The role of texture summary statistics in material recognition from drawings and photographs.

Material depictions in artwork are useful tools for revealing image features that support material categorization. For example, artistic recipes for drawing specific materials make explicit the critical information leading to recognizable material properties (Di Cicco, Wjintjes, & Pont, 2020) and investigating the recognizability of material renderings as a function of their visual features supports conclusions about the vocabulary of material perception. Here, we examined how the recognition of materials from photographs and drawings was affected by the application of the Portilla-Simoncelli texture synthesis model. This manipulation allowed us to examine how categorization may be affected differently across materials and image formats when only summary statistic information about appearance was retained. Further, we compared human performance to the categorization accuracy obtained from a pretrained deep convolutional neural network to determine if observers' performance was reflected in the network. Although we found some similarities between human and network performance for photographic images, the results obtained from drawings differed substantially. Our results demonstrate that texture statistics play a variable role in material categorization across rendering formats and material categories and that the human perception of material drawings is not effectively captured by deep convolutional neural networks trained for object recognition.

Putting people in context: ERP responses to bodies in natural scenes.

The N190 is a body-sensitive ERP component that responds to images of human bodies in different poses. In natural settings, bodies vary in posture and appear within complex, cluttered environments, frequently with other people. In many studies, however, such variability is absent. How does the N190 response change when observers see images that incorporate these sources of variability? In two experiments (N = 16 each), we varied the natural appearance of upright and inverted bodies to examine how the N190 amplitude, latency, and the Body-Inversion Effect (BIE) were affected by natural variability. In Experiment 1, we varied the number of people present in upright and inverted naturalistic scenes such that only one body, a subitizable number of bodies, or a "crowd" was present. In Experiment 2, we varied the natural body appearance by presenting bodies either as silhouettes or with photographic detail. Further, we varied the natural background appearance by either removing it or presenting individual bodies within a rich environment. Using component-based analyses of the N190, we found that the number of bodies in a scene reduced the N190 amplitude, but didn't affect the BIE (Experiment 1). Naturalistic body and background appearance (Experiment 2) also affected the N190, such that component amplitude was dramatically reduced by naturalistic appearance. To complement this analysis, we examined the contribution of spatiotemporal features (i.e., electrode × time point amplitude) via SVM decoding. This technique allows us to examine which timepoints across the entire waveform contribute the most to successful decoding of body orientation in each condition. This analysis revealed that later timepoints (after 300ms) contribute most to successful orientation decoding. These results demonstrate that natural appearance variability affects body processing at the N190 and that later ERP components may make important contributions to body processing in natural scenes.

Not the norm: Face likeness is not the same as similarity to familiar face prototypes.

Face images depicting the same individual can differ substantially from one another. Ecological variation in pose, expression, lighting, and other sources of appearance variability complicates the recognition and matching of unfamiliar faces, but acquired familiarity leads to the ability to cope with these challenges. Among the many ways that face of the same individual can vary, some images are judged to be better likenesses of familiar individuals than others. Simply put, these images look more like the individual under consideration than others. But what does it mean for an image to be a better likeness than another? Does likeness entail typicality, or is it something distinct from this? We examined the relationship between the likeness of face images and the similarity of those images to average images of target individuals using a set of famous faces selected for reciprocal familiarity/unfamiliarity across US and UK participants. We found that though likeness judgments are correlated with similarity-to-prototype judgments made by both familiar and unfamiliar participants, this correlation was smaller than the correlation between similarity judgments made by different participant groups. This implies that while familiarity weakens the relationship between likeness and similarity-to-prototype judgments, it does not change similarity-to-prototype judgments to the same degree.

Robot faces elicit responses intermediate to human faces and objects at face-sensitive ERP components.

Face recognition is supported by selective neural mechanisms that are sensitive to various aspects of facial appearance. These include event-related potential (ERP) components like the P100 and the N170 which exhibit different patterns of selectivity for various aspects of facial appearance. Examining the boundary between faces and non-faces using these responses is one way to develop a more robust understanding of the representation of faces in extrastriate cortex and determine what critical properties an image must possess to be considered face-like. Robot faces are a particularly interesting stimulus class to examine because they can differ markedly from human faces in terms of shape, surface properties, and the configuration of facial features, but are also interpreted as social agents in a range of settings. In the current study, we thus chose to investigate how ERP responses to robot faces may differ from the response to human faces and non-face objects. In two experiments, we examined how the P100 and N170 responded to human faces, robot faces, and non-face objects (clocks). In Experiment 1, we found that robot faces elicit intermediate responses from face-sensitive components relative to non-face objects (clocks) and both real human faces and artificial human faces (computer-generated faces and dolls). These results suggest that while human-like inanimate faces (CG faces and dolls) are processed much like real faces, robot faces are dissimilar enough to human faces to be processed differently. In Experiment 2 we found that the face inversion effect was only partly evident in robot faces. We conclude that robot faces are an intermediate stimulus class that offers insight into the perceptual and cognitive factors that affect how social agents are identified and categorized.

Facial femininity and perceptions of eating disorders: A reverse-correlation study.

Eating disorders are prevalent in college students but college students are not accurate in identifying the presence of eating disorders (ED) especially when race is involved. Much has been researched about diagnostic ability in vignette form, but little outside of this. For example, it is not known how facial features, such as perceived femininity, may affect observers' beliefs about the likelihood of disordered eating depending on race. In the present study, we examined how biases regarding facial appearance and disordered eating may differ depending on the race of face images. Using a technique called reverse correlation, we estimated the image templates associated with perceived likelihood of disordered eating using both White and Black Faces. Specifically, we recruited 28 college students who categorized White and Black faces according to perceived likelihood of an eating disorder diagnosis in the presence of image noise. Subsequently, we asked Amazon Mechanical Turk participants to categorize the resulting race-specific face templates according to perceived ED likelihood and femininity. The templates corresponding to a high likelihood of an ED diagnosis were distinguished from low-likelihood images by this second independent participant sample at above-chance levels. For Black faces, the templates corresponding to a high likelihood of an ED diagnosis were also selected as more feminine than low-likelihood templates at an above-chance level, whereas there was no such effect found for White faces. These results suggest that stereotyped beliefs about both femininity and the likelihood of disordered eating may interact with perceptual processes.

Sensitivity to face animacy and inversion in childhood: Evidence from EEG data.

Adults exhibit relative behavioral difficulties in processing inanimate, artificial faces compared to real human faces, with implications for using artificial faces in research and designing artificial social agents. However, the developmental trajectory of inanimate face perception is unknown. To address this gap, we used electroencephalography to investigate inanimate faces processing in cross-sectional groups of 5-10-year-old children and adults. A face inversion manipulation was used to test whether face animacy processing relies on expert face processing strategies. Groups of 5-7-year-olds (N = 18), 8-10-year-olds (N = 18), and adults (N = 16) watched pictures of real or doll faces presented in an upright or inverted orientation. Analyses of event-related potentials revealed larger N170 amplitudes in response to doll faces, irrespective of age group or face orientation. Thus, the N170 is sensitive to face animacy by 5-7 years of age, but such sensitivity may not reflect high-level, expert face processing. Multivariate pattern analyses of the EEG signal additionally assessed whether animacy information could be reliably extracted during face processing. Face orientation, but not face animacy, could be reliably decoded from occipitotemporal channels in children and adults. Face animacy could be decoded from whole scalp channels in adults, but not children. Together, these results suggest that 5-10-year-old children exhibit some sensitivity to face animacy over occipitotemporal regions that is comparable to adults.

The Fork-and-Knife Illusion.

We describe a transparency illusion that can be observed with an ordinary metal knife and fork. Placed in the correct configuration relative to the fork, the metal knife appears transparent, with some observers experiencing a bistable percept in which transparency alternates with reflective appearance. The effect is related to other illusory percepts that follow from careful placement of mirrored surfaces, but to our knowledge, it is unique in that the key feature of the illusion is how the mirrored surface (in this case, the knife) is perceived rather than how a mirror induces altered perception of other objects and surfaces. We describe conditions that do and do not affect the strength of the illusion and point out its connections to previously reported phenomena.

Reorganization in the representation of face-race categories from 6 to 9 months of age: Behavioral and computational evidence.

Prior research has reported developmental change in how infants represent categories of other-race faces (Developmental Science 19 (2016) 362-371). In particular, Caucasian 6-month-olds were shown to represent African versus Asian face categories, whereas Caucasian 9 month-olds represented different classes of other-race faces in one category, inclusive of African and Asian faces but exclusive of Caucasian faces. The current investigation sought to provide stronger evidence that is convergent with these findings by asking whether infants will generalize looking-time responsiveness from one to another other-race category. In Experiment 1, an experimental group of Caucasian 6-month-olds was familiarized with African (or Asian) faces and then given a novel category preference test with an Asian (or African) face versus a Caucasian face, while a control group of Caucasian 6-month-olds viewed the test faces without prior familiarization. Infants in the experimental group divided attention between the test faces and infants in the control group did not manifest a spontaneous preference. Experiment 2 used the same procedure, but was conducted with Caucasian 9-month-olds. Infants in the experimental group displayed a robust preference for Caucasian faces when considered against the finding that infants in the control group displayed a spontaneous preference for other-race faces. The results offer confirmation that between 6 and 9 months, infants transition to representing own-race versus other-race face categories, with the latter inclusive of multiple other-race face classes with clear perceptual differences. Computational modeling of infant responding suggests that the developmental change is rooted in the statistics of experience with majority versus minority group faces.

Neural sensitivity to natural image statistics changes during middle childhood.

Natural images have properties that adults' behavioral and neural responses are sensitive to, but the development of this sensitivity is not clear. Behaviorally, children acquire adult-like sensitivity to natural image statistics during middle childhood (Ellemberg et al., 2012), but infants exhibit sensitivity to deviations of natural image structure (Balas & Woods, 2014). We used event-related potentials (ERPs) to examine sensitivity to natural image statistics during childhood at distinct processing stages (the P1 and N1 components). We presented children (5-10 years old) and adults with natural images varying in positive/negative contrast, and natural/synthetic texture appearance to compare electrophysiological responses to images that did or did not violate natural statistics. We hypothesized that children would acquire sensitivity to these deviations late in middle childhood. Instead, we observed significant responses to unnatural contrast and texture statistics at the N1 in all age groups. At the P1, however, only young children exhibited sensitivity to contrast polarity. The latter effect suggests greater sensitivity earlier in development to some violations of natural image statistics. We discuss these results in terms of changing patterns of invariant texture processing during middle childhood and ongoing refinement of the representations supporting natural image perception.

Children are sensitive to mutual information in intermediate-complexity face and non-face features.

Understanding developmental changes in children's use of specific visual information for recognizing object categories is essential for understanding how experience shapes recognition. Research on the development of face recognition has focused on children's use of low-level information (e.g. orientation sub-bands), or high-level information. In face categorization tasks, adults also exhibit sensitivity to intermediate complexity features that are diagnostic of the presence of a face. Do children also use intermediate complexity features for categorizing faces and objects, and, if so, how does their sensitivity to such features change during childhood? Intermediate-complexity features bridge the gap between low- and high-level processing: they have computational benefits for object detection and segmentation, and are known to drive neural responses in the ventral visual system. Here, we have investigated the developmental trajectory of children's sensitivity to diagnostic category information in intermediate-complexity features. We presented children (5-10 years old) and adults with image fragments of faces (Experiment 1) and cars (Experiment 2) varying in their mutual information, which quantifies a fragment's diagnosticity of a specific category. Our goal was to determine whether children were sensitive to the amount of mutual information in these fragments, and if their information usage is different from adults. We found that despite better overall categorization performance in adults, all children were sensitive to fragment diagnosticity in both categories, suggesting that intermediate representations of appearance are established early in childhood. Moreover, children's usage of mutual information was not limited to face fragments, suggesting the extracting intermediate-complexity features is a process that is not specific only to faces. We discuss the implications of our findings for developmental theories of face and object recognition.

Children's use of local and global visual features for material perception.

Adults can rapidly recognize material properties in natural images, and children's performance in material categorization tasks suggests that this ability develops slowly during childhood. In the current study, we further examined the information children use to recognize materials during development by asking how the use of local versus global visual features for material perception changes in middle childhood. We recruited adults and 5- to 10-year-old children for three experiments that required participants to distinguish between shape-matched images of real and artificial food. Accurate performance in this task requires participants to distinguish between a wide range of material properties characteristic of each category, thus testing material perception abilities broadly. In two tasks, we applied distinct methods of image scrambling (block scrambling and diffeomorphic scrambling) to parametrically disrupt global appearance while preserving features in small spatial neighborhoods. In the third task, we used image blurring to parametrically disrupt local feature visibility. Our key question was whether or not participant age affected performance differently when local versus global appearance was disrupted. We found that although image blur led to disproportionately poorer performance in young children, this effect was reduced or absent when diffeomorphic scrambling was used. We interpret this outcome as evidence that the ability to recruit large-scale visual features for material perception may develop slowly during middle childhood.

Neural Sensitivity to Mutual Information in Intermediate-Complexity Face Features Changes during Childhood.

One way in which face recognition develops during infancy and childhood is with regard to the visual information that contributes most to recognition judgments. Adult face recognition depends on critical features spanning a hierarchy of complexity, including low-level, intermediate, and high-level visual information. To date, the development of adult-like information biases for face recognition has focused on low-level features, which are computationally well-defined but low in complexity, and high-level features, which are high in complexity, but not defined precisely. To complement this existing literature, we examined the development of children's neural responses to intermediate-level face features characterized using mutual information. Specifically, we examined children's and adults' sensitivity to varying levels of category diagnosticity at the P100 and N170 components. We found that during middle childhood, sensitivity to mutual information shifts from early components to later ones, which may indicate a critical restructuring of face recognition mechanisms that takes place over several years. This approach provides a useful bridge between the study of low- and high-level visual features for face recognition and suggests many intriguing questions for further investigation.

Face perception: A brief journey through recent discoveries and current directions.

Faces are a rich source of information about the people around us. Identity, state of mind, emotions, intentions, age, gender, ethnic background, attractiveness and a host of other attributes about an individual can be gleaned from a face. When face perception fails, dramatic psycho-social consequences can follow at the individual level, as in the case of prosopagnosic parents who are unable to recognize their children at school pick-up. At the species level, social interaction patterns are shaped by human face perception abilities. The computational feat of recognizing faces and facial attributes, and the challenges overcome by the human brain to achieve this feat, have fascinated generations of vision researchers. In this paper, we present a brief overview of some of the milestones of discovery as well as outline a selected set of current directions and open questions on this topic.

Perceiving Animacy in Own-and Other-Species Faces.

Though artificial faces of various kinds are rapidly becoming more and more life-like due to advances in graphics technology (Suwajanakorn et al., 2015; Booth et al., 2017), observers can typically distinguish real faces from artificial faces. In general, face recognition is tuned to experience such that expert-level processing is most evident for faces that we encounter frequently in our visual world, but the extent to which face animacy perception is also tuned to in-group vs. out-group categories remains an open question. In the current study, we chose to examine how the perception of animacy in human faces and dog faces was affected by face inversion and the duration of face images presented to adult observers. We hypothesized that the impact of these manipulations may differ as a function of species category, indicating that face animacy perception is tuned for in-group faces. Briefly, we found evidence of such a differential impact, suggesting either that distinct mechanisms are used to evaluate the "life" in a face for in-group and out-group faces, or that the efficiency of a common mechanism varies substantially as a function of visual expertise.

The Flashed Face Distortion Effect Does Not Depend on Face-Specific Mechanisms.

When normal faces are rapidly presented in the visual periphery, they are perceived as grotesque and distorted. This phenomenon, "The flashed-face distortion effect" (FFDE) is a powerful illusion that may reveal important properties of how faces are coded in peripheral vision. Despite the strength of the illusion (and its popularity), there has been almost no follow-up work to examine what governs the strength of the illusion or to develop a clear account of its phenomenology. Presently, our goal was to address this by manipulating aspects of facial appearance and spatial/temporal properties of the flashed-face stimulus to determine what factors modulate the illusion's strength. In three experiments, we investigated the extent to which local contrast (operationalized by the presence or absence of makeup), image eccentricity, image size, face inversion, and presentation rate of images within the sequence each contributed to the strength of the FFDE. We found that some of these factors (eccentricity and presentation rate) mattered a great deal, while others (makeup, face inversion and image size) made little contribution to the strength of the FFDE. We discuss the implications of these results for a mechanistic account of the FFDE, and suggest several avenues for future research based on this compelling visual illusion.

The Effects of Blur and Inversion on the Recognition of Ambient Face Images.

When viewing unfamiliar faces that vary in expressions, angles, and image quality, observers make many recognition errors. Specifically, in unconstrained identity-sorting tasks, observers struggle to cope with variation across different images of the same person while succeeding at telling different people apart. The use of ambient face images in this simple card-sorting task reveals the magnitude of these face recognition errors and suggests a useful platform to reexamine the nature of face processing using naturalistic stimuli. In the present study, we chose to investigate the impact of two basic stimulus manipulations (image blur and face inversion) on identity sorting with ambient images. Although these manipulations are both known to affect face processing when well-controlled, frontally viewed face images are used, examining how they affect performance for ambient images is an important step toward linking the large body of research using controlled face images to more ecologically valid viewing conditions. Briefly, we observed a high cost of image blur regardless of blur magnitude, and a strong inversion effect that affected observers' sensitivity to extrapersonal variability but did not affect the number of unique identities they estimated were present in the set of images presented to them.

Neural sensitivity to natural texture statistics in infancy.

During infancy, vision becomes tuned to environmental statistics. For example, infant face recognition "narrows" in response to the frequency of face categories in the visual world, inducing out-group effects that disadvantage other-race, other-species, and other-age face recognition. There are many other low-level statistical regularities in visual experience that infants may also become tuned to during this period. In particular, natural scenes have lawful properties that adults and children are sensitive to. To what extent do infants become tuned to these regularities during the first year of life? In particular, do infants exhibit evidence of perceptual narrowing that excludes atypical images from fluent processing? We examined this question by measuring 6- and 9-month-old infants' event-related potentials (ERPs) to natural and artificial textures created by: (a) Disrupting local statistics via contrast negation, (b) Disrupting global statistics via parametric texture synthesis, or (c) both of these. We predicted that younger infants' would be sensitive to both manipulations of natural appearance, but that older infants might not distinguish between different kinds of atypical images. Instead, we found that sensitivity to synthetic appearance is only evident late in infancy. We discuss what these results imply for our understanding of visual statistical learning in infancy.

Low-level orientation information for social evaluation in face images.

Observers make a range of social evaluations based on facial appearance, including judgments of trustworthiness, warmth, competence, and other aspects of personality. What visual information do people use to make these judgments? While links have been made between perceived social characteristics and other high-level properties of facial appearance (e.g., attractiveness, masculinity), there has been comparatively little effort to link social evaluations to low-level visual features, like spatial frequency and orientation sub-bands, known to be critically important for face processing. We explored the extent to which different social evaluations depended critically on horizontal orientation energy vs. vertical orientation energy, as is the case for face identification and emotion recognition. We found that while trustworthiness judgments exhibited this bias for horizontal orientations, competence and dominance did not, suggesting that social evaluations may depend on a multi-channel representation of facial appearance at early stages of visual processing.

Children's use of visual summary statistics for material categorization.

Although adults' ability to recognize materials from complex natural images has been well characterized, we still know very little about the development of material perception. When do children exhibit adult-like abilities to categorize materials? What visual features do they use to do so as a function of age and material category? In the present study, we attempted to address both of these issues in two experiments that we administered to school-age children (5-10 years old) and adults. In both tasks, we asked our participants to categorize natural materials (metal, stone, water, and wood) using original images of these materials as well as synthetic images made with the Portilla-Simoncelli algorithm. By including synthetic images in our stimulus set, we were able to assess both how material categorization develops during childhood and how visual summary statistics are recruited for material perception across age groups. We observed that when asked to provide category labels for individual images (Experiment 1), young children were disproportionately bad at categorizing some materials after they were synthesized, suggesting material-specific changes in information use over the course of development. However, when asked to match real and synthetic images according to material category without labeling (Experiment 2), these effects were weakened. We conclude that while children have adult-like abilities to encode and compare images based on summary statistics, the mapping between summary statistics and category labels undergoes prolonged development during childhood.