Infants use knowledge of emotions to augment face perception: Evidence of top-down modulation of perception early in life

Abstract

While top-down modulation is believed to be central to adult perception, the developmental origins of this ability are unclear. Here, we present a direct, behavioral investigation of top-down modulation of perception in infancy using emotional face perception as a test case. We investigated whether 9-month-olds can modulate their face perception based on predictive, auditory emotional cues without any training or familiarization procedure. Infants first heard a 3-second emotional vocal sound (happy/angry) while their gaze was held in the center of the screen. Then, they were presented with a pair of emotional and neutral faces images without any audio sound. The faces were small (4.70° × 5.80°) and presented in randomized locations outside their focus of attention. We measured the initial latency to shift gaze to look at a congruent emotional face as an index of infants' pre-attentive perception of these faces. We found that infants' face perception was augmented by preceding emotional cues: They were faster to look at the emotional face after hearing an emotionally congruent sound than an incongruent one. Moreover, the emotional sounds boosted perception of congruent faces 200 ms after the onset of the faces. These top-down effects were robust for both happy and angry emotions, indicating a flexible and active control of perception based on different top-down cues. A control study further supported the view that the Congruency effect is due to a top-down influence on face perception rather than a rapid matching of cross-modal emotional signals. Together, these findings demonstrate that top-down modulation of perception is already quite sophisticated early in development. Raw data is available on Github (https://github.com/naiqixiao/CuedEmotion.git).

Keywords: Cross-modal perception; Emotion; Face perception; Infants; Perception; Perceptual development; Top-down.

Fig. 1.

Schematic of infant perception and eye gaze shifting in this task. Perception is presented as the relative salience of the two face images (i.e., image transparency); eye gaze is the yellow dot. Left: Bottom-up input is weak when infants’ attention is held at screen center with no differences in perception between the two faces. Center: If infants are capable of top-down modulation of face perception, preceding congruent emotional vocalizations will boost perception of a congruent emotional face. This better perception of the congruent faces results in attentional orienting (initial looking) to that face. Right: Incongruent emotional sounds do not augment perception of the emotional face and will not guide infants’ initial looking. Top-down modulation of perception will be manifested by faster initial looking latency in the congruent sound condition (center panel) than that in the incongruent condition (right panel).

Fig. 2.

A representative trial for Expt. 1 (Left) and infants’ initial looking latency to the emotional face (Right). Error bars represent unit standard errors. The overlaid shapes represent the distribution of individual response for each condition. The asterisks indicate a significant difference between the Congruent and Incongruent conditions (Tukey’s HSD, FDR corrected p < .050).

Fig. 3.

Demonstration of the size and layout of face images (Upper: happy face, Lower: angry face) in previous cross-modal matching studies as summarized in Table 1 (Left, image width: 15.02° visual angle) and those in the current study (Right, image width: 4.70° visual angle).

Fig. 4.

Mean proportional looking to the emotional faces in the Congruent and Incongruent conditions (Left) and mean proportional looking to the emotional for each time window (Right). Error bars represent unit standard errors. The asterisks indicate a significant difference between the Congruent and Incongruent conditions (FDR corrected p < .050). The overlaid shapes represent the distribution of individual response. The dash line represents the chance level (50%).

Fig. 5.

A representative trial for Expt. 2 (Left) and infants’ initial looking latency to the emotional face (Right). The overlaid shapes represent the distribution of individual response for each condition. Error bars represent unit standard errors.

Fig. 6.

Mean proportional looking to the emotional faces in the Congruent and Incongruent conditions (Left) and mean proportional looking to the emotional for each time window (Right). Error bars represent unit standard errors. The overlaid shapes represent the distribution of individual response. The dash line represents the chance level (50%).

Fig. 7.

Mean latency to look at the happy and angry faces (Upper Left), mean proportional looking to the emotional faces (Upper Right), and mean proportional looking to the emotional face for each time window (Lower). Error bars represent unit standard errors. The asterisks indicate a significant difference between the Congruent and Incongruent conditions (FDR corrected p < .050). The overlaid shapes represent the distribution of individual response. The dash line represents the chance level (50%).