Elsevier

Cognition

Volume 92, Issue 3, July 2004, Pages 305-327
Cognition

The ontogeny of face identity: I. Eight- to 21-week-old infants use internal and external face features in identity

https://doi.org/10.1016/j.cognition.2003.10.004Get rights and content

Abstract

A paradigm was designed to study how infants identify live faces. Eight- to 21-week-old infants were seated comfortably and were presented an adult female, dressed in a white laboratory coat and a white turtle neck sweater, until habituation ensued. The adult then left the room. One minute later either she or an identically garbed confederate returned. Looking time did not increase above habituation levels when the original experimenter returned, but increased substantially when the confederate entered the room. Furthermore, internal features alone could serve as the basis of face identity. Looking times of infants who had habituated to experimenters with masked outer features of hair, ears and neck also markedly increased when a second identically dressed experimenter returned. Identity in this instance could only be based on internal facial characteristics. A second study assessed the contributions of internal and external facial features to identity. Infants were habituated to an experimenter in a short wig. One minute later they saw her again, either in the same short wig or in a long one. Alternatively, they saw a second experimenter wearing either the short wig or the long one. Infants looked longer only to the stranger wearing the long wig. Very brief looking times occurred to the familiar adult, regardless of wig, and to the stranger wearing the familiar wig. This paradigm provides an approach to discover rules used by infants of different ages to process and identify adult faces and to establish the bases of face preference.

Introduction

Recognizing, identifying, and responding appropriately to different faces is a great cognitive achievement of the advanced mammals. Human adults can distinguish among thousands of faces, despite different angles of regard and expressions from occasion to occasion. The bearers of these faces can often be identified, as can their work, relational status and names (Carey, 1992). Affect can also be judged and responded to by either seeking more information from the individual, by approaching, or by fleeing. Appropriately, significant portions of the brain are dedicated to these particular tasks and to the integration of face orientation and expression (Allison et al., 1999, Kanwisher et al., 1997, Perrett et al., 1991). Moreover, severe and often specific deficits in face processing can arise from focal damage (Damasio, Tranel, & Damasio, 1990).

The complexity of face processing has directed some investigators to study it in children and infants to help elucidate its rules during neurologically and cognitively less complex epochs. Face processing ontogeny has been traced through two broad developmental phases of infancy and childhood; the former from birth till about 1 year of age (Johnson, Dziurawiec, Ellis, & Morton, 1991), the latter from the preschool years till mid-teens (Carey and Diamond, 1977, Diamond and Carey, 1986). Children and teenagers have been presented with either photographs or live models and required to indicate recognition of person or mood. In contrast, most studies of infant face perception have used schematic faces to learn about infant face processing characteristics (Goren et al., 1975, Maurer, 1983, Maurer and Barrera, 1981). Remarkably few infant face studies have used photographs of faces (e.g. Bushnell, 1982, Farroni et al., 2002), and only a handful have used live models to study the development of face perception (Bartrip et al., 2001, Bushnell et al., 1983, Pascalis et al., 1995).

Several factors underlie the choice of schematic faces to study infant face perception. First, the pioneering research of Goren et al. (1975) demonstrated that newborns actually looked at schematic faces and preferentially followed them with eyes and head. In contrast “scrambled” and other “impossible-faces” were not followed nearly as extensively (Johnson et al., 1991). These were important empirical advances. They demonstrated, first, that infants were responsive to, and tracked, at least face schemata. Second, they provided accessible and quantifiable behaviors to study the ontogeny of face perception. Moreover, stimulus characteristics could be controlled with precision, markedly reducing stimulus variability inherent in studies of real faces. Stimulus amplitude and phase could be manipulated and light contrast controlled in ways that are not possible with live faces. This degree of stimulus control was useful for determining which aspect(s) of the stimulus layout was critical to preferentially engage eye and head following (Mondloch et al., 1999).

Two-dimensional facial displays have also been employed to study development of face identity (Walton, Bower, & Bower, 1992). Television and computer presentation of faces has not yet been extensively utilized in newborns and infants during the first few months, although it has been successfully used in the study of mother–infant interactions and infant affect (Muir and Hains, 1993, Tronick, 1989). Studies featuring colored slide presentations of faces have been consistent with findings from the schematic face studies (Bushnell, 1982), although they place the initiation of infant face perception and identity at circa 5 weeks of age, somewhat earlier than the 2 months suggested by the schematic studies (Caron, Caron, Caldwell, & Weiss, 1973).

Live faces are the third class of stimuli used to investigate the ontogeny of face perception (Carpenter, 1973). Studies using live faces have generally pitted the familiar mother against a stranger to determine the age at which face recognition can be claimed and the processes underlying identity. These studies have provided strong evidence for maternal preference, even in newborns (Bushnell et al., 1983, Pascalis et al., 1995). Yet preference does not withstand masking the external features of the face. Maternal preference simply vanishes. According to Bartrip et al. (2001) maternal preferences withstand masking hair and other peripheral features, possibly starting at about 5 weeks, although this appears to be transitory.

Some agreement has emerged from these different strategies. As the stimulus material more closely approached the layout of real faces, earlier identity onsets were obtained, and empirically based suggestions were made as to underlying mechanisms. Morton and Johnson (1991) have suggested that at birth, there is available a subcortical mechanism for face-following that utilizes internal facial features, including, possibly, frontal views of the eyes (Farroni et al., 2002). The ‘following-system’ is accompanied by face recognition and identity systems. Within days of birth, infants prefer their mothers to similar looking strangers when both are presented in normal frontal view. Yet, it may not be until 5–8 weeks of age that infants prefer their mother to a stranger when their external facial features are occluded. These data are difficult to interpret statistically and because the preference for the undisguised mother also broke down. In fact, at 135 days of age, the stranger was preferred. We do not know if the problem is perceptual or motivational (Bartrip et al., 2001, Pascalis et al., 1995).

Despite its advantages of simplicity and experimental control, the use of schematic faces raises further concerns. The basic theoretical assumption that face schemata, even one as simple as three spots within an oval, engages the mechanisms that have been selected for over phylogenetic time to process actual live faces has not been rigorously evaluated. This concern is heightened by the ability of newborns to distinguish the mother's face from a stranger's. In contrast, preference for ‘normal’ over ‘scrambled’ schematic faces presented simultaneously does not emerge until about 2 months of age. These asymmetries place constraints on the schematic face as an approximation of the normal one.

Maurer (1985) has raised a second empirical concern. The search pattern of actual faces until about 4–6 weeks of age is characterized by the infant scanning the external face, while paying scant attention to internal features (although newborns look longer at a photograph of a woman in full face with forward eye gaze over the identical photo with displaced gaze; Farroni et al., 2002). In contrast to scanning patterns of live faces, the internal features of schematic faces receive considerable attention from the start. This may reflect the relative impoverishment of the external features, or, possibly, the prominence of the internal features of the schematic faces. Regardless, central utilization of information may differ as a result of scanning pattern differences.

The inanimate nature of schematic faces poses a third concern. In general, schematic faces do not induce the sustained affective changes that are normally elicited by live faces starting at about 2 months of age (Ahrens, 1954, Blass and Camp, 2001). Live engaging stimuli, that provide the bases of normal face identity and preference, should be used to identify the circumstances under which affective changes are elicited, and establish linkage between affective change and looking at novel or familiar faces (Blass & Camp, 2001). In fact, movement is a powerful determinant of infant perception; the use of moving over stationary objects has provided substantial insights into the nature of object perception, for example (Spelke, Breinlinger, Macomber, & Jacobson, 1992).

The preference paradigm also has certain limits for revealing identity. These constraints include infant age, gender, activity, affect at the time of testing and novelty, among others. Unlike newborns, for whom the mother's face was preferred over the strangers by both males and females, only male infants looked longer at the mother between 19 and 65 days of age, but not thereafter (Bartrip et al., 2001). Female infants in this age range never preferred the mother in full face over a stranger, looking equally at both women. In fact, they actually looked longer at strangers beginning at 145–155 days of age. Furthermore, when presented with only the internal features of mothers and strangers, female infants looked longer at the mother over the stranger only during the time frame of 35–40 days. Neither mother nor stranger internal faces were preferred otherwise, further restraining conclusions concerning the development of facial identity. Thus, face preference is strongly influenced by factors that are seemingly outside the domain of face processing per se. These factors constrain utilizing the preference paradigm in the study of face identity.

Issues also remain unresolved concerning internal face representation. The inability of newborns and 1-month-olds to respond to internal features of the face other than by tracking (Morton & Johnson, 1991) has been questioned. Farroni, Cassia, Turati, and Simion (2000) have reported that changing the ‘eyes’ of a three spot schematic face from squares to diamonds is detected by newborns. As indicated, newborns prefer to look at photographs of a female “looking” directly at the infant over identical photos with eyes averted (Farroni et al., 2002). Moreover, newborns imitate complex facial expressions, such as sadness and happiness (Field et al., 1983, Field et al., 1982), suggesting responsiveness to animated facial patterns based on internal features. Newborns also attend to and imitate specific facial features such as an open mouth and extended tongue and may possibly represent these features (Meltzoff and Moore, 1977, Meltzoff and Moore, 1984). The representational capability certainly appears to be in place by 1 month of age when a particular motion provided by a particular adult is replayed with fidelity after a delay of hours in the absence of adult modeling.

Recent studies by Slater et al. (2000) on infant perception of face attractiveness have demonstrated that face photographs judged attractive by adults are looked at longer by newborns, and this is determined by internal facial features. Thus, newborns are broadly responsive to internal facial features, but their ability to use them for identity purposes has not been determined. Although the use of colored slides in identity and discrimination studies is an important step towards aligning experimental stimuli closer to natural stimuli, photographic displays continue to differ from natural faces in at least three important regards. First, they are two- rather than three-dimensional. Second, because they are static, they may not adequately display internal animation or affect. Third, they cannot socially engage the infant.

Our own studies of 2- to 12-week-olds, using live experimenters (Blass and Camp, 2001, Zeifman et al., 1996), also support attention to internal facial features, especially the eyes, starting at about 1 month of age. Zeifman et al. (1996) quieted crying 4-week-old infants by having them taste a sucrose solution during eye contact. Crying infants who tasted sucrose in the absence of eye-to-eye contact did not calm, even though the infants had a full view of the experimenter's face. In this regard, a number of the infants repeatedly raised their heads as if to establish eye contact with the experimenter, whose gaze rested on the infant's forehead. The active seeking of eye contact emphasizes the importance of mutual gaze within the motivational sphere of arresting crying and, therefore, conserving energy. These findings accord with reports on the onset, at 4 weeks of age, of spontaneous infant–mother eye contact in the home setting (Wolff, 1963, Wolff, 1987).

Blass and Camp, 2001, Blass and Camp, 2003 have demonstrated face preference in 6- to 12-week-old infants based on state and on the nature of the interactions between experimenter and infant. Eye contact was necessary for preference to be formed for experimenters with whom infants had interacted for only 3.5 min. This implies the availability of a fast acting face identification system, a conclusion that had also been reached by deSchonen and Mathivet (1989) a number of years ago.

The paradigm developed by Blass and Camp (2001) may help resolve some of the outstanding issues identified above and, thereby, help further elucidate the determinants of face processing during infancy. Live interactive faces are used. This provides the three-dimensional animated configuration that infants normally encounter when looking at a person. Second, animated live faces optimally attract infant attention. Third, the face is ideally suited, one might argue selected, to elicit social affective behaviors (Ahrens, 1954). Indeed, positive affective actions such as smiles and “coos” by 9- to 12-week-olds were the rule during face-to-face interactions with experimenters who sat facing the infants with relaxed, smiling faces (Blass & Camp, 2001). Accordingly, the current study has three goals. The first is to develop an experimental paradigm for the purpose of studying the ontogeny of face identity in which live faces are presented to infants 8–21 weeks of age. Second, within this paradigm, we determine whether such infants could utilize internal facial features alone to identify and remember an individual. Third, we further assess the contribution of internal and external features to face identity when each dimension could be independently manipulated.

The approach uses the well established and accessible habituation paradigm to study the identity of new faces by infants. In following Bushnell's pioneering studies, we precisely mask particular characteristics of live faces. Segregation of features was achieved in the present study by covering the hair, thereby compelling infants to utilize the characteristics of the internal face. Although Slater et al. (2000) have demonstrated that internal features can be attended to by newborns, it remains open whether internal features can be used in face identity and discrimination. We addressed this issue directly by using wigs and different experimenters to identify the external and internal features that could be preferentially used in identity. The paradigm is straightforward: infants are habituated to a living face, in full view or with external features masked, and then re-exposed to the same or another face under different conditions. We developed the paradigm in infants 8–21 weeks of age because of ease of experimentation. Future studies will extend this paradigm to younger and older infants to test hypotheses generated from these and others’ studies.

We now report that these goals have been successfully realized. Eight- to 21-week-old infants readily identify faces when studied in a habituation paradigm. Second, they can identify adults on the basis of internal facial features alone. Third, conditions are specified in which infants can successfully utilize either internal or external conditions in identity tasks. Fourth, infants are proactive in gathering information concerning adult faces and actions during brief encounters with strangers.

Section snippets

Participants

A total of 123 infants 8–21 weeks of age were recruited from the Amherst, MA area by telephone contact. This age range was selected mainly out of pragmatic considerations. Because we were developing new experimental paradigms, we wanted to work within an age range that had proven easy to study. Younger infants, in our experience, have a greater tendency to cry; older infants are more distractible. Twenty-two infants were lost through crying and an additional five through experimental error. The

Experiment 1

Experiment 1 grows out of the confluence of two issues. First, could a standard habituation paradigm be adapted to study face identity in 2- to 5-month-old infants? Second, could this paradigm evaluate whether infants could use an adult's internal facial features exclusively to identify her, remember her in her absence, and distinguish her from another identically attired adult? This goal was achieved by exposing the infant to an experimenter whose outer features were covered and, after

Experiment 2

Although Experiment 1 demonstrated that 8- to 21-week-old infants can perceive the inner face, store that representation, use it to render that face familiar, and distinguish it from other internal faces, it did not reveal whether this information is utilized during the normal course of events of identifying and remembering full faces. Experiment 2 is a first effort at assessing face identification processes used by infants in a more natural situation of seeing an individual and responding to

General discussion

Four new findings have emerged from these studies. (1) The classic habituation paradigm has now been extended to the study of face perception using live models. (2) Infants as young as 8 weeks of age can learn about, store, and utilize information arising from internal facial features, presented alone, to identify individual faces. (3) When encountering a new full face, 8- to 19-week-old infants can learn about and utilize both internal and external facial features in the identity process. The

Acknowledgements

This research was supported by NIMH grant MH 51705 to E.M.B. We gratefully acknowledge the spirit and determination of the mothers, infants and undergraduate research assistants who participated in this research. Professor Rachel Keen's insightful comments on an earlier draft of this manuscript are acknowledged with appreciation.

References (50)

  • J Bartrip et al.

    Responses to mother's face in 3-week to 5-month-old infants

    British Journal of Developmental Psychology

    (2001)
  • E.M Blass et al.

    The ontogeny of face recognition: eye contact and sweet taste induce face preference in 9- and 12-week-old human infants

    Developmental Psychology

    (2001)
  • Blass, E. M., & Camp, C. A (2003). Biological bases of face preference in six-week-old infants. Developmental Science,...
  • Blass, E. M., Lumeng, J. L., & Patil, N (in press). Gaze, preference and feeding determinants in human infants. In R....
  • I.W.R Bushnell et al.

    Neonatal recognition of mother's face

    British Journal of Developmental Psychology

    (1983)
  • S Carey

    Becoming a face expert

  • S Carey et al.

    From piecemeal to configurational representation of faces

    Science

    (1977)
  • A Caron et al.

    Infant perception of the structural properties of the face

    Developmental Psychology

    (1973)
  • G.C Carpenter

    Visual regard of moving and stationary faces in early infancy

    Merrill-Palmer Quarterly

    (1973)
  • R.K Clifton

    The relation of infant cardiac responding to behavioral state and motor activity

  • A.R Damasio et al.

    Face agnosia and the neural substrates of memory

    Annual Review of Neuroscience

    (1990)
  • S deSchonen et al.

    First come, first served: a scenario about the development of specialization in face recognition during infancy

    Cahiers de Psychologie Cognitive

    (1989)
  • R Diamond et al.

    Why faces are and are not special: an effect of expertise

    Journal of Experimental Psychology: General

    (1986)
  • J Dirks et al.

    Infants’ perception of similarity between live people and their photographs

    Child Development

    (1977)
  • H Egeth

    Parallel versus serial processes in multidimensional stimulus discrimination

    Perception and Psychophysics

    (1966)
  • Cited by (14)

    • Infant capacities related to building internal working models of attachment figures: A theoretical and empirical review

      2015, Developmental Review
      Citation Excerpt :

      Research indicates that infants as young as 1 month old can discriminate between two unfamiliar adult faces (de Haan, Johnson, Maurer, & Perrett, 2001). Whether faces are presented as static images or live models, infants between 3 and 5 months are able to discriminate between unfamiliar adults and recognize familiar adults (Bahrick & Newell, 2008; Blass & Camp, 2004; Otsuka et al., 2009). By 6 months of age, infants are able to recognize familiar faces that were presented and tested at differing angles (Pascalis, de Haan, Nelson, & de Schonen, 1998) and with differing emotional expressions (Bornstein & Arterberry, 2003).

    • How infant perceive faces

      2009, Evolution Psychiatrique
    • Facial experience during the first year

      2008, Infant Behavior and Development
    • How face specialization emerges in the first months of life

      2007, Progress in Brain Research
      Citation Excerpt :

      It rests on recognition memory competencies and differs from that of face detection, which refers to the capacity to perceptually discriminate between a face and a non-face visual object. The complexity of face processing in adults has directed some investigators to studying face recognition in preschool-aged children (e.g., Carey and Diamond, 1977; Diamond and Carey, 1986; Sangrigoli and de Schonen, 2004) and infants (e.g., Pascalis et al., 1998; Blass and Camp, 2004) to elucidate its rules in neurologically and cognitively less complex epochs. The following section of the present chapter focuses on newborns’ ability to recognize individual faces.

    View all citing articles on Scopus
    View full text