Facial Image Manipulation

Abstract

Image manipulation of real face photographs, including averaging, morphing, and caricaturing, is widely used in studies of face perception. These methods have led to theoretical insights across topics of interest to social psychologists, including social perception, social categorization, stereotyping, prejudice, impression formation, and individual differences. They may also have practical applications in diagnosing clinical impairments of social perception. Here, we outline key manipulation methods, comment on the strengths and weaknesses of the approach, illustrate the breadth of theoretical insights already achieved, and offer best practice guidelines. We hope that the review stimulates greater use of these powerful methods to understand social perception and encourages studies that bridge theory between visual perception and social psychology.

Keywords

social perception impression formation face perception individual differences stereotyping emotional expression

A wealth of social cues are available from a person’s face (Bruce & Young, 2012; Rhodes, 2006) and facial impressions can have important social outcomes (Todorov, Olivola, Dotsch, & Mende-Siedlecki, 2015; Zebrowitz, 2005). Faces have even been described as the most important human social stimuli (Webster & MacLeod, 2011). Research on face perception is therefore central to an understanding of social cognition (Quinn & Macrae, 2011) and is increasingly being used to understand diverse high-level social processes of person construal, stereotyping, and prejudice (see Adams & Kleck, 2003; Blair, Judd, & Fallman, 2004; Freeman & Ambady, 2011; Hess, Adams Jr, & Kleck, 2008; Hugenberg & Sacco, 2008; Johnson, Freeman, & Pauker, 2012; Quinn & Macrae, 2011; Todorov et al., 2015, for influential papers showcasing the breadth of this literature). This recent drive builds on a long history of research employing face images to understand the social perception of groups and individuals (e.g., Bruce & Young, 1986; Perrett, May, & Yoshikawa, 1994; Rhodes & Tremewan, 1996; Secord, 1958; Taylor, Fiske, Etcoff, & Ruderman, 1978; Zebrowitz & Montepare, 1992).

Studies examining the social perception of faces often use photographs of real faces or alternatively, synthetic, computer-generated “FaceGen” images. Naturally, real photographs offer clear ecological validity, but they can also lack precise experimental control. Conversely, FaceGen images allow for greater experimental control over facial features, helping studies infer causality, but they do not always fully capture real-life cues (Crookes et al., 2015). One can capitalize on the strengths of each method by manipulating real face photographs using image-based techniques, including morphing and averaging. This image-based approach is gaining increasing currency in social psychology (e.g., Corneille, Huart, Becquart, & Brédart, 2004; Freeman, Pauker, & Sanchez, 2016; Oldmeadow, Sutherland, & Young, 2013; Sofer, Dotsch, Wigboldus, & Todorov, 2015). Importantly, facial manipulation techniques can maintain both experimental control and ecological validity, allowing stronger claims about which real-life face cues subserve social judgments.

These morphing techniques originate from a substantial research tradition in face perception (Bruce & Young, 2012; Calder, Rhodes, Johnson, & Haxby, 2011; Rhodes, 2006, review the field). Crucially, this face perception literature also investigates social and personality processes, including emotion perception, social categorization and stereotyping, self-categorization, personality and individual differences, and impression formation (Bruce & Young, 2012; Calder et al., 2011; Rhodes, 2006). Increased theoretical and methodological cross talk between social psychology and face perception is thus timely and would benefit progress in both fields. To encourage future integration, we show how the computer manipulation of face photographs is a powerful technique that can be applied to domains of interest to social psychologists.

We focus here on the manipulation of real face images rather than the complementary approach of reverse correlation, which instead seeks to visualize participants’ inner representations of stereotypes from deliberately minimal visual input (Dotsch, Wigboldus, Langner, & van Knippenberg, 2008; via Mangini & Biederman, 2004; instead, see Dotsch & Todorov, 2011, for a review). We also focus on readily available averaging and morphing techniques rather than more technically demanding facial manipulation approaches, such as the morphable face model approach (see Vetter & Walker, 2011, for a review). Morphable models have generated important insights into social perception and also allow the manipulation of real face photographs with naturalistic results (Walker & Vetter, 2009). However, face photographs are manipulated with reference to a custom-built statistical model of facial attributes (Blanz & Vetter, 1999), which is not readily available. Moreover, manipulation is also constrained to relationships between social and facial attributes found in the original population (e.g., Caucasian young adults; see Walker, Jiang, Vetter, & Sczesny, 2011, for a discussion). In contrast, facial averaging and morphing techniques are easily carried out via widely accessible software that can flexibly manipulate any new face image set.

We have three main aims. First, we outline key facial image manipulation methods, including landmarking, averaging, morphing, and caricaturing. Second, we show how these methods can be used to understand social perception. In doing so, we illustrate the depth and breadth of theoretical insights already achieved, from classic discoveries through to the recent application of these methods to impression formation. Third, we compare different software to help the interested reader get started. Although we concentrate on face photographs, we end by outlining how comparable methods can give insight into social perception from dynamic or 3-D faces, bodies, and voices.

We hope that this review will stimulate new research and help in the assessment of papers using these methods. For researchers who are already familiar with these techniques, we hope that our review will be useful in evaluating their uses and limitations, outlining different software capabilities, and setting out best practice guidelines. Our overall goal is to stimulate increased dialogue between face perception and social psychology for mutual benefit to both fields.

Face Manipulation Methods

We can think of a face photograph as involving two distinct properties (Bruce & Young, 2012). First, it represents 2-D facial shape: the positions and shapes of features (eyes, chin, etc.) as they are located in the photograph. Second, it represents the facial surface: the brightness and color of features, skin, and hair, including shading cues to 3-D facial shape from ambient lighting.

A digital face image is represented as an array of pixels, each with associated color and brightness values. Image manipulation programs can modify the pixel array to change either shape or surface. To represent and manipulate 2-D facial shape, landmarks are placed around key facial features (placing fiducial points or face extraction; Figure 1A). Landmark locations are defined by x and y coordinates. To represent and manipulate the facial surface, a mesh of small regions is created by joining imaginary lines between the landmarks (tessellation; Figure 2). Pixel values defining the facial surface (color, etc.) are then measured separately for each tessellated region.

Figure 1.

(A) Face landmarking: Prespecified feature points are marked out on a face image (at a minimum, defining facial height, width, and eye position). (B) Averaging: A set of landmarked face images can be averaged together (e.g., a set of female faces to create an average female face). The images are first aligned to the average landmarked shape and then the aligned individual images are averaged together (thereby averaging the same features across images, such as the eyes or lips). (C) Morphing: Blending one face image into another, for example, blending a female average (leftmost image) into a male average (rightmost image). (D) Transforming: Transforming an original face image with reference to two other face images by applying the difference between these faces to the original face, for example, blending feminine (leftmost image) and masculine (rightmost image) versions of a single female face (here, taken from A) by transforming the image with reference to average female and male faces (here, taken from the end points of C). (E) Caricaturing: Creating a face image that looks more extreme than the average face by transforming a face image away from an average face, for example, creating a hyperfeminine version of an average female face by transforming it away from an average male face (left image) and vice versa for a hypermasculine face (right image). From “The Karolinska Directed Emotional Faces - KDEF” by D. Lundqvist, A. Flykt, and A. Öhman, 1998, CD Rom from Department of Clinical Neuroscience, Psychology section, Karolinska Institutet. Copyright 1998 by Karolinska Institutet, Department of Clinical Neuroscience, Section of Psychology, Stockholm, Sweden. Reprinted with permission.

Figure 2.

A landmarked image (left) and a diagram of the tessellations between the landmarks (right). Morphing is carried out for each tessellated region separately. Image used with permission from David Perrett.

One way to understand this process is to imagine that the photograph is on a rubber sheet, mounted on a board with a pin through each landmark. If we drag the pins around, the rubber sheet will follow, so that the image shape (landmark positions) shifts while its surface properties (rubber sheet) remain the same, albeit now reshaped. If we have landmarked photographs of two people, and we bring all the landmarks into alignment across the photographs (usually by calculating the average position of each landmark across the images), then the 2-D shapes of each person’s image will now become identical, but their surface properties will still be quite different. Since the shapes are the same, we can now average the surface properties by averaging the pixel values in each of the tessellated regions (Figure 1B). This average image is often called a prototype or a composite.

This averaging method involves equal blends of constituent images. However, it is also possible to weight the shape and/or surface properties of different images in the final mix (e.g., to create a blend of 80% of image A and 20% of image B). This is face morphing or warping. By gradually varying the proportions contributed by each image, series of images can be created to blend one face into another (morphing continua). For example, a female face can be morphed into a male face (Figure 1C). Usually, both image shape and surface are manipulated (Figure 1C), but morphing can also be primarily carried out with either facial shape or surface (although complete separation is unlikely; Sormaz, Young, & Andrews, 2016).

Face transforming is an extension of morphing that transforms a single face image along a continuum by applying the difference between two other face images to the original. For example, a single female target can be changed to make her look more feminine or masculine by transforming her image with reference to male or female averages (Figure 1D).

Morphing and transforming both describe processes that blend faces together (e.g., morphing a female toward a male face). In contrast, face caricaturing involves creating a face image that looks more extreme than the original (a hyperface), by exaggerating differences, through transforming a face image away from a reference (usually an average face). For example, a hyper-male face can be created by transforming an average male face away from an average female face (Figure 1E). An individual face can also be caricatured away from the average to emphasize what is distinctive about that person (like a political caricature). Conversely, an individual face can be blended toward an average, reducing its distinctiveness (anticaricaturing).

Inevitably, these techniques involve trade-offs. For example, the number and positions of landmarks limit accuracy in shape representation: With more landmarks, finer-grained manipulation becomes possible. Less obvious is the problem that while the tessellated regions are usually triangular, photographs are usually comprised of square pixels. Therefore, reshaping a region may involve interpolating new pixels to fill stretched parts or deleting pixels where there is shrinkage. Sophisticated algorithms achieve this reshaping, but our rubber sheet analogy is clearly imperfect. Nevertheless, facial manipulation methods have led to very interesting insights into social perception. One reason why 2-D image manipulation can work so well might be because the images falling on our retinas are also 2-D, making us highly capable of extracting essential social information from this format.

Using Facial Manipulation to Examine Social Perception

Facial Landmarking

Although landmarking is usually only carried out to manipulate facial images, landmarked features can themselves be used to investigate important social cues. Vernon, Sutherland, Young, and Hartley (2014) utilized landmark locations across 1,000 naturalistic face images to define facial features (e.g., eye size) and then used these features to predict facial impressions. This approach gave insights into the pattern of facial cues underlying impressions and allowed impressions to be automatically extracted from new photographs, with obvious applications for real-world impression prediction (Vernon, Sutherland, Young, & Hartley, 2014; see also Lienhard, Ladret, & Caplier, 2015). The approach was inspired by pioneering work in social psychology by Zebrowitz and colleagues who used facial landmarks to test the theory that facial impressions are based on subtle resemblance to emotional expression and to investigate stereotypes (e.g., Zebrowitz, Kikuchi, & Fellous, 2010; Zebrowitz & Montepare, 1992). Once face images have been landmarked, it is straightforward to recover feature points to answer new research questions (O’Toole, 2011, reviews many other studies using facial morphology).

Facial Averaging

Face averaging was first described by Galton (1878) who created facial composites of criminals by photographing a series of mug shots using the same photographic plate. Galton was interested in discovering the physiognomy of criminality, believing that character is displayed in facial features (e.g., Lavater, 1778; an idea that remains hotly debated: Rule, Krendl, Ivcevic, & Ambady, 2013; Stillman, Maner, & Baumeister, 2010). To Galton’s chagrin, instead of demonstrating that criminals shared the same grotesque features, the resulting composites looked more attractive than the individual faces. This insight was the first demonstration of the average-is-attractive effect, with later studies replicating this classic finding with modern computer methods (Langlois & Roggman, 1990; see Rhodes, 2006, for a review).

Although Galton (1878) failed to demonstrate a criminal facial “type,” the logic of averaging across facial images to create prototypical representations of facial cues is elegant and very useful in understanding social perception. For example, by averaging across many female faces, one can visualize the prototypical female face (Figure 1B), because inconsistent facial attributes are averaged out, leaving only features that consistently indicate femininity. Averaging thus offers sensitive control over facial stimuli for hypothesis-driven research. Studies have utilized this approach, often in combination with morphing, to examine emotional expression (Calder, Young, Perrett, Etcoff, & Rowland, 1996), race (Hill, Bruce, & Akamatsu, 1995), sex (Bruce et al., 1993; R. Russell, 2010), age (Tiddeman, Burt, & Perrett, 2001; Figure 3), attractiveness (Perrett et al., 1994; Rhodes & Tremewan, 1996), stereotypes (Sutherland, Young, Mootz, & Oldmeadow, 2015), and first impressions (Sofer et al., 2015) and to demonstrate accurate facial judgments of the Big Five (Penton-Voak, Pound, Little, & Perrett, 2006).

Figure 3.

Face prototypes made by averaging the shape and surface properties for individual faces within 5-year age brackets: (A) 20–24, (B) 25–29, (C) 30–34, (D) 35–39, (E) 40–44, (F) 45–49, and (G) 50–54 years. (H) The dark lines depict the second youngest face prototype, and the shaded area illustrates the difference in average shape between this prototype and the oldest face prototype. (I) A caricature of the color difference between (J) the oldest prototype and an overall prototype face of all age groups matched for shape (not shown). (K) Contrast- and color-enhanced image made by amplifying RGB color differences between the overall prototype and a uniform gray image. From “Perception of age in adult Caucasian male faces: Computer graphic manipulation of shape and colour information” by D. M. Burt and D. I. Perrett, 1995, Proceedings of the Royal Society of London. Series B: Biological Sciences, 259, 137–143. Copyright 1995 by The Royal Society. Figure modified with permission.

To build average faces, most studies start with tightly controlled face images, photographed under standardized conditions (e.g., frontal facing, consistent expression, standard lighting; e.g., Perrett et al., 1994; see Figure 3). These controlled images are well suited to a hypothesis-driven approach aimed at isolating the effect of a manipulated facial cue on social perception. Although this approach has clearly been very productive, it relies on predetermined hypotheses about which cues are important, risks missing important facial cues that were eliminated by standardization, and ignores the natural covariation between cues as well as their relative real-world importance (Dotsch & Todorov, 2011; Vernon et al., 2014).

In contrast, data-driven methods work by sampling a wide range of potential cues as they occur in the world and then examining how naturally occurring combinations of cues influence social perception, rather than prespecifying which cues may be important (Adolphs, Nummenmaa, Todorov, & Haxby, 2016). The value of using a data-driven approach to examine facial impressions was originally noted in the pioneering work by Secord (1958) who anticipated that “impressions would be based on a pattern of cues; indeed the impressions themselves were conceived to be complex in organization” (p. 304). Recent advances in the creation of dimensional models of impressions have validated Secord’s early insights: Facial impressions and their underlying cues are often complex, multicollinear, and interactive (Oosterhof & Todorov, 2008; Sutherland, Young, et al., 2015; Vernon et al., 2014; Walker & Vetter, 2016). Importantly, data-driven methods are particularly effective in understanding complex judgments of high-dimensional stimuli such as faces, because investigation is not limited by experimental manipulation of small sets of cues at a time (see Adolphs et al., 2016).

Face averaging can be adapted into a data-driven approach, because the technique works even with unstandardized images. Recently, we employed data-driven face averaging by first sampling 1,000 naturally varying face images from the Internet and having these images rated on social impressions (Sutherland et al., 2013). By averaging across the individual faces rated highest and lowest on a given attribute, such as trustworthiness, we could create prototypically “trustworthy” and “untrustworthy” faces, thus visualizing the facial features that naturally cue trustworthiness (Figure 4). Similar data-driven face prototypes have been constructed to visualize complex social groups and associated stereotypes to test predictions arising from social psychological theories (Oldmeadow et al., 2013); for example, bankers look less warm but more competent than nurses, supporting the Stereotype Content Model (Fiske, Cuddy, & Glick, 2007). Likewise, data-driven prototypes have been successfully employed to visualize perceived Big Five personality traits (Sutherland, Rowley, et al., 2015) as well as social norms (Ginosar, Rakelly, Sachs, Yin, & Efros, 2015).

Figure 4.

Prototypically high (rightmost) and low (leftmost) attractive (A), trustworthy (B), and dominant (C) face averages, created by averaging together the 20 highest and least trustworthy, attractive, or dominant rated faces from 1,000 original images. Face images lying between the end point faces represent morphed blends of the prototype faces, in 20% steps. The face images demonstrate the facial cues naturally used to cue social impressions. For example, femininity, expression (smiling), head tilt, and age appear to cue perceived trustworthiness. From “Social inferences from faces: Ambient images generate a three-dimensional model,” by C. A. M. Sutherland, J. A. Oldmeadow, I. M. Santos, J. Towler, D. M. Burt, and A. W. Young, 2013, Cognition, 127, 105–118. Copyright 2012 by Elsevier. Figure modified with permission.

In these studies, perceivers’ natural impressions are examined with minimal researcher bias through sampling a large number of potential facial cues. Face averaging thereby offers a complementary approach to another data-driven method, reverse correlation (Dotsch & Todorov, 2011). Reverse correlation also seeks to model participants’ perceptions while minimizing researcher bias. Rather than sampling naturalistic cues to social perception, reverse correlation methods instead aim to visualize mental stereotypes from deliberately minimal visual information.

Facial Morphing and Transforming

Face morphing and transforming go further than averaging by parametrically varying images along a continuum (Figure 1C and D), thereby allowing systematic investigation of how facial cues contribute to social perception (Calder et al., 1996; Young et al., 1997). For example, morphing of facial emotional expressions has been employed to understand whether people represent these important social cues as discrete categories (Ekman & Friesen, 1975) or along emotion dimensions (J. A. Russell, 1980). People are better at discriminating pairs of faces categorized as representing different rather than the same expression, even when face pairs are equidistant along the morphed continuum (Calder et al., 1996; Calder, Young, Rowland, & Perrett, 1997; Etcoff & Magee, 1992). Better discrimination between categories than within categories is a hallmark property of categorical perception, demonstrating that emotional expression is represented categorically. Recently, Harris, Young, and Andrews (2012) combined these methods with neuroimaging to uncover the neural basis of emotion expression perception. They found that some brain regions represent facial emotional expression categorically, whereas others represent expression dimensionally (Figure 5; Harris, Young, & Andrews, 2012). The debate between these competing models (Ekman & Friesen, 1975; J. A. Russell, 1980) was thereby resolved by demonstrating that they hold at different levels of representation. Comparable methods could be used to test the dimensionality of social or personality models (e.g., Fiske et al., 2007; Oosterhof & Todorov, 2008; Osgood, 1969; Walker & Vetter, 2009; Wiggins, 1979).

Figure 5.

Facial morphing used to understand dimensional and categorical representations of emotional expression in the brain for (A) same identity faces and (B) different identity faces. A block design functional magnetic resonance imaging (fMRI)-adaptation paradigm paradigm was used in which “within” and “between” blocks involved images from morphed continua that were equally different in terms of the number of steps apart, but images in a within block crossed a perceived emotion category boundary (e.g., happy vs. sad) and images in a between block did not (e.g., slightly vs. very happy). The “same” condition involved repeating the same image in a trial block to create an estimate of the maximum possible neural adaptation as a baseline. The data (C) show that the posterior superior temporal sulcus was sensitive to any change in facial expression (i.e., it represents emotion dimensions), while the amygdala was sensitive only to the shift in perceived emotional category (i.e., it represents emotion categories). This pattern was observed regardless of changes in face identity. From “Morphing between expressions dissociates continuous from categorical representations of facial expressions in the human brain,” by R. J. Harris, A. W. Young, and T. J. Andrews, 2012, Proceedings of the National Academy of Sciences, 109, 21164–21169. Copyright 2012 by National Academy of Sciences. Reprinted with permission.

Morphing has also been used to examine the effect of social categorization on face processing, since perceived categories can be manipulated while keeping facial features identical (Michel, Corneille, & Rossion, 2007, 2010). Targets can also be manipulated on multiple social categories simultaneously (Hopper, Finklea, Winkielman, & Huber, 2014). Morphing has also been used to examine how people perceive and evaluate ambiguous social categories, such as mixed-race targets (Freeman et al., 2016; Michel et al., 2010; Webster, Kaping, Mizokami, & Duhamel, 2004). For example, mixed-race face morphs are reliably categorized as out-group members, although race category boundaries may be surprisingly flexible with experience (Webster et al., 2004; but see Freeman et al., 2016). Using morphed faces may be particularly useful in future, given the increasing interest in understanding ambiguous, intersectional, and dynamic social categories (e.g., Lick, Johnson, & Riskind, 2015; Pauker et al., 2009).

Whereas morphing creates a direct transition between two images, transforming applies transitions to new photographs (Figures 1D and 6), making this technique highly adaptable. For example, transforming has been used to study self-perception by manipulating images of the participants themselves. Instead of picking their real image, people often choose images as veridical that have actually been morphed to be more attractive (Epley & Whitchurch, 2008; likewise for long-term romantic partners: Penton-Voak, Rowe, & Williams, 2007). Face transforming’s ability to create photo-realistic images thereby offers a powerful new tool for investigating other important aspects of self-representation, including social and personal identity (e.g., Ellemers, Spears, & Doosje, 2002; Sedikides & Brewer, 2015).

Figure 6.

Male (top row) and female (bottom row) faces created to vary from low (left) to high (right) facial adiposity, by transforming with reference to average male or female faces that were either high or low on body mass index (BMI). BMI values of 16, 18, 20, 22, 24, and 26 (all kg/m²) are shown. Only face shape, not face surface, was transformed. Lower levels of facial adiposity were perceived as more attractive, and to a lesser extent as conveying higher leadership ability. Optimal BMI values were found to be relatively lower for female than male faces, perhaps reflecting gendered media representations of ideal body size. From “The effects of facial adiposity on attractiveness and perceived leadership ability,” by D. E. Re and D. I. Perrett, 2014, The Quarterly Journal of Experimental Psychology, 67, 676–686. Copyright 2013 by The Experimental Psychology Society by permission of Taylor & Francis Ltd. Reprinted with permission.

Transforming has also been useful in examining individual differences in social perception, because pairs of similar faces can be generated and people’s sensitivity to subtle differences in image pairs can be measured (Little, Jones, & DeBruine, 2011). For example, less dominant men are more sensitive to others’ dominance (Watkins, Jones, & DeBruine, 2010). Transforming politicians’ images has also demonstrated contextual shifts in leadership preferences, such that people prefer feminine-looking leaders during times of peace and masculine-looking leaders during war (Little, Burriss, Jones, & Verosky, 2007).

Facial Caricaturing

Face caricaturing involves creating a face image that looks more extreme than the original face, by transforming it away from a reference face (Brennan, 1982; Rhodes, Brennan, & Carey, 1987). Caricaturing can be used to exaggerate the effects of facial manipulation (Figure 1E). For example, it has been used to amplify subtle emotional expressions to test theoretical models of emotion representation (Calder et al., 2000, 1997; Figure 7).

Figure 7.

Examples of veridical (0%) and caricatured (+50%) representations of facial expressions. One Ekman face (Ekman & Friesen, 1975; Young et al 2002) is shown posing expressions associated with the six basic emotions: happiness, surprise, fear, sadness, disgust, and anger. Caricatured (+50%) representations were prepared relative to a neutral expression average face, and only shape was manipulated in these faces. From “Computer-enhanced emotion in facial expressions,” by A. J. Calder, A. W. Young, D. Rowland, and D. I. Perrett, 1997, Proceedings of the Royal Society of London B: Biological Sciences, 264, 919–925. Copyright 1997 by The Royal Society. Reprinted with permission.

Caricaturing individual faces away from a prototype reference face has been used to examine how social attributes are coded in faces. For example, caricaturing individual faces away from an average face often makes them look less attractive, supporting the average-is-attractive effect (Rhodes & Tremewan, 1996). Facial manipulation has also demonstrated limits to the averageness benefit, as hyperattractive faces can be made by caricaturing average faces made solely from attractive original faces (Perrett et al., 1994, 1998; Rhodes, Hickford, & Jeffery, 2000). These studies have provided a window into the different (stabilizing vs. directional) selection pressures on attractiveness (Perrett et al., 1994; Rhodes & Tremewan, 1996).

Caricaturing has also been used to investigate the nature of face coding, in combination with perceptual adaptation to shift the reference norm. For instance, after exposing people to extremely feminine caricatured faces, androgynous faces look masculine (i.e., shifted away from the reference face: Pond et al., 2013; Zhao, Seriès, Hancock, & Bednar, 2011). These studies demonstrate the flexible nature of the visual system, which is continuously updating as new social input is encountered. That is, what looks “male,” “attractive,” or “Asian” is also highly dynamic and dependent on experience (Clifford & Rhodes, 2005, review this literature).

Finally, morphing and caricaturing were used in a well-established test of facial emotion perception (Young, Perrett, Calder, Sprengelmeyer, & Ekman, 2002). This test contains individual faces morphed between different emotional expressions, with caricaturing used to generate more extreme expressions to manipulate task difficulty. It has been used to understand emotion perception impairment in many clinical conditions, including autism (Bruce & Young, 2012). Recently, this approach has been extended to examine neuropsychological impairment in formation of key impressions such as trustworthiness (Sprengelmeyer et al., 2016).

Advantages and Limitations

The researcher can choose between using facial manipulation in a hypothesis-driven or data-driven approach, each with advantages and limitations, as outlined previously. In a hypothesis-driven approach, carefully standardized face images are used to gain precise control over prespecified facial cues (Figure 3; cf. Calder et al., 1996; Re & Perrett, 2014), allowing causal inferences to be drawn regarding which facial features are driving social judgments, or to generate controlled but realistic stimuli for use in subsequent studies. Facial manipulation has more recently also been employed as a data-driven approach by averaging across naturally varying face images (Sutherland et al., 2013). Averaging across natural variation loses experimental control but allows prototypical faces to be created that reflect the holistic covariation and relative importance of real-world cues as they naturally occur (Figure 4). Data-driven approaches may be particularly important for understanding impressions in applied settings, where it is critical to study natural variation. Importantly, hypothesis-driven and data-driven approaches are complementary.

Regardless of the overall approach taken, a limitation is that the initial landmarking stage can be time consuming and somewhat subjective. Although automatic landmarking exists, manual adjustment is usually required (Sutherland, 2015, presents detailed guidelines). To overcome subjectivity, one person should landmark all faces or at least markers should not be confounded with the experimental conditions. Ideally, a second person should check landmarks. More landmarks and more consistent landmarking aid photo-realism, but take longer, with more chance of disagreement among markers. One major goal of computer scientists interested in face perception is to develop optimal automatic feature extraction (Kemelmacher-Shlizerman, Suwajanakorn, & Seitz, 2014; Zhu & Ramanan, 2012).

Averaging also tends to produce images that look more attractive than the original photographs (Langlois & Roggman, 1990), making it harder to manipulate negative traits. Thus, where distinctiveness is an important cue, face averages may not be representative. Similarly, older or masculine faces are harder to generate, because averaging tends to smooth out age-related blemishes and wrinkles (Kemelmacher-Shlizerman et al., 2014; Tiddeman et al., 2001) as well as smoothing over masculine features like stubble or angular square jaws (Pond et al., 2013; Zhao et al., 2011). Advanced “wavelet texture” techniques can preserve fine surface detail and can easily run in Psychomorph software (Tiddeman et al., 2001). To minimize loss of shape information when averaging, feature points should be marked in consistent positions on all constituent faces. Including only extreme images can help maintain distinctiveness (e.g., Sutherland et al., 2013, created facial averages from images of elderly adults that appear reasonably old).

A more subtle point is that it is not always obvious what the reference face should be. For example, if manipulating emotional expression, should the reference norm be an unexpressive face, the average of all expressions, or just any other expression? It turns out that any of these reference faces will “work” to create easily recognized caricatures, which is theoretically important but not intuitive (Burton, Jeffery, Skinner, Benton, & Rhodes, 2017; Calder et al., 2000; Young et al., 1997). Although the distinction between emotional expression reference faces may not be vital, other studies have shown that faces do appear to be coded (at least partly) in relation to gender- and race-specific norms (Armann, Jeffery, Calder, & Rhodes, 2011; Little, DeBruine, & Jones, 2005). Do these findings mean that reference and manipulated faces should represent the same social category? Morphing relative to a reference face from the same rather than a different social category is often easier to interpret and generates clearer images (Rennels, Bronstad, & Langlois, 2008; Rhodes, 2006; but see DeBruine, Jones, Smith, & Little, 2010). Yet, where social categories contribute to impressions (e.g., female faces look more trustworthy; Figure 4), controlling them simply removes important cues. Importantly, answering these questions has itself offered insight into social category norms.

Regardless of the choice of reference face, all images used are best kept fairly similar or the process can look unnatural. For example, transforming between open and closed mouths will create artifacts around the lips. Since caricatures push beyond the boundaries of the original images, they will eventually always look grotesque (Pond et al., 2013; Zhao et al., 2011). Manipulating individual photographs can look very realistic (Figure 6) and limiting transformation to face shape (keeping surface consistent) can also help (Watkins et al., 2010). However, surface cues can be important (e.g., for male attractiveness: Said & Todorov, 2011; Torrance, Wincenciak, Hahn, DeBruine, & Jones, 2014). Nevertheless, even if images are too unnatural to be used as stimuli, they can still be very informative to researchers in depicting which attributes are changing.

Finally, a key question affecting all facial manipulation techniques is whether the outcome of multiple linear transforms (in landmarks, pixel colors, etc.) is itself linear, as is widely assumed. As outlined at the start, blending may involve nonlinear transformations when tessellated surface areas are compressed or expanded. Certainly, morphing and transforming steps are often not perceived as linear (Pond et al., 2013), which has provided interesting insights into social categorization (Calder et al., 1996).

A strong approach to address all of these limitations is to first use real face photographs and then provide converging evidence with experimentally manipulated face images. For example, averaging can efficiently replicate findings based on unmanipulated face images with different stimuli (and participants), demonstrating that initial results were not due to noise in the original images (as averaging removes noise). Careful pretesting ensures that images are perceived as intended (Pond et al., 2013). Moreover, unlike FaceGen images, manipulated images as described here are built directly from photographs of real faces, thereby preserving important covariation between cues. They also preserve important surface cues, thus look more natural.

Getting Started

We outline three facial manipulation programs: Fantamorph, Morpheus, and Psychomorph (Table 1). These were chosen because they are readily available, inexpensive/free, and have now been used to study social categorization and stereotyping, impression formation, individual and cultural differences, and clinical impairment (e.g., Frenkel, Lamy, Algom, & Bar-Haim, 2009; Ishii, Miyamoto, Mayama, & Niedenthal, 2011; Sofer et al., 2015; Wu, Laeng, & Magnussen, 2012). GryphonMorph has been used in papers on facial attractiveness (e.g., Rennels et al., 2008) but is now discontinued.

Table 1.

Leading Morphing Software.

Feature	Fantamorph	Morpheus Photo	Psychomorph
Starting number of landmarks	112—can be adapted	Varies—user manually adds landmarks	Varies—user usually reuses a standard template (e.g., 179 landmarks)
Automatic landmarks?	Y—rough	N	Y—rough
Extract landmark locations?	Y	N	Y
Averaging?	Unlimited	Two images at a time	Unlimited
Morphing?	Y	Y	Y
Transforming?	N	N	Y
Caricaturing?	Y—shape only	N	Y
Separate manipulation of shape and surface?	Y—not caricaturing	N	Y
Batch process files?	N	N	Y
Mac/Windows	Both	Both	Both
Unique citation count^a	325	69	239

Note. N = no; Y = yes

^aCitation count via Google Scholar (May 17, 2016) using the titles as search terms plus the word “face.” These counts underestimate how widely these methods have been used, as unfortunately not all papers cite the software used. There were an additional 7 overlapping citations, removed from all citation counts, and an additional 196 unique citations for Gryphon Morph (now discontinued).

Fantamorph (www.fantamorph.com/index.html) and Morpheus (www.morpheussoftware.net/) are commercial programs costing less than US$100, which offer basic face manipulation capability. They come with user guides, although not aimed at researchers. Psychomorph is free software developed by Perrett, Tiddeman, and their colleagues (http://users.aber.ac.uk/bpt/jpsychomorph; Burt & Perrett, 1995; Tiddeman et al., 2001). Psychomorph has unlimited averaging, transforming, and caricaturing and clearer descriptions of underlying calculations. For advanced users, Psychomorph can automatically process multiple images, and the code (JavaScript) can be customized. The Psychomorph site includes help pages, and a user guide exists (Sutherland, 2015). We recommend that new users of morphing software start with extreme faces (e.g., highly different colors/shapes) to understand available features.

Beyond the Face Photograph

Although we have focused on facial image manipulation, the principles outlined here apply to other aspects of social perception, including dynamic or 3-D faces, bodies, and voices (Rowland & Perrett, 1995). MorphAnalyser can manipulate 3-D faces (http://cherry.dcs.aber.ac.uk:8080/wiki/MorphAnalyser) and has been used to investigate ideal facial adiposity and how it varies with sociocultural norms (Coetzee, Re, Perrett, Tiddeman, & Xiao, 2011). As mentioned previously, Vetter and colleagues have also developed sophisticated facial manipulation software (“morphable models”), which can manipulate 3-D (and 2-D) face images on social impressions (Blanz & Vetter, 1999; Vetter & Walker, 2011). Psychomorph can use sequences of frames to morph dynamic input (http://cherry.dcs.aber.ac.uk:8080/wiki/videomorph) and has been used to understand dynamic perception of pro-social traits (Morrison, Clark, Tiddeman, & Penton-Voak, 2010). Psychomorph can also be applied to bodies (Rowland & Perrett, 1995).

Finally, voice manipulation is a relatively new development. Interestingly, preliminary findings appear to parallel those from face perception research (Latinus & Belin, 2011; Schweinberger, Kawahara, Simpson, Skuk, & Zäske, 2014, review this literature). For example, average voices sound attractive (Bruckert et al., 2010). Examining the social perception of dynamic faces, bodies, and voices using averaging and morphing is ripe for further study.

Conclusions

Facial image manipulation has led to many insights into social perception, including the basis of facial attractiveness, the relationship between visual and conceptual stereotyping, and the role of individual differences in impression formation. The approach has practical applications, including the ability to test clinical social impairment. We hope that this review stimulates more researchers to use these techniques, with a better appreciation of the breadth of insights already achieved and the strengths and limitations of the approach. We see great scope for future use of these methods within social psychology.

Footnotes

Acknowledgments

We thank Julian Oldmeadow, Nichola Burton, David Lick, and Stephen Pond for their invaluable comments on an earlier version of the article.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Australian Research Council (ARC) Centre of Excellence in Cognition and its Disorders (CE110001021), an ARC Discovery Outstanding Researcher Award to Rhodes (DP130102300), and an ARC Discovery grant to Rhodes, Sutherland, and Young (DP170104602).

References

Adams

R. B.

Kleck

R. E.

(2003). Perceived gaze direction and the processing of facial displays of emotion. Psychological Science, 14, 644–647. doi:10.1046/j.0956-7976.2003.psci_1479.x

Adolphs

Nummenmaa

Todorov

Haxby

(2016). Data-driven approaches in the investigation of social perception. Philosophical Transactions of the Royal Society B, 371, 20150367. doi:10.1098/rstb.2015.0367

Armann

Jeffery

Calder

A. J.

Rhodes

(2011). Race-specific norms for coding face identity and a functional role for norms. Journal of Vision, 11, 9–9. doi:10.1167/11.13.9

Blair

I. V.

Judd

C. M.

Fallman

J. L.

(2004). The automaticity of race and Afrocentric facial features in social judgments. Journal of Personality and Social Psychology, 87, 763. doi:10.1037/0022-3514.87.6.763

Blanz

Vetter

(1999). A morphable model for the synthesis of 3D faces. In Proceedings of the 26th Annual Conference on Computer Graphics and Interactive Techniques (pp. 187–194). New York, NY: ACM Press/Addison-Wesley Publishing Co.

Brennan

(1982). Caricature generator. Massachusetts Institute of Technology. Retrieved from http://dspace.mit.edu/handle/1721.1/15743

Bruce

Burton

A. M.

Hanna

Healey

Mason

Coombes

… Linney

(1993). Sex discrimination: How do we tell the difference between male and female face? Perception, 22, 131–152. doi:10.1068/p220131

Bruce

Young

A. W.

(1986). Understanding face recognition. British Journal of Psychology, 77, 305–327. doi:10.1111/j.2044-8295.1986.tb02199.x

Bruce

Young

A. W.

(2012). Face perception. New York, NY: Psychology Press.

10.

Bruckert

Bestelmeyer

Latinus

Rouger

Charest

Rousselet

G. A.

… Belin

(2010). Vocal attractiveness increases by averaging. Current Biology, 20, 116–120. doi:10.1016/j.cub.2009.11.034

11.

Burt

D. M.

Perrett

D. I.

(1995). Perception of age in adult Caucasian male faces: Computer graphic manipulation of shape and colour information. Proceedings of the Royal Society of London. Series B: Biological Sciences, 259, 137–143. doi:10.1098/rspb.1995.0021

12.

Burton

Jeffery

Skinner

A. L.

Benton

C. P.

Rhodes

. (2017). Is a neutral or an average expression closer to the norm of expression space? Manuscript submitted for publication.

13.

Calder

A. J.

Rhodes

Johnson

Haxby

J. V.

(Eds.). (2011). Oxford handbook of face perception. Oxford, England: Oxford University Press.

14.

Calder

A. J.

Rowland

Young

A. W.

Nimmo-Smith

Keane

Perrett

D. I.

(2000). Caricaturing facial expressions. Cognition, 76, 105–146. doi:10.1016/S0010-0277(00)00074-3

15.

Calder

A. J.

Young

A. W.

Perrett

D. I.

Etcoff

N. L.

Rowland

(1996). Categorical perception of morphed facial expressions. Visual Cognition, 3, 81–118. doi:10.1080/713756735

16.

Calder

A. J.

Young

A. W.

Rowland

Perrett

D. I.

(1997). Computer-enhanced emotion in facial expressions. Proceedings of the Royal Society of London B: Biological Sciences, 264, 919–925. doi:10.1098/rspb.1997.0127

17.

Clifford

C. W. G.

Rhodes

(2005). Fitting the mind to the world: Adaptation and after-effects in high-level vision (Vol. 2). Oxford, England: Oxford University Press.

18.

Coetzee

Perrett

D. I.

Tiddeman

B. P.

Xiao

(2011). Judging the health and attractiveness of female faces: Is the most attractive level of facial adiposity also considered the healthiest? Body Image, 8, 190–193. doi:10.1016/j.bodyim.2010.12.003

19.

Corneille

Huart

Becquart

Brédart

(2004). When memory shifts toward more typical category exemplars: Accentuation effects in the recollection of ethnically ambiguous faces. Journal of Personality and Social Psychology, 86, 236. doi:10.1037/0022-3514.86.2.236

20.

Crookes

Ewing

Gildenhuys

Kloth

Hayward

W. G.

Oxner

… Rhodes

(2015). How well do computer-generated faces tap face expertise? PloS One, 10, e0141353. doi:10.1371/journal.pone.0141353

21.

DeBruine

L. M.

Jones

B. C.

Smith

F. G.

Little

A. C.

(2010). Are attractive men’s faces masculine or feminine? The importance of controlling confounds in face stimuli. Journal of Experimental Psychology: Human Perception and Performance, 36, 751. doi:10.1037/a0016457

22.

Dotsch

Todorov

(2011). Reverse correlating social face perception. Social Psychological and Personality Science, 3, 562–571. doi:10.1177/1948550611430272

23.

Dotsch

Wigboldus

D. H. J.

Langner

van Knippenberg

(2008). Ethnic out-group faces are biased in the prejudiced mind. Psychological Science, 19, 978–980. doi:10.1111/j.1467-9280.2008.02186.x

24.

Ekman

Friesen

W. V.

(1975). Pictures of facial affect. Palo Alto, CA: Consulting Psychologists Press.

25.

Ellemers

Spears

Doosje

(2002). Self and social identity. Annual Review of Psychology, 53, 161–186. doi:10.1146/annurev.psych.53.100901.135228

26.

Epley

Whitchurch

(2008). Mirror, mirror on the wall: Enhancement in self-recognition. Personality and Social Psychology Bulletin, 34, 1159–1170. doi:10.1177/0146167208318601

27.

Etcoff

N. L.

Magee

J. J.

(1992). Categorical perception of facial expressions. Cognition, 44, 227–240. doi:10.1016/0010-0277(92)90002-Y

28.

Fiske

S. T.

Cuddy

A. J. C.

Glick

(2007). Universal dimensions of social cognition: Warmth and competence. Trends in Cognitive Sciences, 11, 77–83. doi:10.1016/j.tics.2006.11.005

29.

Freeman

J. B.

Ambady

(2011). A dynamic interactive theory of person construal. Psychological Review, 118, 247–279. doi:10.1037/a0022327

30.

Freeman

J. B.

Pauker

Sanchez

D. T.

(2016). A perceptual pathway to bias: Interracial exposure reduces abrupt shifts in real-time race perception that predict mixed-race bias. Psychological Science, 27, 502–517. doi:10.1177/0956797615627418

31.

Frenkel

T. I.

Lamy

Algom

Bar-Haim

(2009). Individual differences in perceptual sensitivity and response bias in anxiety: Evidence from emotional faces. Cognition and Emotion, 23, 688–700. doi:10.1080/02699930802076893

32.

Galton

(1878). Composite portraits. The Journal of the Anthropological Institute of Great Britain and Ireland, 8, 132–144. doi:10.2307/2841021

33.

Ginosar

Rakelly

Sachs

Yin

Efros

A. A.

(2015). A century of portraits: A visual historical record of American high school yearbooks. arXiv:1511.02575 [Cs]. Retrieved from http://arxiv.org/abs/1511.02575

34.

Harris

R. J.

Young

A. W.

Andrews

T. J.

(2012). Morphing between expressions dissociates continuous from categorical representations of facial expression in the human brain. Proceedings of the National Academy of Sciences, 109, 21164–21169. doi:10.1073/pnas.1212207110

35.

Hess

Adams

R. B.

Jr Kleck

R. E.

(2008). The role of facial expression in person perception. In

N. Ambady

Skowronski

J. J.

(Eds.), First impressions (pp. 234–254). New York, NY: Guilford Press.

36.

Hill

Bruce

Akamatsu

(1995). Perceiving the sex and race of faces: The role of shape and colour. Proceedings of the Royal Society of London. Series B: Biological Sciences, 261, 367–373. doi:10.1098/rspb.1995.0161

37.

Hopper

W. J.

Finklea

K. M.

Winkielman

Huber

D. E.

(2014). Measuring sexual dimorphism with a race–gender face space. Journal of Experimental Psychology: Human Perception and Performance, 40, 1779–1788. doi:10.1037/a0037743

38.

Hugenberg

Sacco

D. F.

(2008). Social categorization and stereotyping: How social categorization biases person perception and face memory. Social and Personality Psychology Compass, 2, 1052–1072. doi:10.1111/j.1751-9004.2008.00090.x

39.

Ishii

Miyamoto

Mayama

Niedenthal

P. M.

(2011). When your smile fades away: Cultural differences in sensitivity to the disappearance of smiles. Social Psychological and Personality Science, 2, 516–522. doi:10.1177/1948550611399153

40.

Johnson

K. L.

Freeman

J. B.

Pauker

(2012). Race is gendered: How covarying phenotypes and stereotypes bias sex categorization. Journal of Personality and Social Psychology, 102, 116–131. doi:10.1037/a0025335

41.

Kemelmacher-Shlizerman

Suwajanakorn

Seitz

S. M.

(2014). Illumination-aware age progression. In Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. (pp. 3334–3341). IEEE Computer Society. Retrieved from https://ieeexplore-ieee-org.web.bisu.edu.cn/xpls/abs_all.jsp?arnumber=6909822

42.

Langlois

J. H.

Roggman

L. A.

(1990). Attractive faces are only average. Psychological Science, 1, 115–121. doi:10.1111/j.1467-9280.1990.tb00079.x

43.

Latinus

Belin

(2011). Human voice perception. Current Biology, 21, R143–R145. doi:10.1016/j.cub.2010.12.033

44.

Lavater

J. C.

(1778). Physiognomische Fragmente: zur Beförderung der Menschenkenntniss und Menschenliebe (Vol. 4). Leipzig, Germany: Weidmanns erben und Reich.

45.

Lick

D. J.

Johnson

K. L.

Riskind

R. G.

(2015). Haven’t I seen you before? Straight men who are insecure of their masculinity remember gender-atypical faces. Group Processes & Intergroup Relations, 18, 131–152. doi:10.1177/1368430214538324

46.

Lienhard

Ladret

Caplier

(2015). How to predict the global instantaneous feeling induced by a facial picture? Signal Processing: Image Communication, 39, 473–486. doi:10.1016/j.image.2015.04.002

47.

Little

A. C.

Burriss

R. P.

Jones

B. C.

Verosky

S. C.

(2007). Facial appearance affects voting decisions. Evolution and Human Behavior, 28, 18–27. doi:10.1016/j.evolhumbehav.2006.09.002

48.

Little

A. C.

DeBruine

L. M.

Jones

B. C.

(2005). Sex-contingent face after-effects suggest distinct neural populations code male and female faces. Proceedings of the Royal Society B: Biological Sciences, 272, 2283–2287. doi:10.1098/rspb.2005.3220

49.

Little

A. C.

Jones

B. C.

DeBruine

L. M.

(2011). Facial attractiveness: Evolutionary based research. Philosophical Transactions of the Royal Society B: Biological Sciences, 366, 1638–1659. doi:10.1098/rstb.2010.0404

50.

Lundqvist

Flykt

Öhman

(1998). The Karolinska Directed Emotional Faces - KDEF, [CD ROM]. Solna, Sweden: Department of Clinical Neuroscience, Psychology section, Karolinska Institutet.

51.

Mangini

M. C.

Biederman

(2004). Making the ineffable explicit: Estimating the information employed for face classifications. Cognitive Science, 28, 209–226. doi:10.1016/j.cogsci.2003.11.004

52.

Michel

Corneille

Rossion

(2007). Race categorization modulates holistic face encoding. Cognitive Science, 31, 911–924. doi:10.1080/03640210701530805

53.

Michel

Corneille

Rossion

(2010). Holistic face encoding is modulated by perceived face race: Evidence from perceptual adaptation. Visual Cognition, 18, 434–455. doi:10.1080/13506280902819697

54.

Morrison

E. R.

Clark

A. P.

Tiddeman

B. P.

Penton-Voak

I. S.

(2010). Manipulating shape cues in dynamic human faces: Sexual dimorphism is preferred in female but not male faces. Ethology, 116, 1234–1243. doi:10.1111/j.1439-0310.2010.01839.x

55.

Oldmeadow

J. A.

Sutherland

C. A. M.

Young

A. W.

(2013). Facial stereotype visualization through image averaging. Social Psychological and Personality Science, 4, 615–623. doi:10.1177/1948550612469820

56.

Oosterhof

N. N.

Todorov

(2008). The functional basis of face evaluation. PNAS, 105, 11087–11092. doi:10.1073/pnas.0805664105

57.

Osgood

C. E.

(1969). On the whys and wherefores of E, P, and A. Journal of Personality and Social Psychology, 12, 194–199.

58.

O’Toole

A. J.

(2011). Cognitive and computational approaches to face recognition. In Calder

A. J.

Rhodes

Johnson

Haxby

J. V.

(Eds.), Oxford handbook of face perception (pp. 15–30). Oxford, England: Oxford University Press.

59.

Pauker

Weisbuch

Ambady

Sommers

S. R.

Adams

R. B.

Jr Ivcevic

(2009). Not so black and white: Memory for ambiguous group members. Journal of Personality and Social Psychology, 96, 795. doi:10.1037/a0013265

60.

Penton-Voak

I. S.

Pound

Little

A. C.

Perrett

D. I.

(2006). Personality judgments from natural and composite facial images: More evidence for a ‘kernel of truth’ in social perception. Social Cognition, 24, 607–640. doi:10.1521/soco.2006.24.5.607

61.

Penton-Voak

I. S.

Rowe

A. C.

Williams

(2007). Through rose-tinted glasses: Relationship satisfaction and representations of partners’ facial attractiveness. Journal of Evolutionary Psychology, 5, 169–181. doi:10.1556/JEP.2007.1021

62.

Perrett

D. I.

Lee

K. J.

Penton-Voak

I. S.

Rowland

Yoshikawa

Burt

D. M.

… Akamatsu

(1998). Effects of sexual dimorphism on facial attractiveness. Nature, 394, 884–887. doi:10.1038/29772

63.

Perrett

D. I.

May

K. A.

Yoshikawa

(1994). Facial shape and judgements of female attractiveness. Nature, 368, 239–241. doi:10.1038/368239a0

64.

Pond

Kloth

McKone

Jeffery

Irons

Rhodes

(2013). Aftereffects support opponent coding of face gender. Journal of Vision, 13, 16–16. doi:10.1167/13.14.16

65.

Quinn

K. A.

Macrae

C. N.

(2011). The face and person perception: Insights from social cognition. British Journal of Psychology, 102, 849–867. doi:10.1111/j.2044-8295.2011.02030.x

66.

D. E.

Perrett

D. I.

(2014). The effects of facial adiposity on attractiveness and perceived leadership ability. The Quarterly Journal of Experimental Psychology, 67, 676–686. doi:10.1080/17470218.2013.825635

67.

Rennels

J. L.

Bronstad

P. M.

Langlois

J. H.

(2008). Are attractive men’s faces masculine or feminine? The importance of type of facial stimuli. Journal of Experimental Psychology: Human Perception and Performance, 34, 884. doi:10.1037/0096-1523.34.4.884

68.

Rhodes

(2006). The evolutionary psychology of facial beauty. Annual Review of Psychology, 57, 199–226. doi:10.1146/annurev.psych.57.102904.190208

69.

Rhodes

Brennan

Carey

(1987). Identification and ratings of caricatures: Implications for mental representations of faces. Cognitive Psychology, 19, 473–497. doi:10.1016/0010-0285(87)90016-8

70.

Rhodes

Hickford

Jeffery

(2000). Sex-typicality and attractiveness: Are supermale and superfemale faces super-attractive? British Journal of Psychology, 91, 125–140. doi:10.1348/000712600161718

71.

Rhodes

Tremewan

(1996). Averageness, exaggeration, and facial attractiveness. Psychological Science, 7, 105–110. doi:10.1111/j.1467-9280.1996.tb00338.x

72.

Rowland

D. A.

Perrett

D. I.

(1995). Manipulating facial appearance through shape and color. Computer Graphics and Applications, IEEE, 15, 70–76. doi:10.1109/38.403830

73.

Rule

N. O.

Krendl

A. C.

Ivcevic

Ambady

(2013). Accuracy and consensus in judgments of trustworthiness from faces: Behavioral and neural correlates. Journal of Personality and Social Psychology, 104, 409–426. doi:10.1037/a0031050

74.

Russell

J. A.

(1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161–1178. doi:10.1037/h0077714

75.

Russell

(2010). Why cosmetics work. The Science of Social Vision, 1, 186–204.

76.

Said

C. P.

Todorov

(2011). A statistical model of facial attractiveness. Psychological Science, 22, 1183–1190. doi:10.1177/0956797611419169

77.

Schweinberger

S. R.

Kawahara

Simpson

A. P.

Skuk

V. G.

Zäske

(2014). Speaker perception. Wiley Interdisciplinary Reviews: Cognitive Science, 5, 15–25. doi:10.1002/wcs.1261

78.

Secord

P. F.

(1958). Facial features and inference processes in interpersonal perception. In Tagiuri

Petrullo

(Eds.), Person perception and interpersonal behavior (pp. 300–315). Stanford, CA: Stanford University Press.

79.

Sedikides

Brewer

M. B.

(2015). Individual self, relational self, collective self. Hove, United Kingdom: Psychology Press.

80.

Sofer

Dotsch

Wigboldus

D. H. J.

Todorov

(2015). What is typical is good: The influence of face typicality on perceived trustworthiness. Psychological Science, 26, 39–47. doi:10.1177/0956797614554955

81.

Sormaz

Young

A. W.

Andrews

T. J.

(2016). Contributions of feature shapes and surface cues to the recognition of facial expressions. Vision Research, 127, 1–10. doi:10.1016/j.visres.2016.07.002

82.

Sprengelmeyer

Young

A. W.

Baldas

E.-M.

Ratheiser

Sutherland

C. A. M.

Müller

H.-P.

… Orth

(in press). The neuropsychology of first impressions: Evidence from Huntington’s disease. Cortex, 85, 100–115. doi:10.1016/j.cortex.2016.10.006.

83.

Stillman

T. F.

Maner

J. K.

Baumeister

R. F.

(2010). A thin slice of violence: Distinguishing violent from nonviolent sex offenders at a glance. Evolution and Human Behavior, 31, 298–303. doi:10.1016/j.evolhumbehav.2009.12.001

84.

Sutherland

C. A. M.

(2015). A basic guide to Psychomorph. University of York. Retrieved from https://dlib.york.ac.uk/yodl/app/home/detail?id=york:847354&ref=browse

85.

Sutherland

C. A. M.

Oldmeadow

J. A.

Santos

I. M.

Towler

Burt

D. M.

Young

A. W.

(2013). Social inferences from faces: Ambient images generate a three-dimensional model. Cognition, 127, 105–118. doi:10.1016/j.cognition.2012.12.001

86.

Sutherland

C. A. M.

Rowley

Amoaku

Daguzan

Kidd-Rossiter

Maceviciute

Young

A. W.

(2015). Personality judgements from everyday images of faces. Frontiers in Psychology, 6, 1–11. doi:10.3389/fpsyg.2015.01616

87.

Sutherland

C. A. M.

Young

A. W.

Mootz

C. A.

Oldmeadow

J. A.

(2015). Face gender and stereotypicality influence facial trait evaluation: Counter-stereotypical female faces are negatively evaluated. British Journal of Psychology, 106, 186–208. doi:10.1111/bjop.12085

88.

Taylor

S. E.

Fiske

S. T.

Etcoff

N. L.

Ruderman

A. J.

(1978). Categorical and contextual bases of person memory and stereotyping. Journal of Personality and Social Psychology, 36, 778. doi:10.1037/0022-3514.36.7.778

89.

Tiddeman

Burt

Perrett

D. I.

(2001). Prototyping and transforming facial textures for perception research. Computer Graphics and Applications, IEEE, 21, 42–50. doi:10.1109/38.946630

90.

Todorov

Olivola

C. Y.

Dotsch

Mende-Siedlecki

(2015). Social attributions from faces: Determinants, consequences, accuracy, and functional significance. Annual Review of Psychology, 66, 519–545. doi:10.1146/annurev-psych-113011-143831

91.

Torrance

J. S.

Wincenciak

Hahn

A. C.

DeBruine

L. M.

Jones

B. C.

(2014). The relative contributions of facial shape and surface information to perceptions of attractiveness and dominance. PLOS ONE, 9, e104415. doi:10.1371/journal.pone.0104415

92.

Vernon

R. J. W.

Sutherland

C. A. M.

Young

A. W.

Hartley

(2014). Modeling first impressions from highly variable facial images. PNAS, 111, E3353–E3361. doi:10.1073/pnas.1409860111

93.

Vetter

Walker

(2011). Computer-generated images in face perception. In Calder

A. J.

Rhodes

Johnson

Haxby

J. V.

(Eds.), Oxford handbook of face perception (pp. 388–399). Oxford, England: Oxford University Press.

94.

Walker

Jiang

Vetter

Sczesny

(2011). Universals and cultural differences in forming personality trait judgments from faces. Social Psychological and Personality Science, 2, 609–617. doi:10.1177/1948550611402519

95.

Walker

Vetter

(2009). Portraits made to measure: Manipulating social judgments about individuals with a statistical face model. Journal of Vision, 9, 1–13. doi:10.1167/9.11.12

96.

Walker

Vetter

(2016). Changing the personality of a face: Perceived Big Two and Big Five personality factors modeled in real photographs. Journal of Personality and Social Psychology, 110, 609–624. doi:10.1037/pspp0000064

97.

Watkins

C. D.

Jones

B. C.

DeBruine

L. M.

(2010). Individual differences in dominance perception: Dominant men are less sensitive to facial cues of male dominance. Personality and Individual Differences, 49, 967–971. doi:10.1093/beheco/arq091

98.

Webster

M. A.

Kaping

Mizokami

Duhamel

(2004). Adaptation to natural facial categories. Nature, 428, 557–561. doi:10.1038/nature02420

99.

Webster

M. A.

MacLeod

D. I. A.

(2011). Visual adaptation and face perception. Philosophical Transactions of the Royal Society B: Biological Sciences, 366, 1702–1725. doi:10.1098/rstb.2010.0360

100.

Wiggins

J. S.

(1979). A psychological taxonomy of trait-descriptive terms: The interpersonal domain. Journal of Personality and Social Psychology, 37, 395–412. doi:10.1037/0022-3514.37.3.395

101.

E. X. W.

Laeng

Magnussen

(2012). Through the eyes of the own-race bias: Eye-tracking and pupillometry during face recognition. Social Neuroscience, 7, 202–216. doi:10.1080/17470919.2011.596946

102.

Young

A. W.

Perrett

D. I.

Calder

A. J.

Sprengelmeyer

Ekman

(2002). Facial expressions of emotion: Stimuli and tests (FEEST). Bury St Edmunds, England: Thames Valley Test Company.

103.

Young

A. W.

Rowland

Calder

A. J.

Etcoff

N. L.

Seth

Perrett

D. I.

(1997). Facial expression megamix: Tests of dimensional and category accounts of emotion recognition. Cognition, 63, 271–313. doi:10.1016/S0010-0277(97)00003-6

104.

Zebrowitz

L. A.

(2005). Appearance DOES Matter. Science, 308, 1565–1566. doi:10.1126/science.1114170

105.

Zebrowitz

L. A.

Kikuchi

Fellous

J. M.

(2010). Facial resemblance to emotions: Group differences, impression effects, and race stereotypes. Journal of Personality and Social Psychology, 98, 175–189. doi:10.1037/a0017990

106.

Zebrowitz

L. A.

Montepare

J. M.

(1992). Impressions of babyfaced individuals across the life span. Developmental Psychology, 28, 1143–1152. doi:10.1037/0012-1649.28.6.1143

107.

Zhao

Seriès

Hancock

P. J.

Bednar

J. A.

(2011). Similar neural adaptation mechanisms underlying face gender and tilt aftereffects. Vision Research, 51, 2021–2030. doi:10.1016/j.visres.2011.07.014

108.

Zhu

Ramanan

(2012). Face detection, pose estimation, and landmark localization in the wild. In Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. (pp. 2879–2886). Retrieved from https://ieeexplore-ieee-org.web.bisu.edu.cn/xpls/abs_all.jsp?arnumber=6248014.