories. A majority of these studies have observed affective functions that are consistent with the UVH when using stimuli representing computer-generated morph sequences of human and non-human entities, including non-human animals, robots, and anthropomorphic dolls (MacDorman and Ishiguro, 2006; Seyama and Nagayama, 2007; Burleigh et al., 2013; Ferrey et al., submitted). Many studies have also observed the affective function in response to images of existing artifacts that vary in human-likeness, such as androids, videogame characters, and prosthetic hands (Bartneck et al., 2007; Schneider et al., 2007; Poliakoff et al., 2013); however, it is worth noting that several studies have not found support for the classic account of the UVH (MacDorman et al., 2009; Cheetham et al., 2014). Across those studies which have found support, a general observation is that affective response is positively correlated with human-likeness, except at an intermediate level of human-likeness where there is a maximum of negative affect.

Few studies have examined the possibility that perceptual continua representing non-human entities could produce UCV phenomena. To the best of our knowledge, only two studies have examined this possibility. In Yamada et al. (2013, Experiment 2) morph sequences were generated that represented transitions between cartoon, stuffed, and real dogs. In Ferrey et al. (submitted, Experiment 1), bistable morph sequences were used that represented transitions between various non-human animals (e.g., between a duck and an elephant). In each of these studies, regions of maximal negative affect were found at intermediate levels of the perceptual continua, which is consistent with the generalized account of the UVH (see, Burleigh et al., 2013, Experiment 2). Between general formulations of the UVH and empirical support for UCV-like phenomena, greater theoretical consideration of the affective and cognitive processes is required to define the conditions under which the UCV will be observed as well as to differentiate it from related phenomena.

## EXPLANATIONS FOR THE UNCANNY VALLEY PHENOMENON

Although the UVH provides a description of the non-linear response function, it does not explain why this function occurs, nor does it specify the mechanisms that are responsible. A common explanation is that the negative affect associated with uncanny stimuli might be a consequence of biological adaptations for threat avoidance behaviors (e.g., MacDorman et al., 2009). Stimuli within the valley might be convincing depictions of humans while falling short of a satisfactory level of human-likeness due to imperfections. These imperfections might cause them to be seen as “humans with disease” which triggers an aversive response (MacDorman and Ishiguro, 2006; MacDorman et al., 2009; Burleigh et al., 2013, Experiment 2). There is some evidence supporting this account. For example, Ho et al. (2008) observed that disgust could explain a significant portion of the variance in eeriness ratings. Furthermore, Steckenfinger and Ghazanfar (2009) observed that macaque monkeys displayed an aversion (as measured with looking times) to images of conspecifics that were of intermediate realism, which suggests that there might be an evolutionary basis to the phenomenon.

From this account, it might be reasonable to assume that the UCV phenomenon is specific to observers viewing images of conspecifics—an assumption that would be consistent with the classic UVH. Given that the spread of communicable diseases depends on the genetic similarity between the observer and the stimulus entity, it is possible that a species could have evolved mechanisms that allow them to respond to pathogen cues in conspecifics, but not heterospecifics. Communicable diseases, however, are not the only source of contamination that members of a species have had to contend with in their environments. As Rozin and Fallon (1987) point out, disgust is also a food-related emotion, which serves to prevent the oral incorporation of contaminated substances. As Schoenherr and Burleigh (2014) discuss, food substances that share features from two categories have been associated with aversive responses, such as food taboos (e.g., some refer to a certain transgenic tomato as a “Frankenfood” because it incorporates genes from a winter flounder). This suggests that the UCV phenomenon might not be specific to observers viewing images of conspecifics, but that it might also occur more generally in response to the categorical ambiguity of certain types of stimuli. Even if these accounts are correct, general learning mechanisms would also allow for the adjustment of diagnostic features of disease as well as inclusion and exclusion of categories associated with disease as a result of an individual’s experiences with their environment.

Another theory that accounts for threat avoidance behavior is based on the premise that appearances provide information that allows individuals to predict behavior, and thus to anticipate potential threats in their environment. Some uncanny valley stimuli can be seen to present mismatched features (Seyama and Nagayama, 2007; MacDorman et al., 2009; Mitchell et al., 2011; Saygin et al., 2012), such as a machine with a convincingly human voice, or an android with a physical appearance that is highly realistic but movements that are robotic. In this account, stimulus features, such as physical appearances, drive the automatic selection of a neural model for the purpose of predicting behavior. Stimulus mismatches can therefore lead to the selection of an inaccurate neural model, which is associated with error-related brain activity (Saygin et al., 2012), and error-related processing might result in negative affect. These neural models thus require learning in order to acquire ontological categories that subsequently produce contrasts due to feature mismatch.

## THE UNCANNY VALLEY AS CATEGORICAL PERCEPTION

If feature mismatch is the result of the inclusion of features from neighboring categories, then a crucial feature of any general account of the UVH is the specification of category learning systems that acquire the category structure, as well as the representations that are retained within them (for a recent review, see Goldstone et al., 2012). A number of studies have attempted to qualify the UVH by making reference to principles and processes associated with categorization generally, and categorical perception more specifically (Cheetham et al., 2011, 2013; Moore, 2012; Burleigh et al., 2013; Yamada et al., 2013; Ferrey et al. submitted). Categorical perception (CP) accounts of the UVH suggest that this phenomenon is a consequence of categorical processes associated with stimulus identification. Specifically, stimuli along a human-likeness continuum are perceived as members of either a “human” or “non-human” category, except at the category boundary where their membership is ambiguous. This follows from the position that stimuli at the category boundary should not provide the observer with sufficient perceptual evidence to allow easy or accurate identification on the basis of their representation of the category structure. As a consequence, uncertainty and negative affect are produced due to competition during categorization response selection (Cheetham et al., 2011; Burleigh et al., 2013), which might in turn activate conflict resolution processes like inhibitory devaluation (Ferrey et al., submitted).

Empirical evidence is consistent with accounts of the uncanny valley based on categorical perception. For instance, Cheetham et al. (2011, 2013) demonstrated that participants’ response latencies were longest when categorizing stimuli that were located at, or adjacent to, the category boundary on a human-avatar morph continuum. In addition to this, Burleigh et al. (2013, Experiment 2), Ferrey et al. (submitted), and Yamada et al. (2013), have each observed non-linear affective response functions across between-category (including human-animal and animal-animal) morph sequences that peaked at the midpoint between categories where stimuli were most ambiguous. Relative to the categorization literature, these accounts are underspecified, and therefore do not provide a complete account of the UCV phenomenon. Moreover, whereas Cheetham et al. (2011, 2013) and Yamada et al. (2013) have made a crucial connection between categorization and the response patterns associated with the uncanny valley, we cannot assume that categorization performance will be the only, or even the primary, determinant of affect. As we discuss, the uncanny valley might also be attributed to sub-categorical processes, such as those involved in assessing stimulus frequencies (Zajonc, 1968; Bornstein, 1989).

## CATEGORY BOUNDARY AND EXEMPLAR REPRESENTATIONS

Any explanation of the UCV phenomenon based on categorical perception must consider categorization processes and representational assumptions (e.g., prototype-related models were recently considered by Moore, 2012). Most CP accounts of the UVH appear to have assumed that categorization is governed by a “category boundary” representation (Cheetham et al., 2011, 2013; Burleigh et al., 2013). Category boundary models suggest that when a stimulus is encountered, it is used to locate and modify the location of a decision boundary in perceptual space (Ashby and Gott, 1988). When individuals are presented with a novel stimulus, they will compare its location in perceptual space to that of the category boundary. Proximity to the category boundary thereby increases categorization uncertainty (Paul et al., 2011; Schoenherr and Lacroix, 2014), and according to CP accounts of the UVH, proximity is also assumed to be inversely related to affect.

However, while a category boundary model might provide an adequate explanation of the uncanny “valley,” which is a simple U-shaped quadratic function, it cannot account for the entire UCV response function, which is a more complex N-shaped cubic function (e.g., Mori, 1970). We suggest that models that take into consideration exemplar-based information might account for the additional features of a more complex function. Exemplar-based models assume that a memory trace is encoded each time a stimulus is encountered (Medin and Schaffer, 1978; Nosofsky, 1984). During the course of learning, each instance becomes associated with a category label, and at the end of learning each exemplar is represented by a probability distribution of features. Over the course of learning, an individual’s attentional focus becomes reweighted to different regions of the stimulus continuum (Nosofsky, 1984, 1986), such that attention is sensitized to between-category differences and desensitized to within-category differences. When presented with a novel exemplar, individuals will compare it to all exemplars available in memory, and the similarity between the new item and old items in memory will determine the new item’s category membership.

Thus, a key difference between category boundary and exemplar-based models is how individuals become sensitized to perceptual space. Category boundary models suggest that individuals can only typically become sensitized to a single region of perceptual space, namely where the category boundary is located; whereas exemplar-based models suggest that individuals can become sensitized to multiple regions of perceptual space, due to the distributions of individual members (Nosofsky, 1984, 1986).

## THE UCV AS CATEGORICAL PERCEPTION OR FREQUENCY-BASED EXPOSURE

Crucially, affective processing of stimuli might not require the instantiation of categorical processes. The mere-exposure effect (Zajonc, 1968) suggests that repeated exposure to stimuli can lead to the formation of preferences, and negative affect might therefore be accounted for on the basis of familiarity or perceptual fluency alone (for a review, see Bornstein, 1989). In support of this, Harmon-Jones and Allen (2001) reported physiological evidence (via EMG and EEG) of affective responses that resulted from mere-exposure to stimuli, which corresponded with self-reported evaluations. If the mere-exposure effect can be extended to all members of a perceptual continuum, then an observer’s familiarity with individual members of the continuum might be able to explain non-linear affective response functions. For example, along a human-likeness continuum that is anchored by “human” and “robot,” individuals will have encountered a comparatively larger number of human instances relative to robots. Instances within these two categories should be much more familiar than instances that combine their features (e.g., androids). Thus, in contrast to the categorical perception account, a negative affective peak at an intermediate region in perceptual space might be explained by the fewer number of instances with the conjunction of features represented by stimuli in that region. On this basis, we suggest two distinct accounts of the UCV.

We suggest that at least two broad relationships are possible between cognitive and affective processing of stimuli, which we conceptualize as categorical perception (**Figure 2A**) and frequency-based exposure (**Figure 2B**) stage models. In conceptualizing these models, we limit ourselves to unidirectional processing. We assume that stimulus processing is mediated by the information that is stored in long-term memory, which includes memory traces of past episodes.

The categorical perception model (**Figure 2A**) reflects our understanding of extant categorical perception accounts of the UCV, in that it assumes categorical and affective responses derive from a common processing stage. In this model, individuals process sub-categorical information such as basic perceptual properties (e.g., stimulus magnitude, orientation) and frequency, but this information does not directly influence responding. Subsequent to this stage, category structures stored in long-term memory are activated, and these structures are used to determine both affective and categorical responses.

Alternatively, the frequency-based exposure model (**Figure 2B**) assumes that categorical and affective responses derive from separable processing stages. Specifically, affective responses are also driven by sub-categorical processing, which relies on frequency-based memory representations to provide more basic information such as frequency. The models defined in **Figures 2A,B** are sufficiently distinct that their predictions can be tested in a category-learning paradigm.

## PRESENT STUDY

The present study was designed to test the predictions of the multistage models of the uncanny valley presented in **Figures 2A,B**, and a nested prediction concerning category structures. In the categorical perception model, the affective and categorical responses are derived from the same processing stage. Such a model therefore leads to a prediction of similar patterns of affective and categorical responses, as well as a strong and positive correlation between them. In contrast to this, the frequency-based exposure model implies that categorical and affective responses each account for unique sources of variance. Such a model therefore suggests that under some conditions patterns of responses might be similar, but they need not show a significant correlation.

Importantly, the stage models do not make predictions concerning the specific nature of categorical processing, only the relationship between categorical and affective responses. Therefore, a nested prediction concerns whether categorization will reflect category boundary or exemplar-based representations. The first possibility is that individuals will only have access to a category boundary representation that partitions the response continuum. Therefore, categorization accuracy and affective responses should increase, and response times should decrease, as a function of a stimulus’ distance away from the category boundary. If participants are insensitive to individual characteristics, then categorization uncertainty should also be evidenced by a linear increase in response latencies as a function of proximity to the category boundary. Alternatively, if exemplar-based representations are acquired for two contrasting categories and used for categorical processing, then the location of the central tendencies for each category should determine the location of the maxima and minima of the response functions for affective and categorization responses. Response latencies should evidence a similar trend. Specifically, if uncertainty in category membership is a function of exemplar frequency, then we would expect exemplars presented with comparatively high frequency during training to be associated with fast responses whereas exemplars presented with comparatively low frequency to be associated with slow responses.

In order to test the predictions of these models, our experimental design uses a category-learning paradigm in which we manipulate exemplar frequency along the perceptual continuum. Experiment 1 consists of two training conditions. In the first condition, stimuli within response categories are presented with equal frequency, with each category having an equivalent distribution (EFED). In the second condition, both category distributions are equivalent, but the exemplars were presented with unequal frequency (UFED) such that stimuli near the extrema of Categories A and B training sets are presented with the greatest frequency, and stimuli adjacent to the category boundary were presented with the lowest frequency. An important aspect of our design is that individuals are not exposed to the continuum extrema during the training phase. Thus, while the category boundary of the EFED and the UFED conditions should be identical, differences in exemplar frequency should decrease affective responses outside the training range if frequency-based information is a determinant of categorical and/or affective responses.

The results of Experiment 1 should provide a straightforward tests of our predictions. Left unaddressed, however, is what we consider to be a tacit property of UCV as discussed by Mori (1970): we are presented with less exemplar variability within one category (e.g., human) and greater exemplar variability in the contrasting category (e.g., non-human). In Experiment 2, we used one category defined by exemplars with equal frequencies selected from the EFED condition and another category defined by exemplars with unequal frequencies selected from the UFED condition. This procedure resulted in an unequal frequency, unequal distribution condition (UFUD) which we take as a closer approximation to the properties of the UCV first proposed by Mori (1970). **Table 1** provides training set frequencies.

Crucially, we were also interested in determining whether the affective response patterns could reasonably support a UCV interpretation. We distinguish between “strong” and “weak” interpretation as follows. The UCV function is a non-linear response function that is defined by a slope, indicating a category preference attributable to familiarity (e.g., for humans over robots), and a valley region that is located near the category boundary but skewed toward the preferred category. Thus, support for a strong interpretation of the UCV would be obtained if a response function possessed all of these features; support for a weak interpretation of the UCV would be obtained if a response function possessed some of these features, such as a valley region without a slope. We anticipate the possibility that the EFED and UFED conditions might provide support for a weak interpretation, but not for the strong interpretation, due to their symmetry. In contrast, the UFUD condition might provide support for a strong interpretation of the UCV due to the asymmetry of the response function.

Although the stimuli that we use all represent non-human entities, we believe the findings of these studies are pertinent to human-like stimuli. By using non-human stimuli we hope to minimize the influence of stimulus familiarity or preference for human stimuli. This novelty facilitates the task of training participants to learn different category structures in an experimental setting with practical limitations (e.g., time). This manipulation also allows us to illustrate that response patterns associated with the UCV are generally patterns that can be attributed to stimulus familiarity rather than human-likeness, _per se_.

---

### Table 1 | Stimulus Frequencies for Training Session in Experimental Conditions for equal frequency, even distributions (EFED), unequal frequency, even distributions (UFED), and unequal frequency, uneven distributions (UFUD).

|        | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | Tot. |
|--------|---|---|---|---|---|---|---|---|---|----|----|----|----|----|----|------|
| EFED   | - | - | 4 | 4 | 4 | 4 | 4 | - | 4 | 4  | 4  | 4  | 4  | -  | -  | 40   |
| UFED   | - | - | 8 | 6 | 4 | 2 | - | - | - | 2  | 4  | 6  | 8  | -  | -  | 40   |
| UFUD   | - | - | 8 | 6 | 4 | 2 | - | - | 4 | 4  | 4  | 4  | 4  | -  | -  | 40   |

---

## EXPERIMENT 1

Experiment 1 was designed as an initial test of our predictions derived from the hypothesized multistage models, and to provide evidence in support of the UCV phenomenon. We manipulated the frequency of stimulus presentation to differentially sensitize participants to regions of the stimulus response continuum. An equal frequency condition (EFED) was provided to half of the participants, wherein all stimuli within a category were presented with equal frequency, thereby creating a uniform distribution. An unequal frequency condition (UFED) was provided to the remaining half of the participants, wherein stimuli located within the middle of each category distribution were presented with higher frequency, thereby approximating a normal distribution. In each case, distributions of stimuli from Category A and Category B were symmetrical. Thus, by the end of training we hypothesized that participants should learn the distribution of the training stimuli equally well. Following training, participants responded to stimuli selected from the entire continuum. In the UFED condition, we additionally predicted that participants should show changes in affective responses due to less familiarity with the extreme values that in fact share fewer features with the contrasting category.

### METHODS

**Participants**

A total of 60 participants were recruited online for this study (31 female, _Mage_ = 37.2). Participants were recruited from Amazon’s Mechanical Turk platform and paid a total of $5 if they completed all 4 sessions of the study ($1 for session 1, $1.25 for sessions 2 and 3, and $1.5 for session 4). All participants were registered with Mechanical Turk as United States residents. No participants reported having a visual impairment, and therefore no participants were excluded from our analyses. All participants consented to participate in the study.

**Stimuli**

Three morph sequences were generated, comprising the permutations of three distinct non-human creatures: a beast, a reptile, and an alien. These creatures were selected given our assumption that participants would have less familiarity with these categories thereby allowing us to more readily manipulate their frequency of exposure in the experimental context. Creatures were created using Daz Studio 4.6 Pro (daz3d.com) by modifying the morphology and texture of the _Genesis_ base figure. Morph sequences were then created by stepwise adjustment of morphology and texture parameters corresponding to each creature. For example, the reptile creature had a “head scale” parameter which determined the size of its head, with a value of 32, whereas the alien creature had a value of 40. Therefore, the stimulus at the midpoint on the alien-reptile morph continuum assumed a value for this parameter that was half-way between the alien and reptile values (i.e., 36). Stimuli were then cropped in photo-editing software using an elliptical mask, and saved as images with a vertical resolution of 548 pixels. Stimuli were divided into training and test sets. The following stimuli were excluded from all training sets: stimulus 6 (the category boundary), and stimuli 1, 2, 14, and 15 (the extrema). Other stimuli were excluded depending on the frequency condition. For instance, stimuli 7 and 9 were not included in the training set for the UFED condition due to the frequency manipulation.

**Procedure**

_Training._ At the start of the experiment, participants were presented with stimuli during the training and test phases of the experiment by randomly assigning them to a creature continuum (for an example, see **Figure 3**) and a frequency condition (see **Table 1**). In order to control for the effect of creature continua, we used a counter-balanced design such that an equal number of participants were assigned to each of the (creature x frequency) conditions. This resulted in a total of 5 participants for each cell of the design, or 30 participants in each of the experimental conditions that were of interest. In the EFED condition, participants received an equal presentation of stimuli selected from the training range, whereas in the UFED condition participants received an unequal presentation of stimuli selected from the training range; in each case the frequency distributions were symmetrical.

At the beginning of training, participants were instructed that they would be presented with “models of unfamiliar living creatures” and that their task was to “learn what categories they belonged to.” They were told that each creature was either a “Cax” or a “Miv” and that they were to press the “C” or “M” key depending on which type of creature they thought they saw. Participants were instructed to balance the demands of speed and accuracy. Key assignment was counter-balanced across participants.

Participants completed 1 training session per day over the course of 3 days. Each training session was composed of 10 blocks of 40 trials each, for a total of 400 trials per training session, and each session required approximately 20 min to complete. For each trial, a fixation point was presented for 500 ms, followed by a randomly selected stimulus from the training distribution for 750 ms (these timings were selected to be consistent with Cheetham et al., 2011). At the end of this sequence the response alternatives were presented until a response was registered. After a response was registered, feedback in the form of a “correct” or “incorrect” message was presented for 500 ms.

_Test._ In the test session, all 15 stimuli were presented. Unlike the previous blocks, we sought to limit the amount of exposure to previously unseen stimuli. Therefore, the test session consisted of 4 blocks, in which each stimulus was presented 2 times each, for a total of 120 trials. The training session required approximately 12 min to complete. Stimulus presentation preceded in the same manner as in the training phase with two notable exceptions. Following presentation of a stimulus, participants were asked to rate its eeriness on a scale ranging from 1 (not at all eerie) to 7 (extremely eerie) using the “1” through “7” keys, respectively. After registering their response, participants were then asked to indicate whether it was a “Cax” or “Miv,” as in previous sessions. The ordering of affective and categorization responses was deliberate in order to ensure that the effect of categorical information on ratings of eeriness would be limited.

_Implementation._ The study was developed for the web using HTML and JavaScript programming languages for the frontend, and PHP/MySQL for the backend. Preliminary tests using an automated responder on a test machine revealed that response time noise was within acceptable limits (i.e., less than 35 ms). Our online research environment is comparable to the one used by Crump et al. (2013). Crump et al., used JavaScript and recruited Mechanical Turk participants to successfully replicate numerous reaction time tasks like the Stroop (1935).

---

[Figures and further detailed results, including statistical analyses, are described in the original text above.]

---

## REFERENCES

Ashby, F. G., Alfonso-Reese, L. A., Turken, A. U., and Waldron, E. M. (1998). A neuropsychological theory of multiple systems in category learning. _Psychol. Rev._ 105, 442–481. doi: 10.1037/0033-295x.105.3.442  
Ashby, F. G., and Gott, R. E. (1988). Decision rules in the perception and categorization of multidimensional stimuli. _J. Exp. Psychol. Learn. Mem. Cogn._ 14, 33–53. doi: 10.1037/0278-7393.14.1.33  
Ashby, F. G., and Maddox, W. T. (1993). Relations between prototype, exemplar and decision bound models of categorization. _J. Math. Psychol._ 37, 372–400. doi: 10.1006/jmps.1993.1023  
Bartneck, C., Kanda, T., Ishiguro, H., and Hagita, N. (2007). “Is the uncanny valley an uncanny cliff?,” in _Robot and Human interactive Communication, 2007. RO-MAN 2007. The 16th IEEE International Symposium_ (Jeju: IEEE), 368–373.  
Bless, H., Bohner, G., Schwarz, N., and Strack, F. (1990). Mood and persuasion: a cognitive response analysis. _Pers. Soc. Psychol. Bull._ 16, 311–345. doi: 10.1177/0146167290162013  
Bond, R., and Smith, P. B. (1996). Culture and conformity: a meta-analysis of studies using Asch’s (1952b, 1956) line judgment task. _Psychol. Bull._ 119, 111–137. doi: 10.1037//0033-2909.119.1.111  
Bornstein, R. F. (1989). Exposure and affect: overview and meta-analysis of research, 1968-1987. _Psychol. Bull._ 106, 265–289. doi: 10.1037//0033-2909.106.2.265  
Burleigh, T. J., Schoenherr, J. R., and Lacroix, G. L. (2013). Does the uncanny valley exist? an empirical test of the relationship between eeriness and the human likeness of digitally created faces. _Comput. Hum. Behav._ 29, 759–771. doi: 10.1016/j.chb.2012.11.021  
Burnham, K. P., and Anderson, D. R. (2002). _Model Selection and Multimodel Inference: a Practical Information-Theoretic Approach,_ 2nd Edn. New York, NY: Springer.  
Cacioppo, J. T., and Gardner, W. L. (1999). Emotion. _Annu. Rev. Psychol._ 50, 191–214.  
Cheetham, M., Pavlovic, I., Jordan, N., Suter, P., and Jancke, L. (2013). Category processing and the human likeness dimension of the uncanny valley hypothesis: eye-tracking data. _Front. Psychol._ 4:108. doi: 10.3389/fpsyg.2013.00108  
Cheetham, M., Suter, P., and Jancke, L. (2014). Perceptual discrimination difficulty and familiarity in the uncanny valley: more like a “Happy Valley.” _Front. Psychol._ 5:1219. doi: 10.3389/fpsyg.2014.01219  
Cheetham, M., Suter, P., and Jäncke, L. (2011). The human likeness dimension of the “uncanny valley hypothesis”: behavioral and functional MRI findings. _Front. Hum. Neurosci._ 5:126. doi: 10.3389/fnhum.2011.00126  
Cialdini, R. B., and Goldstein, N. J. (2004). Social influence: compliance and conformity. _Annu. Rev. Psychol._ 55, 591–621. doi: 10.1146/annurev.psych.55.090902.142015  
Crump, M. J., McDonnell, J. V., and Gureckis, T. M. (2013). Evaluating Amazon’s Mechanical Turk as a tool for experimental behavioral research. _PLoS ONE_ 8:e57410. doi: 10.1371/journal.pone.0057410  
Damasio, A. (1994). _Descartes’ Error: Emotion, Reason, and the Human Brain_. New York, NY: Putnam.  
Davis, F. (1930). The relative reliability of words and nonsense syllables as learning material. _J. Exp. Psychol._ XIII, 221–234. doi: 10.1037/h0071308  
Fechner, G. T. (1876). _Vorschule derAesthetik._ Leipzig: Breitkoff and Hartel.  
Fiske, S. T., and Taylor, S. E. (2013). _Social Cognition: From Brains to Culture._ London: Sage Publications Ltd.  
Förster, J., Marguc, J., and Gillebaart, M. (2010). Novelty categorization theory. _Soc. Psychol. Pers. Compass_ 4, 736–755. doi: 10.1111/j.1751-9004.2010.00289.x  
Gillebaart, M., Förster, J., and Rotteveel, M. (2012). Mere exposure revisited: the influence of growth versus security cues on evaluations of novel and familiar stimuli. _J. Exp. Psychol. Gen._ 141, 699–714. doi: 10.1037/a0027612  
Goldstone, R. L., Kersten, A., and Carvalho, P. F. (2012). “Concepts and Categorization,” In _Handbook of Psychology,_ Vol. 4, _Experimental Psychology_, 2nd Edn., eds I. B. Weiner, A. J. Healey, and R. W. Proctor (New York, NY: Wiley), 607–630.  
Haidt, J. (2001). The emotional dog and its rational tail: a social intuitionist approach to moral judgment. _Psychol. Rev._ 108, 814–834. doi: 10.1037/0033-295x.108.4.814  
Harmon-Jones, E., and Allen, J. J. B. (2001). The role of affect in the mere exposure effect: evidence from physiological and individual differences approaches. _Pers. Soc. Psychol Bull._ 27, 889–898. doi: 10.1177/0146167201277011  
Harnad, S. (1987). “Psychophysical and cognitive aspects of categorical perception: a critical overview,” in _Categorical Perception: the Groundwork of Cognition_ S. Harnad (New York, NY: Cambridge University Press), 1–19.  
Ho, C. C., MacDorman, K. F., and Pramono, Z. D. (2008). “Human emotion and the uncanny valley: a GLM, MDS, and Isomap analysis of robot video ratings,” in _Proceedings of the 3rd ACM/IEEE International Conference on Human Robot Interaction_ (ACM), 169–176.  
Hunt, R. R., and Worthen, J. (2006). _Distinctiveness and Memory_. New York, NY: Oxford University Press.  
Kent, C., and Lamberts, K. (2005). An exemplar account of the bow and set-size effects in absolute identification. _J. Exp. Psychol. Learn. Mem. Cogn._ 31, 289–305. doi: 10.1037/0278-7393.31.2.289  
Lacerda, F. (1995). The perceptual-magnet effect: an emergent consequence of exemplar-based phonetic memory. _Proc. XIIIth Int. Congr. Phon. Sci._ 2, 140–147.  
MacDorman, K. (2005). “Androids as an experimental apparatus: why is there an uncanny valley and can we exploit it?” in _Toward Social Mechanisms of Android Science: A CogSci 2005 Workshop_ (Stresa), 106–118.  
MacDorman, K. F., Green, R. D., Ho, C. C., and Koch, C. T. (2009). Too real for comfort? Uncanny responses to computer generated faces. _Comput. Hum. Behav._ 25, 695–710. doi: 10.1016/j.chb.2008.12.026  
MacDorman, K. F., and Ishiguro, H. (2006). The uncanny advantage of using androids in cognitive and social science research. _Interact. Stud._ 7, 297–337. doi: 10.1075/is.7.3.03mac  
Maslow, A. H. (1937). The influence of familiarization on preference. _J. Exp. Psychol._ 21, 162–180. doi: 10.1037/h0053692  
Medin, D. L., and Atran, S. (Eds.). (1999). _Folkbiology_. Cambridge: MIT Press.  
Medin, D. L., and Schaffer, M. M. (1978). Context theory of classification learning. _Psychol. Rev._ 85, 207–238. doi: 10.1037/0033-295x.85.3.207  
Miller, J. L., and Volaitis, L. E. (1989). Effect of speaking rate on the perceptual structure of a phonetic category. _Percept. Psychophys._ 46, 505–512. doi: 10.3758/bf03208147  
Mitchell, W. J., Szerszen Sr, K. A., Lu, A. S., Schermerhorn, P. W., Scheutz, M., and MacDorman, K. F. (2011). A mismatch in the human realism of face and voice produces an uncanny valley. _i-Perception_, 2, 10. doi: 10.1068/i0415  
Moore, R. K. (2012). A Bayesian explanation of the ‘Uncanny Valley’ effect and related psychological phenomena. _Sci. Rep._ 2, 1–5. doi: 10.1038/srep00864  
Mori, (1970). Bukimi no tani [The uncanny valley]. _Energy_ 7, 33–35.  
Nosofsky, R. M. (1984). Choice, similarity, and the context theory of classification. _J. Exp. Psychol. Learn. Mem. Cogn._ 10, 104–114. doi: 10.1037//0278-7393.10.1.104  
Nosofsky, R. M. (1986). Attention, similarity, and the identification-categorization relationship. _J. Exp. Psychol. Gen._ 115, 39–57. doi: 10.1037/0096-3445.115.1.39  
Nosofsky, R. M., Gluck, M. A., Palmeri, T. J., McKinley, S. C., and Glauthier, P. (1994). Comparing models of rule-based classification learning: a replication and extension of Shepard, Hovland, and Jenkins (1961). _Mem. Cogn._ 22, 352–369. doi: 10.3758/bf03200862  
Paul, E. J., Boomer, J., Smith, J. D., and Ashby, F. G. (2011). Information–integration category learning and the human uncertainty response. _Mem. Cogn._ 39, 536–554. doi: 10.3758/s13421-010-0041-4  
Pisoni, D. B., Aslin, R. N., Percy, A. J., and Hennessy, B. L. (1982). Some effects of laboratory training on identification and discrimination of voicing contrasts in stop consonants. _J. Exp. Psychol. Hum. Percept. Perform._ 8, 297–314. doi: 10.1037//0096-1523.8.2.297  
Pisoni, D. B., and Tash, J. (1974). Reaction times to comparisons within and across phonetic categories. _Percept. Psychophys._ 15, 285–290. doi: 10.3758/bf03213946  
Poliakoff, E., Beach, N., Best, R., Howard, T., and Gowen, E. (2013). Can looking at a hand make your skin crawl? peering into the uncanny valley for hands. _Perception_ 42, 998–1000. doi: 10.1068/p7569  
Posner, M. I., and Keele, S. W. (1968). On the genesis of abstract ideas. _J. Exp. Psychol._ 77, 304–363. doi: 10.1037/h0025953  
Regier, T., and Kay, P. (2009). Language, thought, and color: whorf was half right. _Trends Cogn. Sci._ 13, 439–446. doi: 10.1016/j.tics.2009.07.001  
Rosch, E., and Mervis, C. B. (1975). Family resemblances: studies in the internal structure of categories. _Cogn. Psychol._ 7, 573–605. doi: 10.1016/0010-0285(75)90024-9  
Royall, R. M. (1997). Statistical _Evidence: A Likelihood Paradigm_. London: Chapman and Hall.  
Rozin, P., and Fallon, A. E. (1987). A perspective on disgust. _Psychol. Rev._ 94, 23–41. doi: 10.1037/0033-295x.94.1.23  
Saygin, A. P., Chaminade, T., Ishiguro, H., Driver, J., and Frith, C. (2012). The thing that should not be: predictive coding and the uncanny valley in perceiving human and humanoid robot actions. _Soc. Cogn. Affect. Neurosci._ 7, 413–422. doi: 10.1093/scan/nsr025  
Schneider, E., Wang, Y., and Yang, S. (2007). “Exploring the uncanny valley with Japanese video game characters,” in _Proceedings of DiGRA 2007: Situated Play_ (Tokyo), 546–549.  
Schoenherr, J. R., and Lacroix, (2014). _Dissociating Implicit and Explicit Category Learning Systems Using Confidence Reports_. Doctoral Dissertation, Carleton University.  
Schoenherr, J. R., and Burleigh, T. J. (2014). Uncanny sociocultural categories. _Front. Psychol._ 5:1456. doi: 10.3389/fpsyg.2014.01456  
Schoenherr, J. R., and Logan, J. (2013). “Subjective awareness during cross-language speech perception: attending unattended regions of an acoustic continuum,” in _Proceedings of the 34th Annual Meeting of the Cognitive Science Society_ (Berlin).  
Schoenherr, J. R., and Logan, J. (2014). “Attentional and immediate memory capacity limitations in the acquisition of non-native linguistic contrasts,” in _Proceedings of the 35th Annual Meeting of the Cognitive Science Society_ (Québec, QC)  
Schoenherr, J. R., Logan, J., and Winchester, A. (2012). “Subjective confidence of acoustic and phonemic representations during speech perception,” in _Proceedings of the 34th Annual Meeting of the Cognitive Science Society_ (Sapporo).  
Schwarz, N. (2002). “Situated cognition and the wisdom of feelings: cognitive tuning,” in _The Wisdom in Feeling: Psychological Processes in Emotional Intelligence_, eds L. F. Barrett and P. Salovey (New York, NY: Guilford Press), 144–166.  
Schwarz, N. (2010). “Feelings-as-information theory,” in _Handbook of theories of social psychology_, Vol. 1, eds P. A. M. Van Lange, A. Kruglanski, and E. Tory Higgins (London: Sage Publications, Inc), 289–308.  
Schwarz, N., and Clore, G. L. (2007). “Feelings and phenomenal experiences,” in _Social Psychology: Handbook of Basic Principles 2nd Edn._, eds A. Kruglanski and E. Tory Higgins (New York, NY: Guilford Press), 385–407.  
Seyama, J. I., and Nagayama, R. S. (2007). The uncanny valley: effect of realism on the impression of artificial human faces. _Presence_ 16, 337–351. doi: 10.1162/pres.16.4.337  
Slovic, P., Finucane, M. L., Peters, E., and MacGregor, D. G. (2002). “The affect heuristic,” in _Heuristics and Biases: The Psychology of Intuitive Judgment_, eds T. Gilovich, D. Griffin, and D. Kahneman (New York, NY: Cambridge University Press), 397–420.  
Smith, J. D., and Minda, J. P. (1998). Prototypes in the mist: the early epochs of category learning. _J. Exp. Psychol. Learn. Mem. Cogn._ 24, 1411–1436. doi: 10.1037//0278-7393.24.6.1411  
Steckenfinger, S. A., and Ghazanfar, A. A. (2009). Monkey visual behavior falls into the uncanny valley. _Proc. Natl. Acad. Sci. U.S.A._ 106, 18362–18366. doi: 10.1073/pnas.0910063106  
Stewart, N., and Chater, N. (2002). The effect of category variability in perceptual categorization. _J. Exp. Psychol. Learn. Mem. Cogn._ 28, 893–907. doi: 10.1037/0278-7393.28.5.893  
Stroop, J. (1935). Studies of interference in serial verbal reactions. _J.Exp. Psychol._ 18, 643–662. doi: 10.1037/h0054651  
Wagenmakers, E.-J., and Farrell, S. (2004). AIC model selection using Akaike weights. _Psychon. Bull. Rev._ 11, 192–196. doi: 10.3758/bf03206482  
Yamada, Y., Kawabe, T., and Ihaya, K. (2013). Categorization difficulty is associated with negative evaluation in the “uncanny valley” phenomenon. _Japan. Psychol. Res._ 55, 20–32. doi: 10.1111/j.1468-5884.2012.00538.x  
Zajonc, R. B. (1968). Attitudinal effects of mere exposure. _J. Personality and Soc. Psychol._ 9, 1–27. doi: 10.3758/bf03337428

---

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

**Statement of Ethics:** The studies reported in this article were approved by the Research Ethics Board at the University of Guelph (REB #14JA020).

**Acknowledgments:** We would like to thank Karl F. MacDorman and Mark Brown for their constructive feedback on earlier drafts of the manuscript.

---

_Citation: Burleigh TJ and Schoenherr JR (2015) A reappraisal of the uncanny valley: categorical perception or frequency-based sensitization? Front. Psychol. 5:1488. doi: 10.3389/fpsyg.2014.01488_

_Copyright © 2015 Burleigh and Schoenherr. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms._