The Infant Face Preference (Scrambled vs. Normal Faces) – Robert Fantz

For the greater part of the twentieth century, experimental psychology and sensory physiology operated under the premise that the human neonate enters the world visually incompetent, mentally unformed, and perceptually disorganized. In this view, sensory input was treated as a chaotic wash of uncalibrated luminance and color. The infant was thought to be incapable of extracting visual form, pattern, or semantic significance from their visual surroundings until weeks or months of postnatal experience and neural maturation had accumulated. This paradigm rendered the neonate an essentially passive, reflexive organism—a neurobiological slate waiting for environmental contingencies or gradual sensorimotor coordination to instill order upon an otherwise incomprehensible sensory field.

This long-standing empirical consensus was challenged in the late 1950s and early 1960s by the pioneering work of developmental psychologist Robert L. Fantz. Armed with a remarkably simple yet conceptually revolutionary experimental methodology known as the preferential looking paradigm, Fantz demonstrated that newborn and young human infants possess an intrinsic capacity to discriminate between visual patterns. More provocatively, his empirical investigations revealed that infants exhibit systematic, unlearned visual preferences for specific visual configurations. Chief among these was a marked, statistically robust preference for the schematic arrangement of the human face over scrambled, visually equivalent controls.

The demonstration that human infants preferentially fixate on canonical facial arrangements over scrambled features did more than upend early psychophysical models of vision; it catalyzed a foundational paradigm shift across developmental psychology, cognitive science, and evolutionary psychobiology. By providing an objective behavioral index of preverbal perception, Fantz’s work demonstrated that selective attention and form vision operate from the very dawn of postnatal life. This foundational inquiry initiated a multi-decade scientific endeavor to resolve the mechanisms underlying infant face perception: Was this early preference governed by a dedicated, innate neurocognitive module tuned to social conspecifics, or did it emerge from low-level visual filters sensitive to general spatial frequencies and geometric asymmetries? This article explores Robert Fantz’s landmark visual preference experiments, dissecting their historical context, methodological innovations, neurobiological mechanisms, and enduring legacy in contemporary cognitive neuroscience.

1. Historical Foundations and Robert Fantz’s Paradigmatic Shift in Developmental Psychology

1.1 The Pre-Fantz Consensus: The Neonate as a Passive Organism

Prior to the empirical breakthroughs achieved by Robert Fantz, developmental psychology and sensory physiology were dominated by paradigms that systematically underestimated the cognitive, perceptual, and visual capacities of the human infant. The philosophical underpinnings of this consensus were deeply rooted in classical British empiricism, particularly the epistemological frameworks of John Locke and George Berkeley. Locke’s conception of the mind as a tabula rasa—a blank slate upon which sensory experience inscribes all knowledge—implied that structured perception could not exist a priori. Sensory inputs were conceptualized as raw, atomistic sensations that required extensive cross-modal associations, learned environmental contingencies, and motor interactions before resolving into coherent perceptual representations of external objects.

This empiricist heritage was articulated in the late nineteenth century by William James in his foundational work, The Principles of Psychology (1890). James famously characterized the subjective sensory experience of the human newborn as “one great blooming, buzzing confusion.” In Jamesian psychology, the neonate was presumed to be assaulted by an undifferentiated wash of sensory stimuli, lacking the neural organization, selective attentional mechanisms, and cognitive schemas necessary to parse figure from ground, distinguish spatial depth, or recognize patterned forms. The infant was viewed as perceptually blind to the structural invariants of the physical and social world, reacting only through uncoordinated subcortical reflexes.

Throughout the first half of the twentieth century, this passive model of neonatal function was reinforced by the ascendance of radical behaviorism, championed by John B. Watson and later B.F. Skinner. Behaviorist orthodoxy maintained that psychological inquiry must restrict itself exclusively to observable stimulus-response contingencies, explicitly rejecting internal cognitive states, innate perceptual structures, and unobservable central processing. Within this framework, visual perception was treated as an associative response system shaped slowly over time through external reinforcement and conditioning. The neonate was perceived as an organism governed strictly by physiological drives and primitive reflexes, such as rooting, sucking, and grasping, with no endogenous visual organization.

Compounding these theoretical assumptions were severe methodological and technological limitations. Early psychophysicists possessed no reliable techniques to objectively probe the visual world of preverbal subjects. Because infants could not follow verbal instructions, provide psychophysical judgments, or execute coordinated voluntary motor responses, researchers routinely concluded that their sensory systems were non-functional or functionally blind. Standard clinical examinations relied on crude measures, such as the pupillary light reflex or the defensive blink, which revealed little about pattern discrimination. Consequently, the scientific consensus prior to the 1950s remained anchored in the assumption that form perception, pattern discrimination, and social recognition were impossible until several months of postnatal cortical maturation and sensorimotor learning had taken place.

1.2 Robert Fantz and the Origin of Visual Preference Research

The conceptual breakthrough that dismantled this century-old consensus originated not from traditional human clinical psychophysics, but from comparative psychology and classical ethology. During his doctoral and postdoctoral studies at the University of Chicago and Western Reserve University (now Case Western Reserve University), Robert L. Fantz engaged in extensive experimental investigations examining the innate visual capacities of non-human animals, particularly avian neonates. Influenced by European ethologists such as Nikolaas Tinbergen and Konrad Lorenz, who had demonstrated that innate releasing mechanisms guide specific instinctual behaviors in response to sign stimuli, Fantz sought to determine whether visual form perception could operate prior to visual experience.

Fantz designed experimental apparatuses to evaluate the pecking behaviors of newly hatched, dark-reared domestic chicks (Gallus gallus domesticus). Deprived of all prior visual exposure from the moment of incubation, these chicks nonetheless displayed immediate, non-random pecking preferences when presented with three-dimensional geometric stimuli differing in shape, roundness, and angularity. The chicks consistently targeted spherical, grain-sized objects over flat or irregular shapes. Because these animals had never received food reinforcement or visual exposure, Fantz recognized that this selective visual behavior was an innate perceptual adaptation: the visual nervous system was pre-wired to detect, differentiate, and orient toward specific ecological invariants necessary for survival without requiring prior learning trials.

Recognizing the profound epistemological implications of these findings, Fantz hypothesized that the human infant, far from being a passive tabula rasa, might also possess an evolved, unlearned visual architecture. However, translating this comparative paradigm to human subjects presented a severe methodological hurdle: human neonates lack the precocial motor competencies of avian hatchlings. They cannot peck, point, reach, or manipulate objects. Fantz ingeniously realized that while human infants lack coordinated somatic motor output, their oculomotor system—the muscles controlling eye gaze, visual orientation, and foveation—is functional at birth.

Fantz turned the infant’s visual gaze from an uninterpretable physiological movement into an objective dependent variable. He reasoned that if an infant consistently spent more time looking at one visual pattern over another when both were presented simultaneously under controlled, counterbalanced conditions, this differential looking could only be explained by two fundamental cognitive operations: first, the infant’s visual system must be capable of discriminating the sensory differences between the two patterns; and second, an endogenous visual preference or selective attentional bias must be directing ocular fixation. This conceptual leap transformed infant research. By relying on spontaneous exploratory gaze rather than conditioned motor responses, Fantz broke through the methodological impasse of preverbal infancy, shifting the paradigm from stimulus-response behaviorism toward an active, cognitive model of perceptual exploration.

1.3 Epistemological Impact on Cognitive Development

Fantz’s introduction of the preferential looking paradigm fundamentally altered the epistemological architecture of developmental psychology. Prior to his discoveries, the prevailing cognitive developmental paradigm was that formulated by Jean Piaget. In Piaget’s genetic epistemology, the infant during the early sensorimotor stages (specifically Substages 1 and 2, spanning birth to approximately four months) was described as operating through uncoordinated, solipsistic reflex actions. Piaget posited that perception and cognition are constructed through the gradual coordination of sensory modalities with physical action—looking, grasping, and sucking. For Piaget, visual space was initially uncoordinated; an object did not exist independently of the infant’s immediate physical interaction with it, and perceptual awareness of stable external forms emerged only after prolonged physical manipulation of the environment.

Fantz’s empirical demonstrations directly challenged Piagetian timelines regarding early perception. By revealing that infants within the first days and weeks of life systematically differentiated between visual forms—and showed reproducible preferences for structured patterns over unpatterned surfaces—Fantz established that visual organization precedes motor coordination. The infant did not need to physically touch, grasp, or manipulate an object to register its spatial structure or pattern boundaries. Perceptual organization was demonstrated to be an initial condition of human cognitive architecture rather than a late-stage byproduct of sensorimotor integration.

Furthermore, Fantz’s work served as a primary catalyst for the emerging “cognitive revolution” within the behavioral sciences, helping to bridge the divide between developmental psychology, psychobiology, and neurophysiology. By proving that preverbal infants actively process, select, and categorize visual information from birth, his research opened up previously inaccessible domains of developmental psychobiology. Infant cognition was no longer conceptualized as an empty vessel awaiting environmental conditioning, but as an active, information-seeking system equipped with biological constraints and computational biases designed to make sense of the surrounding world.

This shift redefined the ontological status of the infant within developmental science. Fantz transformed the neonate from an inert, solipsistic creature buffeted by sensory chaos into an active epistemic agent. The preferential looking method laid the technical and theoretical foundation for nearly all modern paradigms in infant visual cognition, including habituation-dishabituation techniques, violation-of-expectation experiments, and modern computational eye-tracking. In doing so, it set the stage for one of the most contentious, productive debates in modern cognitive science: the nature, origin, and neural basis of human face recognition.

2. Theoretical Backdrop: Empiricism vs. Nativism in Early Visual Perception

2.1 The Nativist Perspective: Pre-Wired Social Biases

The discovery that neonates demonstrate an elevated visual preference for facial forms over other stimuli reignited the long-standing philosophical and biological debate between nativism and empiricism. Within the cognitive sciences, the nativist perspective—frequently conceptualized through evolutionary psychology and modern cognitive modularity—posits that the human infant arrives in the world equipped with phylogenetically evolved, specialized neural mechanisms dedicated to processing biologically vital social signals. Prominent theorists, drawing on evolutionary biology, argue that because the human infant is born in a state of extreme altriciality, dependent upon adult conspecifics for physical nourishment, thermal regulation, and protection against predators, an innate mechanism designed to detect and visually lock onto conspecific caregivers confers an overwhelming selective advantage.

Nativist models, such as those historically influenced by Jerry Fodor’s modularity thesis and later refined by developmental cognitive neuroscientists like Mark H. Johnson and John Morton, suggest that natural selection has chiseled domain-specific mechanisms directly into the infant nervous system. According to this framework, face recognition is not treated by the neonatal brain as merely another arbitrary visual problem alongside reading text or identifying geometric shapes. Instead, the brain is assumed to contain dedicated, genetically specified neural circuits designed to process the structural geometry of the face prior to any environmental learning. This specialized machinery ensures that the infant’s attention is reliably directed toward social partners, facilitating the immediate initiation of maternal-infant bonding, social communicative signaling, and mutual affective regulation.

From an evolutionary perspective, selective attention to conspecific morphological traits must be operational at the moment of birth. Waiting for an altricial infant to gradually deduce through general associative learning mechanisms that faces are biologically important would represent a substantial survival liability. Nativists argue that natural selection solved this optimization problem by encoding a coarse structural template of the conspecific face into subcortical visual pathways. This template acts as a sensory trigger, ensuring that the neonate preferentially foveates human faces over inanimate visual noise, anchoring the infant within the social matrix from their first postnatal hours.

2.2 The Constructivist and Empiricist Counter-Arguments

Directly opposing the nativist interpretation of early face preferences are empiricist, constructivist, and domain-general perceptual accounts. Constructivist researchers argue that postulating innate, domain-specific face recognition modules is an unnecessary biological assumption. Instead, these theorists propose that neonatal face preferences can be explained through the interaction of domain-general sensory processing biases and visual physics, without requiring an innate psychological concept or structural representation of a “face.”

Central to this empiricist critique is the psychophysics of the immature neonatal visual system. At birth, the visual pathway of the human infant is structurally and functionally immature. The retina exhibits widely spaced, stubby foveal photoreceptors with low pigment density; myelination along the optic nerve and geniculostriate pathway is incomplete; and the primary visual cortex (V1) displays low synaptic density and poorly differentiated laminar architecture. Consequently, the newborn infant operates with low visual acuity (approximating 20/400 to 20/600) and an attenuated contrast sensitivity function (CSF) that is blind to high spatial frequencies (fine details and sharp edges) and capable of resolving only low spatial frequencies (broad luminance boundaries, coarse patterns, and large blobs).

Constructivists, such as Martin S. Banks and Philip Salapatek, argued that infant fixation behavior is driven by low-level physical characteristics of the stimulus—such as total luminance, contour density, spatial frequency spectra, and local contrast energy—rather than by semantic, social content. According to this “sensory hypothesis,” infants look at facial figures not because they represent a human social partner, but because facial configurations happen to possess an optimal distribution of low spatial frequencies and high-contrast transitions (e.g., the dark sclera-pupil margins against pale sclera, or the hairline framing the forehead) that match the tuning curve of the neonatal visual filter. In this view, face recognition is not an innate specialized adaptation, but a developmental derivative—an emergent property constructed over months of visual experience as the infant’s maturing sensory apparatus repeatedly encounters high-contrast, face-like configurations in their immediate caregiving environment.

2.3 Reconciling Nature and Nurture: Interactive Specialization

The historical polarization between strict nativism and radical constructivism has largely given way to a sophisticated, epigenetically informed synthesis within modern developmental cognitive neuroscience. Spearheaded by researchers such as Mark H. Johnson, this modern paradigm is known as the Interactive Specialization framework. Interactive Specialization posits that brain development is neither a purely pre-programmed maturation of isolated, adult-like modules nor an unstructured, domain-general tabula rasa that simply absorbs environmental input.

Instead, this perspective argues that the infant brain begins with minimal innate biases. These biases may not consist of fully formed, adult-like cortical representations of faces, but rather basic, subcortically mediated sensorimotor orientations that guide early behavioral attention toward biologically relevant inputs. By biasing the neonate to orient toward face-like stimuli, this primitive system ensures that the immature, highly plastic cerebral cortex receives continuous exposure to human faces. Through this steady, experience-expectant visual input, cortical pathways—particularly the fusiform face area (FFA) within the ventral occipitotemporal cortex—gradually self-organize, differentiate, and specialize to process the intricate nuances of human facial identity, emotion, and gaze.

This modern neuroconstructivist model emphasizes the concept of sensitive and critical periods in visual development. Experience-expectant plasticity relies upon species-typical environmental input arriving at specific developmental intervals to stabilize, prune, and consolidate neural circuits. If that input is absent or distorted during critical windows—as demonstrated in clinical studies of infants born with bilateral dense congenital cataracts—the normal specialized neural architecture for face processing fails to consolidate properly, resulting in long-term deficits in holistic and configural face perception. Thus, the current consensus views the infant face preference not as an isolated nativist module or a passive empiricist artifact, but as an evolved biological coordination where minimal innate sensory biases channel social experience to construct a specialized, adult-like social brain.

3. The Preferential Looking Paradigm: Methodological Innovations and Technical Apparatus

3.1 Design and Mechanics of the Fantz Looking Chamber

To overcome the methodological challenge of objectively quantifying visual perception in preverbal, non-motorically coordinated infants, Robert Fantz engineered an experimental apparatus that became known as the Fantz Looking Chamber (frequently referred to as the visual preference apparatus). Prior to this innovation, experiments assessing infant vision were plagued by environmental artifacts: fluctuating ambient lighting, distracting acoustic noises, variable experimenter postures, and visual clutter, all of which introduced massive confounds into behavioral data. Fantz recognized that isolating the infant’s visual choices required total environmental control, standardized viewing conditions, and an enclosed testing space.

The looking chamber consisted of a large, rigid, rectangular wooden or fiberboard enclosure painted flat black or uniform neutral gray on the interior surfaces to prevent stray reflections and eliminate visual distractions. The infant was placed securely on a soft, contoured crib cradle situated comfortably within the base of the chamber. The infant lay on their back in a semi-reclined or supine posture, elevated at an angle calculated to keep the head stabilized, minimize fatigue, prevent positional discomfort, and align the eyes symmetrically with the visual targets suspended directly above them on the interior ceiling of the chamber.

Positioned on the ceiling of the chamber, directly within the infant’s binocular visual field (typically at a standardized viewing distance of approximately 30 to 45 centimeters, matching the neonatal focal distance), were dual display holders designed to present pairs of visual stimuli simultaneously. These holders were mounted symmetrically to the left and right of the infant’s central visual axis, separated by a fixed distance to ensure that looking at one stimulus required an unambiguous ocular saccade away from the other. The lighting inside the chamber was provided by concealed, indirect fluorescent lamps positioned to bathe the visual stimuli in uniform, diffuse illumination while casting no visible shadows, hotspots, or specular glare that might bias the infant’s attention toward one side of the apparatus.

3.2 Corneal Reflection Measurement and Observer Reliability

The defining technical breakthrough of the Fantz looking chamber lay in its objective, reliable method for determining where the infant was fixating: the corneal reflection technique. Because an infant’s eye is a curved, reflective optical structure, light reflected from an illuminated stimulus off the outer surface of the cornea acts as a convex mirror, generating a tiny, high-contrast, pinpoint reflection of the stimulus known as the first Purkinje image.

Fantz placed a tiny, concealed pinhole peephole (approximately 0.25 to 0.5 inches in diameter) directly in the ceiling of the chamber, centered between the two stimulus display holders. Through this peephole, a trained human observer, completely concealed from the infant’s view to prevent social distraction, looked directly down into the infant’s eyes. When the infant looked directly at a visual stimulus, the corneal reflection of that specific stimulus was clearly visible, aligned over the center of the infant’s pupil. If the infant looked away, glanced at the alternative stimulus, or gazed randomly into the dark field, the corneal reflection shifted outside the pupillary boundary or vanished entirely.

This optical principle eliminated the subjective guesswork that had compromised previous infant behavioral research. The experimenter did not need to guess whether the infant was attending; visual fixation was operationalized as the presence of the stimulus reflection directly over the pupil. Using silent hand-held stopwatches or continuous electric event-recorders connected to micro-switches, the observer recorded the exact onset, offset, and cumulative duration of each fixation directed toward the left and right targets. To ensure scientific rigor and eliminate observer bias, Fantz incorporated rigorous inter-rater reliability protocols. Two independent observers frequently coded the same trials simultaneously through dual peepholes, consistently yielding inter-observer reliability correlation coefficients exceeding r = 0.90. The primary dependent measures extracted from this paradigm were total cumulative fixation duration (the total time spent looking at a given stimulus across a trial) and fixation frequency (the discrete number of individual glances directed toward each target).

3.3 Methodological Controls and Counterbalancing Protocols

To ensure that differences in looking time reflected true cognitive discrimination and visual preference rather than environmental artifacts, Fantz instituted rigorous experimental controls and counterbalancing procedures. A primary challenge in infant ocular psychophysics is the prevalence of endogenous positional biases; human infants frequently exhibit spontaneous, transient head or gaze preferences toward one side (most commonly the right side), independent of stimulus content. Fantz neutralized this lateral bias by utilizing paired, counterbalanced presentations.

Each pair of stimuli was presented to the infant in two consecutive, matched trials. If Stimulus A (e.g., a canonical schematic face) was presented on the infant’s left side and Stimulus B (e.g., a scrambled face) on the right side during the first trial, their spatial locations were reversed for the second trial, with Stimulus A placed on the right and Stimulus B on the left. The positions were transposed while an opaque shutter or shield blocked the infant’s visual field, preventing tracking during the transition. Total fixation times for each stimulus were then summed across both trials, canceling out any lateralized motor or visual bias.

Furthermore, Fantz standardized trial durations (typically ranging between 30 and 120 seconds per pair) and established strict criteria for managing the behavioral state of the infant. Utilizing what developmentalists later formalized as Prechtl’s behavioral states, Fantz tested infants only when they were in an optimal state of “alert inactivity”—a physiological window characterized by regular respiration, minimal gross somatic motor thrashing, absence of crying or vocal distress, and open, exploratory eyes. If an infant drifted into drowsiness, fussiness, or vigorous crying during a trial, testing was paused or the trial was discarded, ensuring that the empirical data reflected attentional processing unimpeded by physiological distress or fatigue.

4. Experimental Design: Normal vs. Scrambled Facial Configurations

4.1 Stimulus Construction: Geometric Equivalence and Feature Balancing

At the center of Robert Fantz’s empirical inquiries was an experimental condition designed to determine whether infant attention was governed by the holistic, structured form of the human face or merely by its constituent parts: the comparison between normal canonical schematic faces and scrambled facial configurations. Designing visual targets for preverbal infants requires rigorous psychophysical control. If an experimenter simply compares a photograph of a mother’s face to an arbitrary wooden block or a plastic ball, any observed visual preference could be attributed to trivial low-level physical variables—such as differences in color, total reflected luminance, edge contrast, spatial area, or geometric complexity—rather than a selective preference for facial organization.

Fantz addressed this challenge by designing hand-drawn, high-contrast schematic stimulus disks that isolated form and configuration while holding lower-level physical parameters strictly constant. The canonical face stimulus consisted of a two-dimensional, circular or oval flat disk (approximately 6 to 8 inches in diameter) featuring a pale, flesh-toned, or neutral off-white background framed by a dark, high-contrast perimeter. Within this circular boundary, Fantz drew stylized internal features representing the human face in their canonical spatial arrangement: two horizontal black shapes representing eyes, complete with brow markings, positioned symmetrically in the upper hemisphere; a vertical central line or wedge denoting the nose; and a horizontal crescent or bar marking the mouth in the lower hemisphere.

The scrambled face stimulus was constructed with identical components. Fantz cut out or reproduced the exact internal constituent features—the identical eyes, eyebrows, nose, and mouth—drawn with the same black ink, possessing the identical stroke widths, surface areas, and contrast boundaries as the canonical stimulus. However, within this scrambled disk, the features were arranged in a structurally disrupted, non-canonical array. The eyes might be stacked vertically on one side, the nose displaced to an upper corner, and the mouth skewed diagonally across the top margin. In this manner, Fantz engineered an elegant psychophysical control: both stimuli possessed identical overall geometric dimensions, identical background surface areas, identical total amounts of black-to-white contrast energy, and identical constituent elements. Any differential looking time between the two plates could not be attributed to low-level luminance, total contour density, or local feature contrast, but had to arise from the spatial arrangement—the gestalt—of the facial structure itself.

4.2 Categorization of Visual Target Variations

Fantz did not restrict his experimental inquiries to a simple binary comparison between a canonical face and a single scrambled counterpart. To map the visual preferences of infants across a wider spectrum of geometric variables, he constructed a graded battery of stimulus plates. These target variations allowed him to disentangle the relative contributions of social meaning, abstract patterning, structural symmetry, and general visual complexity.

The visual arrays utilized across Fantz’s foundational laboratories typically encompassed four distinct conceptual categories:

• Canonical Schematic Face Plates: Stylized facial line drawings exhibiting standard first-order spatial relations—eyes above nose, nose above mouth, housed symmetrically within an oval or circular frame.
• Scrambled and Disarranged Face Plates: Feature-matched control disks where internal facial elements were randomized, clustered off-center, inverted, or arranged in non-social linear bands, disrupting the relational configuration while preserving local feature identity.
• Non-Social Geometric Pattern Controls: Plates devoid of facial features but matched for high visual complexity, such as nested concentric circles (bullseyes), alternating black-and-white checkerboards, parallel horizontal or vertical stripes, and intersecting geometric lattices.
• Unpatterned Homogeneous Disks: Uniform, low-complexity control surfaces containing no internal contours, consisting of solid red, yellow, white, or neutral gray disks, designed to establish the absolute baseline of visual fixation in the absence of pattern information.

In addition to stylized line drawings, Fantz occasionally utilized high-contrast black-and-white photographic portraits of human faces alongside cut-and-pasted photographic scrambled equivalents. By evaluating infant visual fixation across this systematic gradient, Fantz was able to construct a clear hierarchy of visual preference, observing how fixation times shifted when an infant transitioned from looking at unpatterned disks, to geometric targets, to scrambled features, and ultimately to the intact, canonical human face.

4.3 Experimental Cohort Characteristics

To establish the developmental trajectory of visual preferences and identify whether the face bias was present at birth or acquired through exposure, Fantz evaluated large cohorts of human infants across a broad age span. His investigations included cross-sectional and short-term longitudinal cohorts ranging from neonates tested within hours of delivery to older infants up to six months of age.

Fantz maintained rigorous inclusion and exclusion criteria to ensure that perceptual performance was not confounded by neurodevelopmental trauma, systemic illness, or visual pathology. Neonatal cohorts were recruited directly from hospital maternity wards, requiring documentation of normal full-term gestational delivery (typically between 38 and 42 weeks of gestation), birth weights falling within normal clinical boundaries, unremarkable delivery histories free of severe hypoxic events, and high APGAR scores (typically 8 or higher at both one and five minutes postpartum). By assessing neonates under five days old—some tested within their first 24 to 48 hours of life—Fantz isolated an experimental window where postnatal environmental visual exposure to human faces was minimal.

Older cohorts, encompassing infants aged two weeks to six months, were tested in specialized laboratory chambers equipped for infant comfort. By dividing these cohorts into distinct developmental brackets (e.g., under 5 days, under 1 month, 1–2 months, 2–3 months, 3–4 months, and 5–6 months), Fantz mapped visual development over time. This cross-sectional and longitudinal strategy enabled him to directly track whether the preference for normal faces over scrambled faces remained static, intensified with social experience, or exhibited developmental shifts corresponding to the neuroanatomical maturation of the infant’s visual pathway.

5. Key Findings of Fantz’s Landmark 1961 and 1963 Studies

5.1 Quantitative Analysis of Fixation Durations

The empirical results gathered through Fantz’s preferential looking apparatus yielded definitive quantitative data that refuted the hypothesis of early infant visual indifference. Across hundreds of experimental trials, Fantz demonstrated that human infants do not cast their gaze randomly across visual targets. When presented with paired stimuli, infants systematically allocated their visual fixation toward specific targets, generating statistically significant differences in both total cumulative fixation duration and initial orienting frequencies.

When normal schematic faces were presented alongside scrambled facial configurations, the data revealed a consistent, statistically robust bias favoring the intact, canonically arranged face. In his systematic recordings, Fantz found that infants across various age brackets allocated a clear majority of their visual fixation time—frequently exceeding 60% to 70% of total looking time—to the canonical face plate rather than the scrambled control. The scrambled face, despite containing the identical black-and-white internal features, identical contour perimeters, and identical total luminance, failed to sustain visual attention to the same degree.

Furthermore, when looking times were compared across the wider spectrum of non-social visual targets, Fantz discovered a structured hierarchy of visual preference that remained remarkably consistent across subjects. The intact schematic face sat at the apex of this hierarchy, closely followed by complex, multi-element geometric patterns such as the concentric bullseye and fine checkerboards. Moderately complex patterns, such as linear stripes, occupied the middle tier, while uniform, unpatterned disks (solid red, yellow, or white circles) consistently occupied the bottom of the distribution, eliciting negligible fixation times. The quantitative data demonstrated that the human infant’s visual system is endogenously tuned to prioritize patterned visual input over unpatterned stimuli, with the canonical facial arrangement serving as the most powerful visual attractant in the early postnatal environment.

5.2 The 1961 Science Paper: ‘The Origin of Form Perception’

In May 1961, Robert Fantz published a concise four-page report in Science entitled “The Origin of Form Perception”. This paper dealt a major empirical blow to the behaviorist and classical empiricist consensus, providing verifiable evidence that visual form perception is operational in human infants prior to prolonged visual experience or motor practice.

In this landmark paper, Fantz presented findings from a study evaluating infants ranging from four days to six months of age. The infants were presented with a series of six flat, circular visual targets, each six inches in diameter, consisting of: a stylized line drawing of a human face; concentric black-and-white rings (a bullseye); a page of printed text; and three uniform disks of distinct colors (yellow, white, and red). The results were unequivocal across every age bracket tested. Even the youngest infants, under five days of age, spent significantly more time fixating on the patterned disks—with the face drawing eliciting the highest overall percentage of total looking time—than on the bright, uniform, unpatterned disks.

Fantz highlighted the theoretical weight of this discovery: the human infant does not begin life in a state of formless sensory confusion. The finding that neonates under five days old reliably separated, categorized, and preferred the patterned face stimulus over unpatterned controls dismantled the assertion that form perception requires months of sensorimotor association. Fantz asserted that the capacity to perceive visual form is innate or emerges through immediate neurobiological maturation independently of prolonged practice. This 1961 publication provided developmental psychology with an objective methodology, demonstrating that the human infant possesses an active, organized visual mind from birth.

5.3 The 1963 Scientific American Report: Developmental Trajectories

Two years after his paper in Science, Fantz expanded upon his findings in an article for Scientific American (May 1963), titled “Pattern Vision in Newborn Infants.” This publication refined the developmental timelines of early vision, presenting comprehensive longitudinal and cross-sectional data detailing how visual preferences evolve from birth through the first six months of life.

In this work, Fantz presented detailed data specifically contrasting the canonical schematic face with both a scrambled face plate and a control plate where the internal features were replaced by concentric arcs or stripes matching the area of the features. The data confirmed that while neonates under one month exhibited a robust visual preference for both the normal and scrambled faces over the unpatterned or simple striped controls, they showed an elevated interest in the canonically arranged face. As infants matured into the two-to-three-month and four-to-six-month brackets, their visual behavior became increasingly refined. Older infants demonstrated sustained discrimination, systematically rejecting the scrambled configuration in favor of the normal facial pattern.

Fantz contextualized these developmental shifts within the rapid maturation of the underlying visual apparatus. As the infant’s visual acuity improves, contrast sensitivity expands to encompass higher spatial frequencies, and cortical visual networks begin to take over control from early subcortical reflex structures, visual preferences become increasingly attuned to subtle morphological distinctions. The 1963 paper demonstrated that while the preference for pattern and general face-likeness is present from birth, it undergoes an orderly, experience-dependent refinement. This process transitions from an early, coarse sensitivity to high-contrast features into an advanced cognitive system capable of evaluating complex facial expressions, social cues, and familiar human identities.

6. The Structural Hypothesis vs. Feature-Based Visual Preferences

6.1 Configural Processing versus Isolated Feature Saliency

The robust visual preference for normal over scrambled faces demonstrated by Fantz ignited a protracted empirical controversy concerning the exact cognitive and perceptual mechanisms driving the behavior. At the center of this debate was the distinction between configural processing and isolated feature saliency. Proponents of the structural hypothesis asserted that the infant visual system is attuned to first-order relational properties—the global, spatial layout of facial morphology, where two eyes reside horizontally above a central vertical nose, which in turn resides above a horizontal mouth. According to this view, the infant visual system evaluates the stimulus holistically, recognizing the canonical “gestalt” of the human face and rejecting configurations where these spatial relations are violated.

Conversely, skepticism arose regarding whether preverbal infants truly process global face configuration, or whether their attention is captured by specific, hyper-salient local visual features acting independently of the global structure. Prominent among these feature-based accounts was the “eye hypothesis.” The human eye is an exceptionally salient, high-contrast visual signal: the dark, circular boundary of the iris and pupil set within the bright white sclera creates an intense local contrast gradient. Researchers questioned whether the infant’s preference for normal faces was driven simply by the canonical placement of two high-contrast, horizontally aligned spots in the upper hemisphere of the stimulus. If neonates are drawn to paired dark-light contrast boundaries, a normal face might win looking-time competitions not because it is recognized as a face, but because the upper visual field contains an optimal, symmetrical pair of visual attractors that trigger primitive sensory reflexes.

To resolve this question, researchers conducted experiments dissecting the stimulus matrix. Experiments utilizing isolated features, such as presenting infants with single eyes, solitary mouths, or isolated geometric shapes, demonstrated that feature saliency alone cannot explain the preference. While infants are drawn to eye-like elements, these elements elicit maximal looking times only when they are situated in an upright, symmetrical, canonical relationship relative to the other internal features and the perimeter frame. The spatial configuration—the structural relationship binding the features together—consistently modulated the infant’s attention, providing strong support for the hypothesis that early vision is sensitive to basic relational arrangements rather than merely responding to isolated, high-contrast features.

6.2 The Linear Systems Analysis and Spatial Frequency Hypotheses

As psychophysical methods advanced during the 1970s and 1980s, the debate over infant face preference took a mathematical turn with the application of linear systems analysis and Fourier optics to infant vision. Psychophysicists such as Martin Banks, Philip Salapatek, and later Kirk Kleiner, challenged both nativist and configural explanations, proposing that early visual preferences could be predicted by modeling the infant’s contrast sensitivity function (CSF) alongside the two-dimensional spatial frequency spectra of the stimuli.

Linear systems theory posits that any visual image can be mathematically decomposed into an array of sinusoidal luminance gratings varying in spatial frequency (cycles per degree), amplitude, phase, and orientation. Because the newborn’s visual apparatus acts as an aggressive low-pass spatial frequency filter—attenuating or blocking high spatial frequencies (fine details, sharp edges) while transmitting only low spatial frequencies (broad structural blobs, large contrast transitions)—Banks and colleagues argued that the infant does not see the sharp, detailed lines seen by an adult observer. Instead, the infant perceives a blurred, low-resolution visual representation.

Kleiner and Banks (1987) conducted experiments to test whether neonatal preferences were driven by the stimulus’s amplitude spectrum (the amount of energy present at different spatial frequencies) or its phase spectrum (the spatial alignment and relational configuration of those frequencies). They constructed stimuli where the amplitude spectrum was held constant while the phase spectrum was altered to create normal versus scrambled facial layouts. Under certain experimental conditions, Kleiner observed that neonatal looking preferences were predicted largely by the low-pass energy distribution of the amplitude spectrum: targets whose spatial frequency distributions matched the optimal bandpass of the neonatal CSF were fixated longer, irrespective of whether the phase information produced a canonically recognizable face. These findings led some researchers to claim that the early preference for normal schematic faces was an epiphenomenon of low-level visual filtering. However, subsequent studies revealed that when stimuli are rigorously controlled so that their amplitude spectra are identical, human infants continue to reliably fixate the canonically phased facial layout over scrambled versions, demonstrating that linear energy models alone cannot account for the infant’s sensitivity to relational structure.

6.3 Upper-Heavy Visual Biases: The Top-Heavy Effect

At the turn of the twenty-first century, another sensory challenge to the domain-specific structural hypothesis emerged: the top-heavy bias. Pioneered by Italian cognitive scientists Chiara Turati, Francesca Simion, and their colleagues at the University of Padua, this line of research questioned whether the infant’s preference for canonical facial layouts stemmed from a dedicated social template or from an innate, non-specific geometric preference for stimuli with more visual elements in their upper half than in their lower half.

In standard human facial morphology, the upper half of the face naturally contains more high-contrast visual features (two eyebrows, two eyes with dark irises, eyelashes, and upper eyelid creases) than the lower half (which contains only the vertical philtrum, the nose tip, and the horizontal mouth contour). Turati and Simion designed a series of non-facial geometric experiments to isolate this variable. They created pairs of identical geometric borders (such as rectangles or broad ovals) containing arbitrary, non-social internal geometric shapes—such as small black squares. In one condition, three squares were placed in the upper half and one square in the lower half (a “top-heavy” geometric configuration). In the paired control, this arrangement was inverted, placing one square in the upper half and three in the lower half (a “bottom-heavy” configuration).

Their findings were striking: newborn infants under three days old demonstrated a consistent, statistically significant preference for the top-heavy geometric configurations over the bottom-heavy configurations, even though the stimuli bore no visual resemblance to human faces. When Turati and colleagues pitted an upright schematic face against a non-facial top-heavy geometric stimulus, neonatal preferences were often split or modulated by the relative distribution of upper-field contrast. These results led the Padua researchers to argue that the early preference for normal over scrambled faces is driven not by an innate cortical or subcortical representation of a face, but by an innate visual orienting mechanism biased toward general vertical asymmetry—specifically, an attentional attraction toward configurations with higher visual contour density in the upper visual field. This “top-heavy hypothesis” remains one of the most prominent domain-general counter-explanations to the classic structural face preference hypothesis.

7. Subsequent Replications and Refinements: Goren, Siqueland, and Johnson & Morton

7.1 The Goren, Sarty, and Wu (1975) Breakthrough

While Robert Fantz had demonstrated face preferences using static visual stimuli presented to infants lying in a stationary chamber, his findings were frequently contested on grounds that the preference was inconsistent during the first days of life, or that low-level sensory variables had not been fully decoupled from social configuration. In 1975, a clinical and experimental breakthrough occurred with the publication of a study by Carolyn C. Goren, Merrill Sarty, and Paul Y. K. Wu in Pediatrics, titled “Visual Following and Pattern Discrimination of Face-like Stimuli by Newborn Infants.”

Goren and colleagues introduced a major methodological innovation: instead of using static stimuli inside an enclosed chamber, they utilized dynamic, moving visual stimuli presented to newborns directly in the delivery and recovery rooms. The researchers constructed three flat, life-sized, paddle-shaped visual targets mounted on handles: a canonical schematic face, a scrambled face matched for feature surface area and contrast, and a blank, unpatterned white paddle. The infant was held comfortably in the experimenter’s lap in a semi-upright posture, and the paddle was introduced directly into the infant’s central line of sight at a distance of approximately 12 inches. Once foveated, the paddle was moved slowly in a continuous horizontal arc of 180 degrees around the infant’s head at a standardized velocity.

Using calibrated protractor markers situated above the infant’s head, the researchers recorded two discrete dependent measures: the degree of ocular visual tracking (eye-turn) and the degree of somatic head turning (head-turn) elicited as the infant tracked the moving target across the visual field. The experimental cohort was unprecedented in its youth: the infants were tested at a mean age of approximately nine minutes postpartum—prior to any sustained visual exposure to human faces outside of the delivery staff, who were masked. The findings demonstrated that these minutes-old neonates tracked the canonical schematic face paddle through significantly larger visual arcs (for both head and eye movements) than they tracked the scrambled or blank paddles. By moving the visual target, Goren and colleagues engaged the subcortical visual system, demonstrating that the structural preference for canonical facial configurations is functional immediately upon exit from the womb.

7.2 Methodological Disputes: The Maurer and Young (1983) Critiques

Despite the findings of Goren, Sarty, and Wu, the scientific community met their conclusions with cautious skepticism. In developmental psychobiology, claims of complex innate perceptual templates require extensive empirical verification, and during the late 1970s and early 1980s, several high-profile attempts to replicate the delivery-room dynamic tracking findings produced mixed or outright negative results.

The most prominent critique came from Daphne Maurer and Russell Young in a widely cited 1983 study published in Child Development. Maurer and Young attempted to replicate the paddle-tracking methodology using cohorts of one-month-old and two-month-old infants, as well as newborns. Their results diverged from the Goren et al. findings: while they confirmed that infants tracked structured patterns (both normal and scrambled faces) farther than blank paddles, they failed to find a statistically robust difference between the normal schematic face and the scrambled face plate. They observed that infants frequently tracked the scrambled paddle through visual arcs comparable to those elicited by the canonical face.

Maurer and Young raised methodological concerns regarding earlier studies, pointing to variations in the degree of feature scrambling, differences in background illumination, and the potential for experimenter bias. They noted that when an unblinded experimenter moves a paddle manually, subtle, unconscious adjustments in movement speed or arc trajectory could bias infant tracking toward the canonical face. Furthermore, they emphasized that the spatial arrangement of the internal features used across different laboratories was not standardized; some researchers scrambled only the peripheral contours while leaving internal eyes intact, while others dispersed the eyes and mouth randomly. These contradictory empirical reports cast doubt on the robustness of the innate face preference, prompting researchers to seek out more rigorous experimental designs, automated tracking systems, and mechanistic models to resolve these conflicting findings.

7.3 The Johnson and Morton (1991) Replications and Synthesis

The empirical ambiguities surrounding infant face perception were resolved in 1991 with the publication of a landmark monograph and series of studies by cognitive neuroscientists Mark H. Johnson and John Morton. Working at the Medical Research Council Cognitive Development Unit in London, Johnson, Morton, and their colleagues conducted rigorous, pre-registered replications of the Goren paradigm using large cohorts of newborns tested within the first hour of life, alongside systematic testing of one-, two-, three-, and five-month-old infants.

Johnson and Morton addressed the earlier methodological critiques by standardizing every parameter of the paddle-tracking paradigm. They utilized precisely engineered stimulus paddles: a canonical schematic face disk; an “inverted” face disk; a scrambled face disk where features were randomized while preserving local edge counts; and a blank paddle. Observers were blinded to the stimulus paddle being presented, and paddle velocity was mechanically or metronomically paced across standardized arcs. The results verified the findings of Goren et al.: newborn infants under one hour of age systematically turned both their eyes and their heads significantly farther to follow the moving canonical schematic face than to follow the scrambled, inverted, or blank paddles.

Crucially, Johnson and Morton extended their testing longitudinally and cross-sectionally across the first six months of life, uncovering a critical developmental phenomenon: the temporal dip. They discovered that while newborns under one month exhibited robust tracking preferences for the canonical face paddle over the scrambled paddle, this behavioral tracking preference declined precipitously between four and eight weeks of age, disappearing almost entirely around two months before re-emerging in a different behavioral form (in static preferential looking chambers) around three to four months of age. This U-shaped developmental trajectory explained why earlier researchers, such as Maurer and Young—who had tested infants primarily around one to two months of age—had failed to replicate the neonatal effect. By uncovering this developmental curve, Johnson and Morton laid the empirical foundation for their influential two-process neural model of face processing.

8. Two-Process Theory: CONSPEC and CONLERN Mechanisms

8.1 CONSPEC: The Subcortical Orienting Mechanism

To explain the empirical reality of the neonatal face preference and its subsequent developmental changes, Mark H. Johnson and John Morton formulated the Two-Process Theory of Infant Face Recognition, introducing the theoretical constructs of CONSPEC and CONLERN. CONSPEC represents an innate, phylogenetically ancient, subcortically mediated structural mechanism whose functional architecture is operational at birth.

The CONSPEC mechanism is mediated primarily by subcortical retinotectal and tectopulvinar visual pathways, centered on the superior colliculus and the pulvinar nucleus of the thalamus, with reciprocal connections to the amygdala. This subcortical visual system bypasses the primary visual cortex (V1), which is immature and largely unmyelinated in the human newborn. The superior colliculus is phylogenetically ancient and structurally optimized to process visual information from the peripheral retina. It is broadly tuned to low spatial frequencies and responds rapidly to dynamic motion, transient luminance changes, and coarse spatial forms.

CONSPEC does not contain an adult-like, high-resolution representation of a human face. Instead, it contains a minimal, coarse sensory template consisting of three high-contrast blobs arranged in an inverted triangle: two blobs corresponding to the eyes and one corresponding to the nose-mouth complex, framed within a symmetrical, face-like perimeter. Because CONSPEC operates primarily through the superior colliculus and the peripheral visual field, it is triggered most effectively by dynamic, moving stimuli—explaining why the paddle-tracking paradigms of Goren, Johnson, and Morton elicited profound preferences in newborns, whereas static preferential looking in stationary chambers produced more variable results. The evolutionary function of CONSPEC is not to identify specific human individuals, but to serve as a directional orienting device: an automatic, subcortical compass that steers the neonate’s eyes and head toward face-like objects in the immediate environment.

8.2 CONLERN: The Cortical Specialization System

The second arm of the two-process theory is CONLERN, a cortical, experience-dependent visual learning system. Unlike CONSPEC, which is innate and subcortical, CONLERN is an acquired, plasticity-driven neurocognitive system localized within the ventral visual processing stream, encompassing the lateral geniculate nucleus (LGN), primary visual cortex (V1), and the extrastriate and ventral occipitotemporal cortices—culminating in the fusiform gyrus (specifically the Fusiform Face Area, or FFA) and the superior temporal sulcus (STS).

CONLERN does not possess an innate structural template of the human face. Instead, it begins life as a highly plastic, general-purpose cortical visual matrix equipped with flexible neural networks capable of fine-grained pattern analysis, statistical feature extraction, and complex perceptual learning. As the infant matures over the first two to four months of life, synaptogenesis explodes within the cerebral cortex, laminar boundaries differentiate, dendritic branching expands, and myelination of the geniculostriate pathway accelerates. With these neuroanatomical developments, CONLERN becomes functional, gradually assuming control over ocular saccades, foveation, and visual processing from the primitive subcortical reflex circuits.

Under the CONLERN regime, facial processing shifts from coarse, low-spatial-frequency detection to sophisticated, high-resolution configural and holistic analysis. CONLERN is sensitive to second-order spatial relations—the precise geometric distances separating the eyes, the spatial distance between the nose and upper lip, and the subtle variations in surface texture, shading, and skin luminance that differentiate one human individual from another. CONLERN enables the infant to extract identity, recognize familiar caregivers, decode subtle emotional expressions, and track eye gaze for joint attention. It represents the specialized, mature face processing system that persists throughout human adulthood.

8.3 Developmental Transition and Interaction Between the Two Systems

The core insight of Johnson and Morton’s two-process theory lies in the dynamic interaction and developmental handoff between CONSPEC and CONLERN. The two systems do not operate in isolation; rather, CONSPEC acts as a biological “bootstrapper” for CONLERN. Because the immature human cortex lacks an innate face template, it depends on extensive, repetitive visual exposure to faces to tune and specialize its emerging neural circuitry. If the infant’s gaze were left to random environmental scanning, this cortical exposure might occur infrequently or inconsistently.

CONSPEC resolves this developmental problem. By mechanically orienting the newborn’s eyes toward canonical facial arrangements, CONSPEC ensures that during the first two months of life—when the infant is held in close physical proximity to adult caregivers—the infant’s fovea is repeatedly aligned with human faces. The subcortical orienting mechanism channels millions of facial exposures directly into the visual pathway, providing the neural stimulation required to drive experience-dependent synaptic organization and metabolic development within the ventral visual stream.

This theoretical framework provides an explanation for the classic U-shaped developmental curve in infant face preferences:

• Phase 1 (Birth to 4 Weeks): Visual behavior is dominated by subcortical CONSPEC mechanisms, resulting in robust tracking of moving schematic faces over scrambled controls across peripheral visual space.
• Phase 2 (4 to 8 Weeks): A transitional developmental dip occurs. Subcortical visual structures begin to come under descending cortical inhibition, but the ventral visual stream and CONLERN circuitry have not yet reached functional maturity. Consequently, primitive tracking preferences for schematic paddles diminish or become inconsistent.
• Phase 3 (2 to 3 Months and Beyond): The mature, cortical CONLERN system comes online, operating primarily through foveal central vision. Preferences for canonical faces re-emerge in static visual tasks, now accompanied by individual recognition, holistic processing, and identity differentiation.

This interaction demonstrates that the nativist and constructivist positions are not mutually exclusive. CONSPEC provides the initial nativist bias, which in turn secures the visual environment required for the constructivist assembly of CONLERN’s specialized cortical architecture.

9. Neurobiological Correlates of Early Face Perception

9.1 Subcortical Architectures vs. Cortical Immature Networks

The dual-process behavioral model is supported by neuroanatomical and electrophysiological investigations of the developing infant brain. In the human newborn, the subcortical and cortical visual pathways exhibit distinct maturation rates. The retinotectal pathway, which projects directly from the retina to the superior colliculus and the pulvinar nucleus of the thalamus, is phylogenetically older and achieves functional myelination and synaptic competence early in prenatal ontogeny. The superior colliculus possesses a retinotopic organization and is wired to coordinate rapid visual orienting responses, saccadic eye movements, and spatial orientation via descending tectospinal pathways. Its receptive fields are large and responsive to low spatial frequencies, but structurally unsuited for high-resolution pattern discrimination.

Conversely, the geniculostriate and ventral visual processing streams—responsible for high-acuity pattern vision and identity recognition—remain structural works in progress at birth. In the primary visual cortex (striate cortex, Brodmann Area 17), layer IV afferents from the lateral geniculate nucleus are sparsely organized, horizontal intra-cortical connections are underdeveloped, and the process of selective synaptic pruning has not yet commenced. Myelination of the optic radiations connecting the thalamus to the occipital pole progresses slowly across the first six months of life, accompanied by a continuous increase in visual evoked potential (VEP) conduction velocities.

Electrophysiological studies utilizing Event-Related Potentials (ERPs) provide insight into the emergence of cortical face processing. In adults, face perception is characterized by the N170, a prominent, negative-going occipitotemporal ERP component occurring approximately 170 milliseconds post-stimulus onset, which selectively responds to faces compared to non-face objects. In infants, researchers have identified developmental precursors to the N170: the N290 and the P400. The N290, observable in infants from approximately two to three months of age, displays an extended latency (often between 290 and 350 milliseconds) and shows sensitive amplitude modulation when infants view canonical versus scrambled or inverted faces. Over the first year of life, as axonal myelination and synaptic efficiency increase, the latency of the N290 steadily contracts, gradually shifting toward the adult-like N170 profile.

9.2 Hemispheric Lateralization in Infant Face Processing

A central question in developmental cognitive neuroscience concerns when and how hemispheric lateralization for face recognition emerges. In typical adults, face processing exhibits a pronounced right-hemisphere dominance, localized primarily to the right lateral fusiform gyrus (the right FFA) and the right superior temporal sulcus. This right-hemispheric specialization is associated with holistic, configural processing (integrating the face into a unified whole), whereas the left hemisphere is more engaged by analytical, local, feature-based processing.

Neuroimaging investigations utilizing Functional Near-Infrared Spectroscopy (fNIRS) and high-density electroencephalography (EEG) have demonstrated that right-hemisphere dominance for face perception is observable within the first months of life. Studies evaluating infants between two and six months old reveal that viewing canonical schematic faces and natural human faces elicits a significantly higher concentration of oxygenated hemoglobin ([HbO]) over right lateral occipital and temporal channels compared to homologous left-hemisphere channels. When infants view scrambled facial matrices or inverted faces, this right-lateralized hemodynamic activation is attenuated, and blood flow patterns become more diffuse or symmetrical across both hemispheres.

This early lateralization is hypothesized to stem from the intrinsic computational biases of the two cerebral hemispheres. The right hemisphere matures structurally faster than the left hemisphere during early postnatal life and possesses an endogenous processing bias toward low spatial frequency visual information and global structural relations. Because early face perception relies almost entirely on low-pass visual configurations transmitted by subcortical inputs, the emerging cortical representation naturally anchors within right-hemisphere ventral regions. Thus, the infant’s preference for canonical over scrambled facial layouts serves as an early driver of the functional lateralization that characterizes the adult social brain.

9.3 The Role of the Amygdala and Subcortical Emotion Circuitry

Beyond the classic visual sensory pathways, the neurobiological apparatus supporting early face perception is integrated with subcortical limbic circuits, particularly the amygdala. The human amygdala receives direct, non-striate visual input via a dedicated subcortical route traveling from the retina, through the superior colliculus, directly through the pulvinar nucleus of the thalamus, and into the basolateral amygdala. This pulvinar-amygdala pathway bypasses the primary visual cortex entirely, providing a rapid channel for low-spatial-frequency visual forms.

Neuroimaging and neuroethological studies confirm that this subcortical limbic circuit is tuned to detect biologically significant social stimuli, including facial patterns and eye-contact configurations. In neonates, direct eye contact and canonical facial arrays trigger activation within the amygdaloid nuclei, which in turn interface directly with autonomic, neuroendocrine, and arousal centers within the hypothalamus and brainstem. This pathway ensures that the visual detection of a conspecific face is paired with physiological arousal, pupil dilation, heart-rate deceleration (a marker of sustained orienting and focused attention), and the suppression of competing motor behaviors.

This direct subcortical emotion circuitry explains why early face perception is intrinsically tied to affective resonance and mutual social bonding. When a neonate foveates a canonical facial arrangement, the activation of this pulvinar-amygdala pathway serves not merely an analytical visual function, but an emotional and regulatory one. The canonical face operates as a soothing, stabilizing visual input, anchoring the infant’s attention and establishing the biological baseline required for reciprocal caregiver-infant interactions and autonomic regulation.

10. Evolutionary and Adaptive Significance of Innate Facial Biases

10.1 Attachment and Maternal Investment Optimization

The phylogenetic conservation of an innate visual preference for canonical facial structures is best understood through the lens of evolutionary biology and human life-history theory. In his foundational formulations of attachment theory, British psychoanalyst and ethologist John Bowlby posited that human infant survival depends upon a behavioral system designed to maintain proximity to protective, nurturing adult caregivers. Bowlby recognized that human neonates are unable to physically cling to their mothers in the manner of other primate species. Consequently, natural selection had to evolve psychological and perceptual mechanisms to establish and sustain this social bond.

The neonatal face preference functions as a proximate behavioral trigger that elicits parental caregiving and maximizes maternal investment. When an infant locks its gaze onto an adult caregiver’s eyes, looks selectively at the caregiver’s face, and follows the caregiver’s movements across the room, it generates an immediate affective response in the adult partner. Cross-cultural observations show that adult caregivers perceive sustained infant eye gaze as an intentional sign of social recognition, connection, and infant vitality. This visual locking stimulates neurochemical release within the caregiver—triggering surges of oxytocin and dopamine—which strengthens maternal bonding, suppresses parental neglect, and promotes intensive, sustained caregiving behaviors.

From an evolutionary perspective, an infant born without an innate bias to orient toward faces would be at an adaptive disadvantage. Such an infant might look indifferently at patterned wallpaper, ceiling fans, or shadows, failing to provide the mutual gaze cues that reinforce parental investment. By ensuring that the infant selectively targets the human face, CONSPEC optimizes parental attachment behavior. The infant’s selective looking at the canonical face is not simply an exercise in visual perception; it is an evolved survival strategy engineered to ensure protection and nourishment during altricial infancy.

10.2 Social Referencing and the Scaffolding of Intersubjectivity

The early visual bias toward canonical facial structures serves as the primary developmental foundation upon which human communication and cultural transmission are erected. Developmental psychologist Colwyn Trevarthen conceptualized the early face-to-face engagements between human infants and their caregivers as primary intersubjectivity. Primary intersubjectivity encompasses the coordinated exchanges of eye contact, vocalizations, smiles, and hand gestures that emerge around two months of age, forming what Trevarthen described as “communicative musicality” and proto-conversational turn-taking.

This intersubjective communicative system depends on the infant remaining continuously focused on the caregiver’s facial matrix. The infant’s bias toward canonical arrangements ensures that attention is directed toward the specific anatomical structures where social signals are transmitted: the eyes (which signal attentional focus, fear, and interest) and the mouth (which signals vocalization, pleasure, and communicative intent). As the infant matures, this early face-looking behavior provides the visual scaffolding for joint attention and social referencing. Between six and nine months of age, infants begin to follow an adult’s directional gaze shifts, using the eyes as a directional pointer to discover objects in the shared external environment.

Furthermore, in ambiguous or novel situations, infants engage in social referencing, looking to the caregiver’s facial expression to evaluate safety or threat before acting. The face is the primary medium through which emotional states are decoded, language acquisition is visually mapped (via the observation of articulatory lip movements), and cultural expectations are transmitted. The early preferential looking paradigm documented by Fantz is the initial developmental domino: by ensuring that the infant looks at the face, the visual system bootstraps the developmental progression of social cognition, culminating in complex theory of mind and symbolic communication.

10.3 Comparative Perspectives: Primate and Avian Homologues

The evolutionary roots of the infant face preference are illuminated by comparative studies across non-human primates and precocial avian species. Research with non-human primates confirms that innate visual biases toward conspecific facial morphology are shared across the primate order. Landmark investigations conducted with newborn rhesus macaques (Macaca mulatta) reared in visual isolation under controlled nursery conditions reveal that baby monkeys, tested within hours of delivery, demonstrate significant preferential looking toward schematic and natural macaque faces over scrambled and non-social controls.

Work led by Atsushi Senju, Masako Myowa-Yamakoshi, and colleagues with chimpanzees (Pan troglodytes) and macaques demonstrated that the subcortical paddle-tracking behavior documented in human infants is shared with infant non-human primates. Newborn chimpanzees track moving schematic faces through larger visual arcs than they track scrambled controls. Furthermore, infant monkeys reared from birth by human caretakers wearing masks—completely depriving the monkeys of visual exposure to faces—still display an immediate preference for faces when the masks are removed months later. This confirms that primate visual systems contain a robust, phylogenetically conserved subcortical mechanism that operates independently of prior visual exposure.

Beyond mammals, remarkable homologues exist in precocial avian species. In newly hatched domestic chicks (Gallus gallus), an innate mechanism known as filial imprinting drives the hatchling to attach to and follow its mother. Experimental work by Gabriel Horn, Mark H. Johnson, and colleagues revealed that the chick’s imprinting system is divided into two distinct components: a subcortical, unlearned predisposed bias toward a visual “head and neck” configuration (analogous to CONSPEC), and an acquired, forebrain-mediated learning system known as the intermediate and medial mesopallium (IMM, analogous to CONLERN). Newly hatched, dark-reared chicks with no visual experience consistently prefer to approach stuffed, taxidermic hens or stylized visual models displaying the canonical head-and-neck morphology over scrambled or inverted control models. The discovery of structurally analogous, dual-process visual orienting mechanisms across avian and mammalian taxa demonstrates that subcortical systems designed to orient neonates toward conspecific morphological structures represent an evolutionary strategy conserved across hundreds of millions of years of vertebrate evolution.

11. Methodological Critiques, Controversies, and Competing Explanations

11.1 The General Sensory/Perceptual Account

Throughout the history of infant visual perception research, the domain-specific interpretation of Fantz’s findings has faced persistent skepticism from psychophysicists advocating for general sensory explanations. The central thesis of the sensory account, championed by Martin Banks and colleagues, is that visual behavior in the first months of life can be explained without attributing social knowledge, internal templates, or domain-specific modules to the infant mind. Instead, this perspective models the neonate’s gaze as an automated response dictated by the optical properties of the neonatal eye, retinal photoreceptor geometry, and early neural transfer functions.

Sensory theorists argued that when an infant is presented with a canonical schematic face and a scrambled face, the two stimuli, while containing identical individual features, generate subtle differences in their two-dimensional Fourier power spectra. The spatial layout of features in a canonical face creates specific horizontal and vertical luminance profiles that may interact differently with the infant’s contrast sensitivity filter compared to the dispersed features of a scrambled plate. In some scrambled configurations, features placed close to the high-contrast circular border can create local visual crowding, high-frequency spatial interactions, or altered peripheral contrast ratios that may be less visible or less salient to the infant’s immature visual system.

To rigorously probe these sensory boundaries, researchers introduced stimulus transformations that altered social content while preserving physical energy distributions. One experimental approach utilized contrast-inverted stimuli (photographic negatives). A photographic negative preserves all spatial frequencies, geometric relations, contour densities, and feature configurations of an intact face, but reverses the luminance polarities: sclera become dark, pupils become bright white, and skin highlights turn black. If infant preferences are governed strictly by non-specific spatial frequency geometry, an infant should prefer a photographic negative face just as strongly as a positive face. However, empirical investigations show that while newborns will track contrast-inverted faces, their preference for them is reduced compared to normal, contrast-positive faces. This divergence suggests that while low-level sensory models explain significant aspects of neonatal visual processing, they cannot account for the preference for canonical faces without acknowledging sensitivity to naturalistic, species-typical luminance relations.

11.2 The Other-Species and Other-Race Effects in Early Infancy

Another challenge to understanding the infant face preference emerged through studies of perceptual narrowing—a developmental phenomenon that illuminates how broad, uncalibrated visual capacities refine into culturally and ecologically specific expertise. While Robert Fantz and early researchers often assumed that the neonatal face preference was tuned specifically to the human species, subsequent empirical research revealed that the neonatal visual template is broadly tuned.

Studies evaluating newborns under three days of age reveal that neonates demonstrate a lack of species specificity: they display looking preferences for canonical non-human animal faces (such as upright domestic cat or Barbary macaque faces) over scrambled controls that are comparable to their preferences for human schematic faces. Furthermore, newborns show no initial preference for faces belonging to their own ethnic group over faces belonging to other racial groups; they prefer any canonical face-like configuration over non-face patterns, regardless of ethnic morphology or species boundaries.

This initial broad tuning undergoes systematic perceptual narrowing across the first year of life, driven by environmental visual exposure:

• The Other-Species Effect: At six months of age, human infants can discriminate between individual human faces and can also discriminate between individual monkey faces with equal perceptual facility. However, by nine months of age, this capacity narrows: infants retain the ability to distinguish between individual human faces, but lose the ability to differentiate between individual monkey faces, treating non-human faces as a homogeneous group unless they are provided with regular, daily exposure to monkey faces during this critical window (as demonstrated by Olivier Pascalis and colleagues).
• The Other-Race Effect: A parallel narrowing process occurs with racial categorization. Newborn infants show no preference for faces of their own ethnic group over unfamiliar ethnic groups. However, by three months of age, infants reared predominantly in mono-ethnic caregiving environments begin to show a preference for faces matching the racial profile of their primary caregivers. By nine months of age, infants exhibit the classic other-race effect, demonstrating superior configural processing and individual recognition for own-race faces while exhibiting impaired discrimination for other-race faces.

The existence of perceptual narrowing supports the Interactive Specialization model over strict nativist models. The neonatal brain does not emerge with an innate catalog of human ethnic groups or species-specific traits. Instead, it provides a broadly tuned, coarse orienting mechanism (CONSPEC) that captures any face-like form, leaving it to environmental experience to tune and specialize the cortical processing machinery (CONLERN) to the human faces found in the infant’s specific cultural environment.

11.3 Replication Variations and Laboratory Discrepancies

The history of infant face perception research has not been without empirical discrepancies and failed replications. Variations across laboratories often generated contentious debates regarding the reliability of the infant face preference, with some investigators arguing that the phenomenon was an artifact of specific laboratory configurations, and others asserting that it was an unshakeable biological fact.

A meta-analysis of these conflicting findings reveals that these empirical discrepancies can be attributed to four major methodological parameters:

1. Dynamic versus Static Presentation: Laboratories that failed to observe a preference in newborns frequently utilized static stimuli presented inside stationary looking chambers, where the infant’s gaze was free to drift across non-moving displays. Conversely, laboratories that utilized dynamic, moving stimuli (such as the Goren and Johnson paddle-tracking paradigms) observed consistent, statistically significant preferences. Because subcortical pathways are sensitive to movement, dynamic paradigms systematically engage neonatal attention where static paradigms may fail.
2. Degree of Feature Scrambling: Early studies exhibited wide divergence in how “scrambled” control stimuli were constructed. In some laboratories, features were shifted by only a few millimeters, leaving the general top-heavy or three-blob triangular organization intact; these “weakly scrambled” controls were often treated by the neonatal visual system as acceptable face representations. When features were fully randomized or inverted—destroying the first-order relational structure—preferences for canonical faces emerged reliably.
3. Infant Behavioral State: Early studies did not always control for subtle shifts in infant state. Testing an infant during the onset of drowsiness or immediately following feeding introduces physiological noise that attenuates preferential orienting.
4. Observer Bias in Manual Coding: Prior to the advent of high-speed infrared corneal-reflection eye-tracking, manual recording through peepholes or from analog video footage introduced inter-rater variance, particularly when observers were not blinded to the spatial positioning of the targets.

When these methodological parameters are standardized—utilizing blinded protocols, dynamic or high-contrast calibrated stimuli, and infants in verified states of alert inactivity—the visual preference for canonical facial configurations over scrambled controls replicates across independent developmental laboratories worldwide.

12. Modern Legacy: Clinical Implications, Computational Modeling, and Future Trajectories

12.1 Atypical Face Processing as Early Biomarkers for ASD

Robert Fantz’s insights into early visual preferences have laid the groundwork for modern translational medicine, particularly in the early identification and diagnosis of neurodevelopmental conditions such as Autism Spectrum Disorder (ASD). Autism is characterized by social-communication challenges, atypical eye contact, and altered socio-emotional reciprocity. For decades, clinical diagnoses could not be made until the emergence of overt behavioral and linguistic markers between two and four years of age. However, using modern preferential looking paradigms and high-speed automated eye-tracking, researchers have identified atypicalities in visual orienting toward social stimuli within the first year of life.

Seminal longitudinal investigations conducted by researchers such as Ami Klin, Warren Jones, and Karen Pierce have evaluated infants at high familial likelihood for ASD (infants with an older biological sibling diagnosed with ASD) from birth through 24 months of age. These prospective studies have revealed distinct differences in visual scanning trajectories:

• Decline in Eye Fixation: Typically developing infants exhibit high, increasing levels of preferential fixation directed toward the eyes of human faces between two and six months of age. Conversely, infants who later receive a clinical diagnosis of ASD often show a progressive decline in eye fixation over this identical developmental window, shifting their gaze away from the eye region toward the mouth or non-social background regions.
• Geometric over Social Preference: When presented with paired dynamic stimuli featuring biological motion and social faces on one side, and complex, repetitive geometric patterns on the other, a significant subgroup of infants at elevated likelihood for ASD demonstrate a visual preference for the geometric patterns, fixating on repetitive physical animations rather than human faces.

These findings suggest that a disturbance in the innate, subcortical orienting mechanism (CONSPEC) or its transition to cortical specialization (CONLERN) may represent a core neurodevelopmental pathway toward atypical social development. If an infant’s brain does not preferentially orient toward faces during the first months of life, the immature cortical networks fail to receive the environmental exposure required to construct the social brain. These translational discoveries open up opportunities for early behavioral and perceptual interventions: by identifying atypical visual orienting during infancy, targeted social-communicative interventions can be deployed during windows of high neuroplasticity to support healthy social-cognitive developmental trajectories.

12.2 Artificial Intelligence and Deep Neural Network Models

Fantz’s questions regarding innate templates, perceptual learning, and structural processing have found a testing ground in modern computational neuroscience and artificial intelligence. Researchers in machine learning and computer vision are utilizing deep convolutional neural networks (CNNs) and biologically inspired vision architectures to resolve the debate between nativist and constructivist models of face perception.

A central question in computational developmental neuroscience is whether a neural network requires an explicit, pre-programmed “face module” to develop face selectivity, or whether face-selective units emerge spontaneously from domain-general visual training. Researchers have demonstrated that deep neural networks trained purely on generic, non-face object classification tasks (such as the ImageNet database) spontaneously develop internal artificial neurons that respond selectively to human and animal faces. These artificial “face units” emerge naturally in the intermediate layers of the networks as an optimal computational solution for categorizing visual space, mirroring the self-organization posited by the Interactive Specialization hypothesis.

Furthermore, computational neuroscientists have constructed dual-stream artificial neural architectures that explicitly model the two-process theory of infant vision. In these models, a simulated subcortical pathway—implemented as a low-resolution, low-pass visual filter operating across broad peripheral coordinates—is coupled with a high-resolution, deep cortical processing stream. In silico experiments demonstrate that when the subcortical stream is initialized with a simple, coarse three-blob bias (modeling CONSPEC), it reliably drives the camera sensors to orient toward faces in naturalistic video streams. This automated orienting funnels facial images into the cortical network, which rapidly develops robust representations of facial identity, expression, and gaze. These computational experiments demonstrate the mathematical viability of the two-process theory, proving that a minimal innate orienting bias is computationally sufficient to bootstrap adult-level facial expertise in a plastic learning system.

12.3 Contemporary High-Resolution Eye-Tracking and Neuroimaging Frontiers

Six decades after Robert Fantz peeked through a pinhole in a wooden looking chamber, the empirical methodologies of developmental psychology have been transformed by modern technological breakthroughs. Today’s developmental cognitive neuroscience laboratories utilize automated, contactless, high-speed infrared corneal-reflection eye-tracking systems operating at sampling rates between 300 and 1200 Hz. These systems calculate gaze position with spatial accuracy under 0.5 degrees of visual angle, allowing researchers to track micro-saccades, pupil dilation dynamics, and fixations across complex, dynamic scenes without requiring physical restraint of the infant.

Furthermore, contemporary research has moved beyond static schematic line drawings to evaluate the infant’s natural visual ecology. By outfitting human infants with lightweight, head-mounted “wearable baby-cameras” and mobile eye-trackers, researchers such as Linda Smith and Chen Yu have captured the visual world from the infant’s own anatomical perspective. These investigations reveal that the infant’s everyday visual environment is not a uniform distribution of objects, but is characterized by sudden, massive bursts of facial input. During the first six months of life, human faces occupy a substantial proportion of the infant’s visual field, appearing at close, optimized focal distances (15 to 30 cm) that match the low-pass spatial frequency filtering of the neonatal visual system.

Simultaneously, the integration of high-density EEG and wearable, wireless Functional Near-Infrared Spectroscopy (fNIRS) caps allows investigators to record cortical and subcortical hemodynamic and electrophysiological responses in real-time as infants interact with real, living caregivers in naturalistic environments. These advanced neuroimaging tools are mapping the developmental progression of early vision, documenting how neural synchrony emerges between parent and child during face-to-face communication.

The foundational insight generated by Robert L. Fantz—that the infant mind is active, discerning, and endogenously organized to prioritize pattern over chaos, and the human face over non-social noise—remains an enduring principle of developmental science. By recognizing that the human newborn looks with purpose and selectivity, Fantz provided psychology with a powerful methodology, laying the empirical foundations for our modern understanding of the social human brain.

Conclusion: Synthesizing the Fantzian Legacy

The trajectory of developmental psychology across the past seventy years reflects the enduring impact of Robert Fantz’s empirical discoveries. By challenging the philosophical dogmas of radical empiricism and behaviorist passivity, Fantz reclaimed the preverbal infant as a subject of rigorous scientific inquiry. His demonstration that newborn infants preferentially fixate on canonical facial arrangements over visually equivalent scrambled controls provided the empirical catalyst for the modern discipline of developmental cognitive neuroscience.

Today, the infant face preference is understood not as an isolated nativist module or a sensory artifact, but as a biological coordination between nature and nurture. The unlearned, subcortical orienting bias embodied by CONSPEC guarantees that the altricial infant will immediately gaze upon its caregivers, securing maternal investment, fostering intersubjective communication, and feeding visual experiences into an emerging, highly plastic cerebral cortex. This experience-expectant exposure drives the self-organization of CONLERN within the ventral occipitotemporal stream, culminating in the specialized, right-lateralized neural circuitry that characterizes adult facial perception. From early clinical screening for neurodevelopmental conditions to the design of artificial intelligence architectures, the conceptual paradigm pioneered by Fantz continues to shape contemporary inquiries into the nature of human cognition. By looking into the eyes of the newborn infant, Robert Fantz illuminated the origins of the human social mind.

References

• Banks, M. S., & Salapatek, P. (1981). Infant pattern vision: A new approach based on the contrast sensitivity function. Journal of Experimental Child Psychology, 31(1), 1–45. https://doi.org/10.1016/0022-0965(81)90002-3
• Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. Basic Books.
• Fantz, R. L. (1956). A method for studying early visual development. Perceptual and Motor Skills, 6(1), 13–15. https://doi.org/10.2466/pms.1956.6.1.13
• Fantz, R. L. (1958). Pattern vision in young infants. The Psychological Record, 8(2), 43–47. https://doi.org/10.1007/BF03393306
• Fantz, R. L. (1961). The origin of form perception. Science, 133(3464), 1568–1572. https://doi.org/10.1126/science.133.3464.1568
• Fantz, R. L. (1963). Pattern vision in newborn infants. Scientific American, 208(5), 84–97. https://doi.org/10.1038/scientificamerican0563-84
• Goren, C. C., Sarty, M., & Wu, P. Y. (1975). Visual following and pattern discrimination of face-like stimuli by newborn infants. Pediatrics, 56(4), 544–549. https://doi.org/10.1542/peds.56.4.544
• James, W. (1890). The principles of psychology. Henry Holt and Company.
• Johnson, M. H. (2005). Subcortical face processing. Nature Reviews Neuroscience, 6(10), 766–774. https://doi.org/10.1038/nrn1766
• Johnson, M. H., Dziurawiec, S., Ellis, H., & Morton, J. (1991). Newborns’ preferential tracking of face-like stimuli and its subsequent decline. Cognition, 40(1–2), 1–19. https://doi.org/10.1016/0010-0277(91)90045-6
• Jones, W., & Klin, A. (2013). Attention to eyes is present but in decline in 2–6-month-old infants later diagnosed with autism. Nature, 504(7480), 427–431. https://doi.org/10.1038/nature12715
• Kleiner, K. A., & Banks, M. S. (1987). Stimulus energy does not account for 2-month-olds’ face preferences. Journal of Experimental Psychology: Human Perception and Performance, 13(4), 594–600. https://doi.org/10.1037/0096-1523.13.4.594
• Maurer, D., & Young, R. (1983). Newborn’s following of visual stimuli. Child Development, 54(3), 748–755. https://doi.org/10.2307/1130062
• Morton, J., & Johnson, M. H. (1991). CONSPEC and CONLERN: A two-process theory of infant face recognition. Biology and Society, 6(4), 175–197. https://doi.org/10.1037/0033-295X.98.2.164
• Pascalis, O., de Haan, M., & Nelson, C. A. (2002). Is face processing species-specific during the first year of life? Science, 296(5571), 1321–1323. https://doi.org/10.1126/science.1070223
• Piaget, J. (1952). The origins of intelligence in children. International Universities Press.
• Pierce, K., Conant, D., Hazin, R., Stoner, R., & Desmond, J. (2011). Preference for geometric patterns early in life as a risk marker for autism. Archives of General Psychiatry, 68(1), 101–109. https://doi.org/10.1001/archgenpsychiatry.2010.166
• Simion, F., Valenza, E., Macchi Cassia, V., Turati, C., & Umiltà, C. (2002). Newborns’ preference for up-down asymmetrical configurations. Developmental Science, 5(4), 427–434. https://doi.org/10.1111/1467-7687.00237
• Trevarthen, C. (1979). Communication and cooperation in early infancy: A description of primary intersubjectivity. In M. Bullowa (Ed.), Before speech: The beginning of interpersonal communication (pp. 321–347). Cambridge University Press.
• Turati, C., Simion, F., Milani, I., & Umiltà, C. (2002). Newborns’ preference for faces: What is crucial? Cognition, 85(3), 245–274. https://doi.org/10.1016/S0010-0277(02)00107-1