Recognition of individuals at first sight is important for social species and can be achieved by attending to facial or body information. Previous research suggests that infants possess a perceptual template for evolutionarily relevant stimuli, which may include humans, dangerous animals (e.g. snakes), but not non-dangerous animals. To be effective, such a mechanism should result in a systematic preference for attending to humans over non-dangerous animals. Using a preferential looking paradigm, the present studies investigated the nature of infants' early representation of humans. We show that 3.5- and six-month-old infants attend more to human beings than non-human primates (a gorilla or monkey) which are examplars of non-dangerous animals. This occurred when infants were presented with head or body information in isolation, as well as when both are presented simultaneously. This early preference for humans by 3.5 months of age suggests that there is a basic representation for humans, which includes both head and/or body information. However, neonates demonstrated a preference only for human faces over non-human primate faces, not for humans over non-human primates when the stimuli were presented with both head and body simultaneously. The results show that although neonates display a preference for human faces over others, preference for the human body only develops later, in the first few months of life. This suggests that infants have acquired some knowledge about the human body at 3.5 months of age that may have developed from their privileged experience with other humans in the first few months of life, rather than an innate ability to detect humans in their entirety.
To survive in life threatening and in social situations, individuals of any given species need to detect dangerous species, as well as conspecifics, and to discriminate in-group and out-group members. In other words, individuals need to categorize stimuli, the propensity to group items into distinct morphological sets, and to recognize them within a category. Recognition of individuals at first sight is especially important for social species, and may have been pivotal for the development of primate societies with strong social relationships . Some species recognize individuals using their olfactory (e.g. hamster; ) or auditory (e.g. birds; ) capabilities. Primates individuate using their visual system. Recognition can be based on several visual elements of an individual, such as body shape or gait, but examination of the face leads to the fastest and most accurate identification . Before individual recognition occurs, categorization occurs at a fast rate and may even facilitate it . Indeed several studies have shown that categorization occurs faster than recognition in human adults, suggesting that it constitutes a separate process . The rapidity of the process is likely to permit fast reaction to threatening stimuli like predators. Moreover, such categorization may facilitate approaches of non-threatening stimuli like conspecifics and further individuation. Whereas we possess some understanding about the development of face processing supporting individual recognition, we know less about the development of whole body categorization. How and when do human infants generate a representation of what a human body looks like? In this paper, we are investigating if the representation that human infants have of humans includes the whole body or not. Considering the limitation of the visual system during the first few months of life and the nature of the infant–carer interaction, which is usually face to face, it is sensible to suggest that the whole human body representation emerges after the face representation. Alternatively, it is possible that, given its adaptive value, whole body representation is present from birth and is a hard-wired ability passed through evolution.
In this paper, we will review research on adults' and infants' representation of their own and other species. We will also compare infant abilities for human face processing to those for whole body processing. Finally, we will present new experimental evidence of an early representation of the human face and whole body during infancy.
2. Representation of humans and other species
Some stimuli in our environment require more attention than other stimuli as our survival may depend on our ability to detect them. For example, harmful stimuli such as snakes and spiders may require high levels of attention, to avoid negative consequences such as being bitten and poisoned. More generally, monitoring non-human animals was of great importance to our ancestors because these animals could potentially be either predators or food, or they could be fellow humans of interest for social purposes. Thus, given the adaptive value of categorizing other living animals, including humans, we can hypothesize that humans, over evolution, developed a system that affords privileged processing of living things.
Animals change their status far more frequently than other objects in our environment (e.g. plants, artefacts), and thus require more frequent and constant monitoring . The animate monitoring hypothesis proposes that the human attention system evolved to differentially monitor animals/humans versus other objects: animals and humans should recruit more spontaneous attention. To test this hypothesis, New et al.  used a change-detection paradigm, in which adults were exposed to alternations between complex natural scenes and duplicates with a single change. Adults were substantially faster and more accurate at detecting changes in animals and people relative to changes in other familiar objects (e.g. plants, buildings), including vehicles, which adults have been trained for years to monitor to avoid collision and death. In some experiments, there was a further advantage for human detection over other animals, such that accuracy was better (i.e. more changes were detected) when the target was a person than an animal. Kirchner & Thorpe  have demonstrated that human participants can detect animal presence in pictures even if they were presented extremely rapidly. Crouzet et al.  have extended this work, demonstrating that participants can elicit a saccade towards a human face in 100 ms! Together, these studies suggest that the human attention system evolved to differentially monitor animals/humans, which are highly relevant objects in our environment and potentially essential for our survival.
In terms of harmful or negative stimuli, Seligman's  theory of prepared learning proposes that as adults we most readily learn to fear classes of threats, such as snakes, that were recurrent throughout evolutionary history . Thus, it appears that some stimuli (both positive and negative) are more relevant to survival of a species than others, and adults are tuned to attend to those stimuli accordingly. In the context of human development, therefore, it is of interest to investigate when these stimuli are first perceived as highly relevant or attention-worthy during development.
From birth, infants face the difficult task of extracting relevant information from a complex world and deciding what is essential to learn in order to survive. One hypothesis about how infants categorize and process stimuli is that mechanisms which have evolved from our ancestral history will prime our attention towards specific categories of stimuli and will be present early in development. Evidence supporting this hypothesis comes from the finding that infants attend to evolutionarily relevant stimuli that are dangerous. Five month old infants look longer at a moving image of a spider than at reconfigured or scrambled versions of the same image. This pattern of responding does not occur when infants are presented with a neutral stimulus (a flower) in an analogous task . Similarly, 7–18 month olds more readily associate a fearful voice with a movie of a snake than other moving objects . Infants were simultaneously presented with two films—one of a snake and the other of an exotic animal—accompanied by a recording of either a very frightened or very happy human voice. Infants looked longer at films of snakes while listening to a frightened human voice than while listening to a happy voice. These studies suggest that human infants are primed towards evolutionarily relevant stimuli, in this case towards detecting specific animals that were potentially dangerous throughout evolutionary history. This allows for rapid identification and facilitates learning about harmful objects early in life.
Not all highly relevant evolutionary stimuli are negative. Humans are essential for the infant's survival as well as for social interaction and comforting, making them a positive evolutionary stimulus. Detecting and orienting rapidly to humans is therefore crucial for infants. Infants attend to and have knowledge about humans from birth. Newborns prefer to look at typical faces rather than faces where the parts have been scrambled [13,14]. Infants prefer to look at biological motion than non-biological motion during the first week of life . The preference for humans also extends to the auditory domain: newborns prefer human speech than non-speech analogues (which contain similar spectral and temporal parameters as speech) . Together, the above research strongly suggests that infants are born with mechanisms which draw their attention towards evolutionarily relevant stimuli (humans) and predispose them to learn about conspecifics.
How specific is the mechanism which recruits attention to humans? Is it species-specific, or a broad representation which also includes non-human animals? How early do infants recognize or treat their own species differently?
A series of studies by Quinn et al. (e.g. ) demonstrates that three-month-olds represent humans and non-human animals (cats and horses) differently. By three months of age infants can categorize non-human animals on the basis of perceptual features. They can form a categorical representation for domestic cats that excludes birds, horses, dogs, tigers and female lions, this categorization being primarily based on face/head information [18–20]. This has been demonstrated using the habituation–dishabituation procedure, which involves presenting infants with exemplars from one category (e.g. dogs) until they reach habituation criterion. Following this, two exemplars are presented: a novel exemplar from the already familiarized category (e.g. a new dog) and a novel exemplar from a novel category (e.g. a cat). If infants can differentiate the two categories they should look longer at the exemplar from the novel category. Using this procedure, it has been shown that there is an asymmetry in three month olds' categorization such that when habituated to horses or cats infants will dishabituate to a human (i.e. they differentiate the two categories). However, when familiarized with humans, infants do not dishabituate to a cat or horse, and instead include cats and horses in their category for humans . This indicates that infants' representation for humans is quite broad, based on many exemplars, and thus can include other animals. By contrast, the representation for other animals is much narrower, based on a prototype; thus infants do not include humans in their representation for other animals .
Interestingly, this asymmetry in categorization is only observed when infants are presented with whole animal stimuli, not when provided with information from just the head or the body alone of the exemplars . This suggests that young infants' representation for humans is based on the overall structure of the stimuli (i.e. a head on top of an elongated body with appendages) and contrasts with the finding that representations for non-human animals appear to be based on part/featural information such as heads [18,19]. This leads to the question of how human faces and bodies are processed when presented separately.
3. Body processing
Recent brain imaging suggests that human bodies are a special category of objects for adults, and that we have specific brain areas dedicated for processing human bodies . In particular, the extrastriate body area (EBA) in the lateral occipitotemporal cortex [23,24] was shown to respond selectively to images of the human body. Furthermore, the strength of visual representations in the EBA is determined by long-term visual experience, such that representations are strongest for stimuli in their usual combinations of visual field and side . Another cortical area that is active when bodies are being processed is the fusiform body area (FBA), located in the lateral posterior fusiform gyrus. These two cortical areas appear to have different functions within the domain of processing bodies. The EBA seems to specialize in processing the human body at the part level, whereas the FBA appears to specialize in processing the body in terms of its configuration as a whole [24,26].
Event-related potential (ERP) studies also suggest a distinct neural representation of the human body. Images of the human body (without a head) elicited a negative ERP peaking at 190 ms (the N190). While similar to the N170 for faces, the N190 differs in both spatial distribution and amplitude, and peaks later than the N170. Furthermore, scrambled bodies do not elicit the N190 . Taken together, these studies suggest that there are dedicated neural mechanisms that process the human body in adulthood.
An indicator of expertise which can be measured behaviourally involves assessing whether objects are processed holistically (i.e. viewed as a whole instead of various parts) or configurally (i.e. processing spatial relations between an object's parts). Configural processing indicates expertise and can be assessed by the ability to detect and recognize an object once it has been inverted. The inversion effect indicates a disturbance in the recognition and processing of objects which are processed configurally, since spatial relations cannot be processed as easily once objects are presented upside down. Research indicates that human bodies are subject to an inversion effect, such that detection and recognition of postures and normal human body configurations are disrupted if the bodies are inverted, suggesting configural processing [28,29]. This is in contrast with recognition of different houses, which are recognized equally well in upright and inverted orientations. These studies suggest that adults most easily detect and recognize human bodies by analysing their configurations, and that human bodies are a special class of objects for adults owing to expertise and experience.
Infants' knowledge about the human body varies widely depending on the information available to the infants as well the nature of the task. The earliest demonstration of knowledge related to humans is not species specific. Infants demonstrate knowledge of biological motion from birth . Newborns selectively prefer to attend to biological motion (hen-walking animations) and the preference is orientation specific: newborns looked longer at upright displays than upside-down displays of biological motion. This study suggests that detection of biological motion is an intrinsic capacity of the visual system, which is most probably part of an evolutionarily relevant system that applies to many species and predisposes animals to attend to other animals.
The earliest evidence of specific knowledge about the human body shape is demonstrated at four to six months of age, when the human body model is real and moving naturally . In this study infants were habituated to a series of different normal human body postures and then presented with scrambled human bodies (e.g. arms lowered to hips) on the test trial. The human model was a real live human moving naturally, who could be manipulated to look both normal and scrambled with the aid of a curtain and a second experimenter. Infants dishabituated to the scrambled bodies, indicating recognition of typical and scrambled bodies as distinct categories. However, movement is crucial to early detection of violations of the configuration of the body shape. When the same habituation–dishabituation procedure was used with the one key difference that the real human bodies remained static, the earliest response to scrambled static human bodies was at nine months of age . In line with this, Zieber et al.  report that nine month olds are sensitive to the relative proportions of human body parts, while five month olds are not. Infants were presented with pairs of photos depicting a normal versus a proportionally distorted body (created by lengthening the neck and torso and shortening the legs). Nine month olds exhibited a preference for the normal body when images were presented upright but not when they were inverted, controlling for the possibility that infants were simply responding to some low-level feature. Taken together, these studies show emergence of knowledge about the configuration of the human body shape in its static form at nine months of age.
There is some evidence that sensitivity to the human body shape may develop before infants demonstrate this ability in visual tasks. Gliga & Dehaene-Lambertz  report different ERP recordings for typical and scrambled bodies in three-month-old infants. In this study, infants were presented with images of human bodies where head information had been removed. The mean amplitudes for the scrambled bodies were significantly greater than for the typical body images. This may indicate discrimination of typical and scrambled bodies or it may be that the infants were responding simply to the low-level configuration information in the stimuli (e.g. overall symmetry) without recognizing the images as human bodies . More research is required before firm conclusions can be drawn.
4. Face processing
The core of the human neural system for face processing as revealed by functional MRI studies is of three cortical areas: the inferior occipital gyrus (occipital face area; OFA) that allows the creation of a global stimulus based on the parts of a face; the middle fusiform gyrus (fusiform face area; FFA) supporting recognition/discrimination of individuals and the superior temporal sulcus that carries information relative to the direction of gaze, orientation and features of the faces . These three areas are more activated in humans viewing human faces than other visual objects [34–36] with the lateral FFA showing the strongest activation in the right hemisphere .
At the physiological level, there is evidence suggesting that we process human and other species' faces differently. Face selective electrophysiological activity has been observed in ERP (recorded from the scalp) studies with adults, and consists in a negative deflection, with peak latency around 170 ms after stimulus onset (N170). This potential tends to be of larger amplitude and shorter latency for faces than other objects . It is influenced by stimulus inversion as the N170 is of larger amplitude and longer latency for inverted human faces compared with upright human faces. This inversion effect is particular to human face stimuli and has not been observed for animal faces . At a behavioural level, Dufour et al.  showed that humans recognized human faces better than monkey faces in a two alternatives forced choice task. Likewise, studies conducted with a visual paired comparison task where no instruction of recognition is given, showed that humans discriminate automatically between two human faces but not between two macaque faces .
Taken together, these elements provide solid evidence that the adult face processing system is somewhat species specific and lacks the flexibility to allow recognition of faces of other species at an individual level. How does such a highly specialized face processing system in adults develop? What is the impact of experience on its development?
From the first moments of life, newborn infants prefer to look at human faces over almost any other form of stimuli ([13,14,42,43], suggesting the existence of a specialized cognitive system operating from very early in life. Johnson & Morton  have proposed that this natural orientation and attentiveness towards faces may be driven by CONSPEC—a subcortical system containing very basic information regarding the visual structural characteristics of members of one's own species. These structural characteristics probably include a bounded area, an asymmetrical featural pattern with more elements on the upper portion of the bounded area, and a positive stimulus contrast . On the other hand, CONLERN refers to a cortical system which accrues and retains fine details regarding the visual characteristics of conspecifics via experience with such conspecifics . Thus, one perspective holds that an initial biological predisposition to attend to faces is subsequently complemented by visual experience with conspecifics to develop face expertise. However, it should also be noted that the existing evidence regarding infants' preference for the first-order configuration of schematic faces does not directly implicate a preference for own-species configuration. Thus, if neural mechanisms, such as CONSPEC and CONLERN do exist, they might not necessarily be species specific. In addition to face preference, human infants also demonstrate evidence of categorization and recognition .
Nelson  hypothesized that in humans the representation of faces at birth is broad and develops according to the type of facial input received, tuning towards the predominant faces in the environment. In term of categorization, Quinn et al.  found that face representation in three-month-olds is biased towards their primary carer. Infants primarily raised by their mother prefer to look at female faces when paired with male faces, whereas a population of three month olds who had been primarily raised by their fathers show a preference for male faces. Kelly et al.  further highlighted the role of experience in early infancy, demonstrating that three-month-old Caucasian infants prefer to look at faces from their own racial group when paired with faces from other racial groups in a visual preference task. Caucasian newborns tested in an identical manner demonstrated no preference for faces from either their own or other racial groups.
Growing evidence suggests that greater experience with a particular face type leads to improved face processing abilities (e.g. better recognition abilities), whereas a lack of experience with a particular face type leads to relatively poorer face processing abilities (e.g. poor recognition abilities). Pascalis et al.  investigated the ability of six- and nine-month-old infants to recognize faces from their own species (human) and those from other species (rhesus macaque) using a standard infant recognition paradigm. As expected, infants at both ages were able to demonstrate recognition with human faces. However, when tested with the monkey faces, only the six-month-old group showed evidence of recognition, suggesting that the face system becomes tuned to human faces between six and nine months of age. This perceptual narrowing process is akin to a similar phenomenon in the language domain whereby younger infants are able to discriminate almost all phonemes in any language in the world but later become particularly sensitive to phonemes of the language to which they receive the most exposure in their environment [51,52].
It is possible to maintain accurate recognition memory for monkey faces if six-month-old infants are exposed regularly to other species' faces . However, the exposure should be associated with consistent individuation between such faces (i.e. repeatedly referring to each monkey face by a specific name; ). Such species-specific perceptual narrowing is also evident at the intersensory level by eight months of age. Four to six-month-old infants demonstrated an ability to match a visual image of a monkey producing a given vocalization with its congruent auditory call, while eight-month-olds did not .
This perceptual narrowing in face perception at the broader species level has also been found at the within species level—that is, pertaining to different human races. Kelly et al.  showed that Caucasian infants at three months of age show recognition memory for individual faces within Caucasian, Chinese, Middle Eastern and African races, while nine-month-olds only showed recognition memory for own-race Caucasian faces. In addition, Kelly et al.  showed that Chinese infants undergo a similar course of perceptual narrowing in that increased exposure to Chinese faces leads to a recognition memory only for own-race Chinese faces at nine months of age.
Only a few studies have investigated the electrophysiological response elicited by own versus other species faces during infancy. De Haan et al.  have found an ‘infant N170’ in six-month-olds that was elicited by faces at a latency of 290 ms followed by a positivity at 400 ms. They also examined the influence of stimulus inversion, for both human and monkey faces and found that in adults, inversion affected only the processing of human faces and not monkey faces, while in six-month-olds, inversion affected the ERPs similarly for human and monkey faces. Around 12 months of age, the adult ERP patterns are observed . These results suggest that humans possess an evolved system for processing faces that becomes specialized as a consequence of exposure exclusively to faces from a single species.
This review of the literature, though not exhaustive, shows that at birth, human infants most probably benefit from a perceptual system that allows them to process a variety of ecologically relevant stimuli selected over evolution. Longitudinal examination of this system highlights its adaptive and plastic nature enabling recognition to be best tuned to the precise stimuli infants encounter in their environment. These plastic properties are observed for different species, body regions (faces versus the rest of the body) and for face stimuli this plasticity applies at different superordinate levels (races, gender). In the following sections of this paper, we will present experiments conducted in our laboratory aimed at understanding the nature of infants' representation of humans between 3 days and nine months of age. Previous literature suggests that infants possess a perceptual template for evolutionarily relevant stimuli, which may include humans, dangerous animals (e.g. snakes), but not non-dangerous animals. Such a mechanism should result in a systematic preference for humans over non-dangerous animals. However, other primates share the same general body shape and face arrangement as humans; thus it is important to determine if any preferences for humans are human or primate specific.
This series of studies investigated whether infants could discriminate humans from upright non-human primates, and if there is a preference for humans from birth and across the first six months of life (experiment 1). Next, we investigated whether infants could discriminate humans from non-human primates using only face information (experiment 2) or body information only (experiment 3). Together, these experiments take the first step to investigate infants' early representation of human beings in an inter-species framework.
5. Experiment 1: head and body information available
This experiment investigated whether or not there is a preference to attend to humans over non-human primates at birth, three months and six months of age.
Participants were recruited from the Royal Hallamshire Hospital, Sheffield, UK. In total, 14 full-term neonates (eight females; age range 9–102 h old, M = 60 h), fourteen 3.5-month-old infants (seven females; age range 105–114 days, M = 110 days) and 16 six-month-old infants (eight females; age range 180–190 days, M = 184 days) were included in the final sample. A further six neonates were excluded from the final sample due to fussiness (N = 3) or falling asleep (N = 3), and seven 3.5-month-olds were excluded due to side bias (n = 6) or fussiness (n = 1). Side bias was defined as looking at one image in the pair for 95 per cent or more of the total looking time.
The stimuli were four colour photographs of humans and non-human primates (figure 1). Pairs of one male adult and one non-human primate were shown to infants (two pairs in total). Non-human primates included a monkey and a gorilla. Humans were presented wearing long-sleeved jackets and trousers so that minimal skin was exposed. The human's clothing was coloured to match the fur of the non-human primate with which it was paired. The human male had closely shaved hair on his head. Photographs were presented against a white background. In both pairs the human and non-human primates were presented standing and matched on height.
(b) Testing procedure
Newborns were tested in a quiet room, seated in a semi-upright position in a padded infant car chair, which was secured to a table, limiting movement and ensuring safety, approximately 40 cm from a screen (measuring 30 × 45 cm) onto which the paired images were projected. Infants of 3.5- and six-months-old were tested in an anechoic chamber at the University of Sheffield, UK. Infants were seated on their mother's lap approximately 60 cm away from a screen, which displayed the images. For all age groups the experimenter remained out of sight during testing, and both the mother and the experimenter remained quiet. The two pairs of photographs were presented until 10 s of cumulative looking had been obtained for each pair. When projected onto the screen all images measured 18 × 18 cm (16° visual angle) and were positioned side by side separated with a 9 cm gap. If the infant spent 10 s looking away from the projected images, the trial was terminated. Between the two image pairings, a blank screen was presented for 3–5 s. Images were counterbalanced for side.
A black and white CCD camera (specialized for low light conditions) was used to film the infant's eye movements. This was displayed to the experimenter, during recording, on an ITC control monitor. Time was recorded and displayed on the control monitor using a Horita II (TG-50) at 25 frames per second. The film was subsequently digitized to be analysed frame by frame on a computer using specialized software. An independent observer recoded 25 per cent of the data for reliability. Both observers were blind to condition. The average level of inter-observer agreement was high (Pearson r = 0.98 for 3.5- and six-month-olds; r = 0.85 for neonates).
Preliminary examination of the data revealed no significant gender differences, so the data were combined for further analysis. A paired-samples two-tailed t-test conducted on the total time spent looking at the humans versus non-human primate stimuli revealed that overall 3.5- and six-month-old infants attended more to the humans than the non-human primates, t(13) = 3.34, p = 0.007 and t(15) = 6.62, p < 0.001 (table 1). In contrast, a paired-samples t-test revealed that neonates did not look significantly longer at the humans when compared with the non-human primates, t(13) < 1, p = 0.530 (table 1).
A 2 × 3 (stimuli: human versus animal × age: neonate, 3.5-, six-month-old) mixed model ANOVA was computed on total looking time to the stimuli. There was a significant main effect of stimuli F1,41 = 41.68, p < 0.001, η2 = 0.49; however, this was subsumed by a significant stimuli × age interaction, F2,41 = 5.42, p = 0.008, η2 = 0.20. Follow-up independent sample t-tests with a Bonferroni correction, adjusted for 3 comparison (α = 0.017) were conducted. t-tests were computed on difference scores, which were calculated by subtracting the infant's looking time to the non-human primate from their looking time to the human (thus positive scores reflect greater looking to the human stimuli). Independent group t-tests revealed that the difference scores for the 3.5- and six-month-old infants were significantly different from those of the neonates, t(26) = 2.65, p = 0.014 and t(28) = 3.58, p = 0.001, respectively. This indicates that while the 3.5- and six-month-old infants had large difference scores due to preferentially attending to the humans, the neonates did not. There was no significant difference in the difference scores for the 3.5- and six-month-olds, t(28) < 1, p = 0.641. This is because both age groups had large difference scores indicating greater looking to the human than non-human primate stimuli.
The results of this experiment confirm that 3.5- and six-month-old infants preferentially attend to humans over other primates when the stimuli are presented with head and body information. This ability is not demonstrated at birth in the present study, at least for humans in their entirety. This suggests that infants' preference for humans at 3.5 months is at least partly due to their rich experience with humans over the first few months of life, relative to the very minimal experience they have with non-human primates. The null preference for the newborn infants is not surprising as newborns primarily see faces, with exposure to only a select few other body parts (hands, breasts). Given that a newborn's vision is blurred beyond 40 cm and that their vertical field is approximately 120°, it allows a clear image of only 85 cm height, which is smaller than an adult human body. This means that newborn infants have very limited viewing of human bodies in their entirety and this condition likely accounts for the null preference observed. This finding is consistent with research involving older infants, which indicates that knowledge of human faces and bodies follows different developmental trajectories, with the former evident from birth and the latter not emerging until later in the first year . Detailed body knowledge is not as important as facial information, at least initially, from an evolutionary perspective. Facial information is more useful in this respect because it enables us more easily to recognize individuals, identify gender and infer emotional information such as mood states.
Therefore, some body parts may be more important than others for discriminating humans versus non-human primates, at least early in development. One might hypothesize that facial information is integral, given that newborn infants demonstrate knowledge about faces . In the current study, the head region of the pictures represented only a small proportion of the body, making it difficult for neonates to use facial information to make the discrimination. Experiments 2 and 3 test whether head or body information is necessary for newborn, 3.5- and six-month-old infants to discriminate humans versus non-human primates.
6. Experiment 2: head information only
This experiment investigated whether or not there is a preference to attend to humans over non-human primates in newborns and at 3.5 and six months of age when only head information is available. Infants demonstrate knowledge about faces from birth ; however, knowledge about the configuration of the human body shape does not emerge until nine months of age . Thus, it may be that infants rely on head information to discriminate human from non-human primates, and that a preference for human faces is responsible for the results obtained in experiment 1.
Eighteen healthy, full-term newborn babies (11 boys and seven girls, mean age 2.64 days, s.d. = 1.3) were selected from the maternity ward of the Jessop Wing of the Royal Hallamshire Hospital, Sheffield, UK and tested. A further 37 babies were selected but removed from the study for the following reasons: four babies failed to complete testing, 11 babies had a strong side bias, 12 babies changed their state during testing such that they were no longer alert, and 10 babies because of an error on the part of the experimenter. Infants of 3.5 and six months old were recruited in an identical manner to experiment 1. In total, 15 full-term 3.5-month-old infants (six females; age range 104–114 days, M = 108 days) and 16 six-month-old infants (eight females; age range 180–190 days, M = 185 days) were included in the final sample. A further four 3.5-month-olds were excluded due to side bias (n = 2) or fussiness (n = 2).
Neonates were presented with two pairs of stimuli composed of a full face (i.e. external as well as internal features were displayed), black and white photograph of an adult male or a female, depicted from the crown of the head to the jaw, and a full face, black and white photograph of a monkey's face, depicted from the crown of the head to the jaw (figure 2). For 3.5- and six-month-olds the stimuli were photographs depicting the four faces of the human and non-human primate stimuli in experiment 1 (figure 1). Two pairs of photographs (each containing a human and non-primate human face) were presented to infants. The human/gorilla faces were 14 × 10 cm and the human/monkey faces were 10.7 × 8.5 cm.
Infants were tested in a quiet room, seated in a semi-upright position in a padded infant car chair, which was secured to a table, limiting movement and ensuring safety, approximately 30 cm from a screen onto which the paired images were projected. Presentation of the images on the left- and right-hand side was counterbalanced. All aspects of stimuli presentation were identical to study 1. The average level of inter-observer agreement was high: Pearson r = 0.87 for neonates and r = 0.98 for 3.5- and six-month-olds.
A paired-samples two-tailed t-test conducted on total looking time revealed that overall neonates, 3.5- and six-month-old infants attended more to the human faces than the non-human primate faces, t(17) = 2.24, p = 0.03, t(14) = 4.33, p = 0.001 and t(15) = 4.93, p < 0.001, respectively (table 1).
The results of this experiment confirm that infants preferentially attend to humans over other primates when only face information is available. Therefore, body information is not necessary for infants to discriminate between the two species. Further, it appears that the presence of body information actually eliminates the preference for human faces in newborns.
The neonatal preference for human faces indicates that infants have learned something about human faces during the first few days of life. However, the stimuli used in this study do not allow us to determine if the preference is based on a difference in contrast: the human eyes have more contrast between the sclera and the iris than the monkey eyes. This finding needs to be interpreted cautiously until further studies can rule out low-level perceptual features as an explanation for the present results. However, if infants are indeed making the discrimination based on facial information, this is consistent with previous research showing that infants have a mechanism which draws them towards human faces or allows newborn infants to encode something about the appearance of the human face and subsequently prefer images which most closely match this template . A neonatal preference for human faces is also consistent with previous research showing knowledge of human faces develops before knowledge about the human body shape .
7. Experiment 3: body information only
This experiment investigated whether or not there is a preference to attend to humans over non-human primates at 3.5 and six months of age when only body information is available. The greater amount of time spent looking at humans when the whole body is present (experiment 1) does not rule out the possibility that the presence of the face drives the preference. It is possible that infants' early representation for humans is restricted to a limited number of features and does not take into account the whole body. This would be consistent with research showing that infants do not have detailed knowledge about human bodies until at least nine months of age .
We did not test neonates in this condition as no preference was observed when neonates were presented with head and body information simultaneously.
Participants were recruited in an identical manner to experiment 1. In total, 14 full-term 3.5-month-old infants (six females; age range 104–114 days, M = 107 days) and 16 six-month-old infants (eight females; age range 180–190 days, M = 184 days) were included in the final sample. A further seven 3.5-month-olds were excluded due to side bias (n = 5) or fussiness (n = 2).
The stimuli were the same four colour photographs of humans and non-human primates used for experiment 1, the key difference being that the heads were occluded using a grey strip so that only body information was available (figure 1).
Infants were tested in an identical manner to the 3.5- and six-month-olds in experiment 1. The average level of inter-observer agreement was high (Pearson r = 0.97).
A paired-samples two-tailed t-test conducted on total looking time revealed that overall 3.5- and six-month-old infants attended more to the human bodies than the non-human primate bodies, t(13) = 2.96, p = 0.010 and t(15) = 5.27, p < 0.001 (table 1).
The results of this experiment confirm that infants preferentially attend to human bodies over non-human primates' bodies even when face information is unavailable. Therefore, infants' demonstrate basic knowledge about the human body shape from 3.5 months of age, at least in terms of preferring to attend to it over other non-human primate bodies. This suggests that development of knowledge about the human body shape is gradual, with detailed knowledge about the configuration of the body shape developing later in the first year of life . Thus, while young infants may not have a detailed representation of the human body (i.e. know where parts go) they still recognize a human body when compared with a non-human primate.
8. General discussion of experiments 1–3
This series of experiments shows that at 3.5 months of age, infants attend more to human beings than non-human primates (a gorilla or monkey) when presented with head or body information in isolation, as well as when both are presented simultaneously. The preference for humans observed by 3.5 months of age suggests that infants already have a representation for humans which is flexible and accurate enough to allow recognition of humans even with the head information missing. This flexibility appears between birth and three months, as neonates show a preference only for the human head.
The fact that neonates are only tuned to faces and not human bodies might have two main reasons. First, face to face interaction is a privileged way of communication with the carers, thus prioritizing face rather than body processing. Second, neonates have very little opportunity to view a human body in its entirety because of limited vision at birth, allowing infants to see clearly an image only 85 cm in height from a 40 cm distance. This does not match a full adult height, and once newborn infants are more than 40 cm away from a person the image will become blurred. The infant could use eye movements to expand their visual field; however, given that newborns spend the vast majority of their time in the supine position, and have limited body and neck mobility, it is unlikely that the newborn infant would have experience of viewing human bodies in their entirety. These limitations fade as infants get older, allowing a full and flexible representation of the body to emerge. The exact timeline of this emergence remains unknown as there was no longitudinal day to day evaluation of this skill, but we can hypothesize that the representation builds gradually as infants have more and more interactions with other humans.
While a general body representation appears to be learned in the first three months, neonates already demonstrate a preference for human faces over non-human primate faces. This is consistent with previous research showing that from birth, infants encode something about the appearance of the human face and subsequently prefer images which most closely match this template . The early face processing system will allow infants to recognize others and possibly to convey and interpret emotions , which is essential for bonding and attachment with their carers. The exact mechanisms supporting this face preference remain unclear and could include both fast learning based on early interactions in the post-partum period as well as innate system selected by evolution . It has been suggested that infants have a broad face processing system that narrows with experience and support the hypothesis that perceptual systems tune to stimuli present in the environment [47,60]. This is supported by the fact that infants show individual recognition abilities for human and non-human primates until six months of age, a faculty which can be maintained until nine months with longer exposure to the other species [50,53,54]. This is not in contradiction with our current results, as within species face recognition is one level down from discrimination at the species level. In other words, discriminating between two samples from two species is less difficult than discriminating between two samples within the same species. Our finding denotes an adaptive behaviour that is consistent with studies demonstrating knowledge of human faces and biological motion in the first few days of life [13,61]. Our results are also in line with the theoretical framework of the human first hypothesis  which proposes that infants possess information about their conspecifics and use it to identify and count objects.
Other species present a very similar adaptative system from birth . In adult rhesus monkeys, faces processing is also a species specific system (see  for a review). However, this species preference can be reversed in favour of humans if monkeys first experience human faces .
One intriguing finding of our experiment is that although neonates discriminated humans versus non-human primates based on face information alone, this behaviour disappeared when neonates were presented with whole body information. Rather than reflecting an absence of discrimination of humans per se, this lack of preference is most probably owing to the small size of the face when presented with the body and/or to body information being distracting, both preventing newborns from extracting sufficient facial information to support discrimination. However, these results show that newborns have not yet learned a template of the human body that withstands size reduction, and enables them to recognize humans when scale information is altered. To decide on this issue, experiments using stimuli projected in actual size may be needed.
In contrast, 3.5-month-olds discriminate and prefer a human body when compared with a non-human primate body. Their human representation is strong enough to overcome both size reduction and altered information, as infants could discriminate human headless bodies from a monkey's body presented in the same fashion. This quite robust discrimination could be supported by a coarse representation of human bodies which includes conspecifics, rather than by a detailed one which would include knowledge about the location of individual body parts. This interpretation is supported by the findings that knowledge about the configural structure of the human body was not demonstrated until nine months of age . Similarly, infants tested for their knowledge about the relative proportions of human body parts did not show accurate discrimination of body proportions until nine months . Both of these tasks are more challenging, and require more detailed knowledge than simply recognizing that stimuli are human or not, as was required in the current studies. Nonetheless, at 3.5 months of age infants are able to discriminate a human from a non-human primate who is perceptually similar in terms of overall body shape and facial structure. It is unclear which features allow infants to make the discrimination; it could be colour, texture or subtle differences in overall shape and proportion. It would be of interest to run the current experiments with 3.5- and six-month-olds using eye tracking, to establish which features they use to discriminate human and non-human primates in each condition.
Whereas our results allow us to draw conclusions on the presence of overall body representation at an early age, future research should test for neonatal preferences for other human body parts such as hands, which are presented within infants' visual field on a regular basis. This would further our understanding on infants' perceptions of relevant stimuli.
Preparation of this paper was supported by NIH grant NIH R01 HD046526. Michelle Heron-Delaney was supported by ESRC grant RES-000-22-3246.
One contribution of 10 to a Theme Issue ‘Face perception: social, neuropsychological and comparative perspectives’.
- This journal is © 2011 The Royal Society