The development of facial emotion recognition: The role of configural information

Abstract

The development of children’s ability to recognize facial emotions and the role of configural information in this development were investigated. In the study, 100 5-, 7-, 9-, and 11-year-olds and 26 adults needed to recognize the emotion displayed by upright and upside-down faces. The same participants needed to recognize the emotion displayed by the top half of an upright or upside-down face that was or was not aligned with a bottom half that displayed another emotion. The results showed that the ability to recognize facial emotion develops with age, with a developmental course that depends on the emotion to be recognized. Moreover, children at all ages and adults exhibited both an inversion effect and a composite effect, suggesting that children rely on configural information to recognize facial emotions.

Introduction

Processing of facial emotions emerges early (e.g., Barrera and Maurer, 1981, Walker-Andrews, 1997), but full proficiency seems not to be acquired before 10 years of age. Although preschoolers can label facial emotions at above chance levels (e.g., Markham and Adams, 1992, Russell and Widen, 2002, Widen and Russell, 2003), they are substantially less accurate than adults. The purpose of our study was to provide new evidence on development change in recognition of facial emotions during childhood.

Few studies have investigated the developmental course of facial emotion recognition, contrary to face recognition abilities, and the results are often inconsistent, mainly because of the great variety of methods used. A study by Bruce et al. (2000) suggested that development of facial emotion recognition depends on task demands. When children needed to point to which of two faces was happy, sad, angry, or surprised, they achieved nearly perfect accuracy by 6 years of age. However, when they needed to select which of two emotional faces expressed the same emotion as a third face, a good accuracy level was not reached until 10 years of age. In a similar study in which children needed to match an emotional photograph to one of four possibilities (neutral, surprise, happiness, or disgust), Mondloch, Geldart, Maurer, and Le Grand (2003) reported an increase in accuracy between 6 and 8 years of age, when performance reached the adult level. In a study by Kolb, Wilson, and Taylor (1992), children and adults were shown a single emotional photograph or a cartoon depicting an emotional situation and then needed to select from a panel of six different emotional photographs (happiness, sadness, fear, anger, disgust, and surprise) the face that expressed the same or correct emotion. Recognition of facial emotions improved between 6 and 8 years and between 8 and 10 years of age, depending on the task, and improved again between age 14–15 years of age and adulthood. There is also evidence that the developmental pattern is not uniform across emotions. Expressions of happiness and sadness seem to be correctly categorized earlier than those of fear and disgust (Boyatzis et al., 1993, Camras and Allison, 1985, Gosselin, 1995; see also Gosselin, 2005, Gosselin and Larocque, 2000). The developmental pattern for anger is less clear, with some results indicating a pattern similar to that for happiness and sadness (e.g., Gosselin, 1995) and others indicating that anger was less well categorized than happiness and sadness but also less well categorized than fear and disgust (e.g., Boyatzis et al., 1993).

Despite task- and emotion-related differences, performance generally improves with age. One hypothesis developed to explain improvement in face recognition is that adults are more likely than children to rely on the configural properties of faces (e.g., Leder and Bruce, 1998, Tanaka and Farah, 1993, Thompson, 1980, Yin, 1969). Three types of configural information are distinguished (for a review, see Maurer, Le Grand, & Mondloch, 2002): first-order relations (relative position of features, e.g., the eyes above the nose), second-order relations (fine spatial information, e.g., distance between features), and holistic information (features are perceived as a unique gestalt). Configural processing is distinguished from featural, componential, or local processing, where features are processed in a piecemeal or analytic way. Carey and Diamond (1977) suggested that children do not use configural information until they are 10 years of age. Children under age 10 would tend to base their judgments on local information such as features or paraphernalia. These authors noted that children under age 10 are highly sensitive to paraphernalia when recognizing unfamiliar faces. Moreover, they found no inversion effect (i.e., children were not impaired by presenting a face upside down), whereas such an effect is known to be strong in adults (Yin, 1969). The inversion effect is considered to demonstrate the weight of configural information; by modifying the relative position of features (e.g., the eyes are above the nose in an upright face, whereas they are below the nose in an upside-down face), inversion of a face interferes with configural processing (Maurer et al., 2002).

However, Carey and Diamond’s (1977) hypothesis has been challenged by studies demonstrating that children under 10 years of age can process faces configurally (Baenninger, 1994, Carey and Diamond, 1994, Flin, 1985, Freire and Lee, 2001, Tanaka et al., 1998). Notably, an inversion effect was reported at younger than 10 years of age. Flin (1985) found that inversion altered children’s performance as young as 7 years in the same way as older children and adults. In Brace and colleagues’ (2001) study, children as young as 5 years showed an inversion effect, but 2- to 4-year-olds did not. Recently, Sangrigoli and de Schonen (2004) showed an inversion effect in 3-year-olds. Even newborns and infants are sensitive to face orientation (Gallay et al., 2006, Slater et al., 2000, Turati et al., 2004). Carey and Diamond (1994) themselves reported an increasing inversion effect from 6 to 10 years of age, suggesting a development of the ability to process configural information.

Configural processing of facial information by children was also observed with paradigms used specifically to provide evidence for holistic processing of faces, namely, the composite face paradigm and the whole-part paradigm. In the composite face paradigm, a composite stimulus is made up by joining the top half of a face with the bottom half of another face. Adults are slower and less accurate to recognize either half part when the top and bottom parts are vertically aligned (creating a new face stimulus) than when the same top and bottom parts are misaligned. A similar composite effect (i.e., the recognition of a person from a half-part of the face is interfered by the alignment of a counter-part from another face) was observed in 4-year-olds and adults (De Heering, Houthuys, & Rossion, 2007; for a similar observation from 6- to 10-year-olds, see Carey & Diamond, 1994). In the whole-part paradigm, participants need to recognize face features either embedded in the whole face stimulus or presented in isolation. A whole-part advantage effect (i.e., features are better recognized in the context of a normal face than when isolated) was reported in 4- and 6-year-olds, again of similar amplitude to older children (Pellicano and Rhodes, 2003, Tanaka et al., 1998). Such effects have also been demonstrated in infants. Notably, Cashon and Cohen, 2003, Cohen and Cashon, 2001 reported that 4-month-olds look longer at a “switched” (or composite) face, made up of the internal features of a familiar face and the external features of another familiar face, than at each whole familiar face. Thus, the composite face was perceived as a new face by infants despite its being built up with only familiar parts. This observation was extended to 8-month-olds by switching single features rather than all internal features; infants reacted when only the eyes, nose, or mouth of a familiar face was put into another familiar face (Schwarzer & Zauner, 2003). Thus, there is now sufficient evidence demonstrating the early ability of children to process configural information.

Little is known about the role of configural information in the development of the ability to recognize facial emotions. This topic of investigation is quite important because configural information also plays a significant role in facial emotion recognition by adults. Calder et al., 2000, Calder and Jansen, 2005 observed a composite effect in emotion recognition similar to the one reported by Young, Hellawell, and Hay (1987) in face recognition; that is, when the top and bottom halves of a composite face depict different emotions, recognition of the emotion in either half is slower and less accurate than when the composite face is inverted or the two halves are offset laterally. Thus, it can be hypothesized, as for face recognition, that the development of the ability to process configural properties of faces also underlies the development of the ability to process facial emotions.

The main aim of the current research concerned the possible role of configural information in the development of recognition of facial emotions. Configural information plays an important role in adults’ recognition of facial emotions (e.g., Calder and Jansen, 2005, Calder et al., 2000), but the role of configural information in children’s recognition of facial emotions is unknown. In particular, testing children’s sensitivity to facial emotion orientation and composite emotion would help to reveal the processes that underlie developmental change in recognition of facial emotions. Testing both the inversion and the composite effects is important because many authors have suggested that the various paradigms do not tap into the same type of configural information or are differentially sensitive to this information (Baudouin and Humphreys, 2006, Calder and Jansen, 2005, Maurer et al., 2002, Mondloch et al., 2002). Consequently, testing configural effects provides a more complete view of the role of configural information in children; the inversion effect informs us about the consequences of disturbing the processing of configural information by altering first-order relations, whereas the composite effect tests more specifically the role of holistic information in the recognition of facial emotion.

Children of four ages and adults participated in two tasks. In the first task, they were asked to recognize six different emotional categories: happiness, anger, disgust, fear, sadness, and neutrality. The faces were upright in one session and upside down in another session. This task allowed us to study the development of the ability to process each emotion and informed us about the presence of an inversion effect in emotion recognition at each age. In the second task, participants were asked to recognize the emotion depicted in the top half of faces when, in each case, the bottom half was the same person’s face displaying another emotion. The two halves were either aligned or shifted, and they were presented either upright or upside down. In adults, alignment in upright faces creates a new emotional configuration (the composite effect) that interferes with the recognition of the emotion displayed by each half; recognition is slower than it is in the shifted and upside-down conditions (Calder et al., 2000). This task allowed us to find out whether children would exhibit the same composite effect as do adults. More precisely, it was informative about whether holistic information would interfere in the processing of facial emotion to a similar extent in children and adults.