Origins of neonates’ face preference: Domain-specific or domain-general accounts

Faces carry rich social information (identity, emotion, intention) and are used to identify people and communicate meaning [1]. Humans preferentially attend to faces even from birth. Fantz [2] famously showed that neonates spend more time looking at simple face-like patterns than at control patterns. Remarkably, studies in utero find that fetuses around 7–9 months’ gestation turn more often toward upright face-like dot configurations than toward inverted ones [3, 4]. Face-oriented preferences also appear across species. For example, infant monkeys, visually naïve tortoise hatchlings, and newly hatched domestic chicks all show spontaneous attraction to face-like stimuli [5-7 ]. These findings raise the question: why do neonates preferentially orient to faces

Two dominant accounts have been offered to explain neonates’ face preferences. The domain- specific view holds that infants are born with a hardwired face-detection system (sometimes called CONSPEC) subserved by subcortical structures (superior colliculus, pulvinar) [8]. In this account, faces form a special stimulus category with innate behavioral relevance. For instance, Johnson[8] and Morton’s[9] two-process model argues that a subcortical route (CONSPEC) initially orients infants to faces, while a cortical route (CONLERN) later refines face recognition with experience. Consistent with this, infants exhibit early face biases that do not require visual learning. For example, Sugita[5] reported that infant monkeys reared without seeing faces still preferred face stimuli in later months. Likewise, newborn humans prefer naturalistic faces with correct (positive) contrast polarity [10]. Finally, recent neural data show face-selective neurons in a high-level visual area of face-naïve chicks, supporting a primitive face detector [7]. Together, these findings suggest that faces have a privileged status in the newborn visual system.

By contrast, the domain-general view attributes neonates’ face bias to basic visual features that the immature visual system detects well. For example, newborns are highly sensitive to low spatial frequencies and to high-contrast patterns [1, 11]. Among proposed constraints, a top-heavy bias (more elements in the upper visual field) is especially important: faces tend to be top-heavy, and even abstract patterns with upper-heavy layouts draw neonates’ gaze [12, 13]. Indeed, Cassia et al.[14] showed that neonates prefer any top-heavy pattern, even non-face patterns. This account also notes that subcortical pathways (superior colliculus, pulvinar) specialize in low- frequency, high-contrast input and have a natural bias for stimuli in the upper visual field [15]. Because typical faces are high-contrast, coarse (low-frequency), and top-heavy, they strongly activate these pathways. Crucially, the domain- general perspective maintains that these biases are not face-specific: faces simply happen to match the newborns’ visual preferences [16]. Thus, no innate face template is required; instead, early face preference emerges from general perceptual constraints and experience.

In summary, both views agree that neonates’ face preferences reflect innate, evolutionarily- derived visual capacities. However, they differ in logic: the domain-specific account is “top-down” (faces are intrinsically meaningful stimuli), whereas the domain-general account is “bottom-up” (faces merely fit general constraints). Conceptually, the debate hinges on whether faces are a special stimulus category for neonates. At the neural level, the question is whether specialized circuits for faces exist at birth, or whether face preference arises from general visual processing pathways. Both accounts highlight the superior colliculus- pulvinar pathway as important for early face orienting, but it is not known if this pathway is truly face-selective. Clarifying this is key to deciding between the models. One promising approach is to record directly from subcortical neurons in newborn primates (as Kobylkov et al. did in chicks) while presenting faces and other stimuli.