Neonatal imitation and early social experience predict gaze following abilities in infant monkeys

Individuals vary in their social skills and motivation, the causes of which remain largely unknown. Here we investigated whether an individual’s propensity to interact with others measured within days after birth, and differences in infants’ early social environment, may predict a later social skill. Specifically, we tested whether neonatal imitation—newborns’ capacity to match modelled actions—and social experience in the first months of life predict gaze following (directing attention to locations where others look), in infant macaques (Macaca mulatta; n = 119). Facial gesture imitation in the first week of life predicted gaze following at 7 months of age. Imitators were better at gaze following than non-imitators, suggesting neonatal imitation may be an early marker predicting socio-cognitive functioning. In addition, infants with rich social environments outperformed infants with less socialization, suggesting early social experiences also support the development of infants’ gaze following competence. The present study offers compelling evidence that an individual difference present from birth predicts a functional social cognitive skill in later infancy. In addition, this foundational skill—gaze following—is plastic, and can be improved through social interactions, providing infants with a strong foundation for later social interaction and learning.


Method Subjects
On the day of birth, infants were separated from their mothers and raised in a primate nursery.
Infants could see and hear other infants, but remained physically separated. Incubators contained one soft toy, two hard toys, three fleece blankets, and a surrogate made from an upright, flexible polypropylene cylinder covered in fleece. A two-ounce bottle containing Similac formula was attached to the surrogate for the first week of life. Toys and fleece blankets were replaced daily. At eight days of age, the plastic incubator top was replaced with a wire top, and the polypropylene cylinder surrogate was replaced with a hanging, fleece-covered surrogate pouch, into which the infant could climb. Two-ounce bottles of formula were attached to the side of the cage. Additionally, cages contained four hanging enrichment toys, two hard toys, and three blankets. Hard toys and fleece blankets were replaced daily. At fifteen days of age, infants were moved from the incubator room into an adjacent room lined with 61 × 61 × 76 cm cages. Infants were placed into one of the cages, along with three blankets, three floor toys, two hanging toys, and a hanging surrogate. The surrogates were changed weekly, and enrichment toys were rotated daily. At fifteen days of age infants also began receiving six minutes of handling by a member of the nursery staff on weekday mornings. The monkeys were rotated so that staff members did not hold the same monkey more than once every four days. On day sixteen infants began drinking formula through 500 ml bottles with metal sipper tube.
Infants were assigned to rearing conditions on their day of birth. Peer-reared (highsocialization) infants were raised in groups of three to five similar-aged peers in a 71 × 81 × 152 cm cage. High-socialization infants were placed into their peer groups when the youngest member of the group reached 37 days of age. Until then they remained in their individual cages. Each peer cage contained three floor toys, one fleece blanket, and one hanging surrogate per infant, in addition to two hanging toys. When each infant in the peer group turned four months of age, the single hanging surrogate was removed from the cage. Thus,

S3
when the youngest infant turned four months of age, no surrogates remained in the cage.
Infants remained in their peer groups until they were weaned around seven to eight months of age (after gaze following testing was completed).
Unlike peer-reared infants, surrogate-reared (low-socialization) infants were individually housed in 61 × 61 × 76 cm cages throughout their stay in the nursery. Cages contained three fleece blankets, three floor enrichment toys, two hanging enrichment toys, and a hanging surrogate. Low-socialization reared infants had access to their surrogate throughout their stay in the nursery. To facilitate socialization, low-socialization infants were assigned to playgroups composed of three to four similar-aged infants. Once the youngest member of the playgroup reached 37 days of age, infants in were placed into a 71 × 81 × 152 cm cage with their playgroup for two hours a day, five days a week. Only half of the play cage was accessible to the infants until the youngest infant turned two months old. The play cage contained enrichment toys, one hanging surrogate, three to four fleece blankets, and one water bottle. Once the youngest infant turned two months old, the entire cage was made accessible, and another hanging surrogate and water bottle was added to the cage.
For the first thirty days of life, infants were fed bottles of Similac formula every two hours from 8am to 8pm. After 30 days, infants were fed at 8am and 4pm, Monday through Friday, and at 7:30am and 1:30pm on weekends. The amount of Similac provided to infants varied by infants' age and weight. At 16 days of age, Purina LabDiet 5045 High Protein Monkey Diet chow was introduced to their diet, in addition to Similac. The feeding regimen eventually transitioned from a Similac-only diet, to a chow-only diet by six months of age. Infants also received a daily afternoon snack of nuts, fruit, grains, or seeds beginning at two months of age.
Water was available ad libitum from a lixet system. In the incubator room and the large cage room, lights were on from 7:00 to 21:00.

Details of the Neonatal Imitation Procedure and Classification
Infants were tested three times a day, every other day, in the first week of life (days 1-2, 3-4, 5-6, and 7-8), for up to four days. There was at least an hour between each test session. A demonstrator presented infants with two stimuli, one during each session, at a distance of S4 approximately 30 cm at eye-level with the infant: a lipsmacking gesture (rapid opening and closing of the mouth) and a nonsocial control condition (a white plastic disk with orthogonal stripes-which were either black/red or green/yellow-slowly rotated clockwise and counterclockwise). Each stimulus type was presented once a day to infants; the order of stimulus presentations remained the same for each infant but was randomized between infants. In each test session, one experimenter held the infant, a second experimenter-the demonstratorserved as the source of the stimuli, and a third experimenter was the time-keeper who ensured stimuli were presented for appropriate lengths. All sessions were videotaped. Individual demonstrators were randomly assigned to conditions but remained consistent across days within each infant.
At the beginning of a trial, a 40 sec baseline was conducted, in which the demonstrator displayed a calm, neutral facial expression (or the still disk in control). The demonstrator then displayed a facial movement (lipsmacking) or rotated the disk for 20 seconds, followed by a still/neutral facial expression (still disk in control) period for 20 sec. This movement-still face sequence was repeated three times.
Infants' mouth movements were coded off-line, frame-by-frame (30 frames per second) from video. Lipsmacking was operationally defined as a high frequency opening and closing of the mouth without sound production in which the lips were required to part and then rejoin within 2 seconds. Observers were blind to the stimulus. Inter-observer agreement was high, r = .950, p < .001, n = 66.

Categorical Classification: Imitators and Non-Imitators. Lipsmacking gestures rates (per 60 sec)
from Baseline and Stimulus were averaged across all test days in each condition. Infants were classified as imitators and non-imitators. Infants were LPS imitators if they produced an increase in LPS from the baseline to the stimulus period in the LPS condition (matching the model), to a greater extent than the increase in LPS from the baseline to the stimulus period in the control condition, averaged across days [1][2][3].

Continuous Classification: Imitation Index.
We computed a LPS Imitation Index (imitation strength score), using the average gesture rates across days, while considering both baseline response rates and responses to the Control (CTRL) condition. We used the formula: LPS Imitation Index = [(LPS Stimulus -LPS Baseline ) LPS Cond ] -[(LPS Stimulus -LPS Baseline ) CTRL Cond ]. We calculated a difference score: LPS produced in Stimulus and subtracted from it LPS produced in Baseline (Still-Face). This difference score was computed for the LPS and CTRL conditions, and we subtracted the CTRL condition from the LPS condition to obtain the difference of the difference scores. The resulting value is small or negative if there was a greater imitative response in the CTRL condition (non-Imitators), and greater than zero if there was a greater imitative response in the LPS condition (imitators). For further classification details, see [3].

Preliminary Neonatal Imitation Analyses
We first explored interindividual variability in neonatal imitation. Imitators produced a larger 96. In addition to the categorical classification of infants as imitators and non-imitators, we also calculated a LPS imitation index score for each infant, which revealed varying levels of imitative capacity, even within the imitators, Figure S1. We obtained the same pattern of results when imitation was a continuous variable (see Fig. S2); therefore, for simplicity, we hereafter use the categorical classification, i.e., labeling infants imitators and non-imitators, consistent with previous approaches [e.g., 1,2,5,6]. Our data are consistent with previous reports in newborn humans and macaques, suggesting that infants vary in their imitative capacity, with approximately half of infants S6 imitating consistently, while approximately half do not imitate [1,4]. In sum, these results lend support for the proposal that neonatal imitation is a good skill to assess in terms of interindividual variability.

Preliminary Gaze Following Analyses
We excluded 24% of trials because infants failed to produce a usable response (e.g., looked up or straight-ahead). This resulted in the exclusion of an equal proportion of trials for imitators (head: r = -.13, p = .17; head+torso: r = -.14, p = .14).

Consistency of Imitation and Gaze Following Responses
We carried out a repeated measures ANOVA on the neonatal imitation index with the within subjects factor of test session (1, 3, 5, 7 days of age). There was no significant change in imitation across days, F(3,333) = .52, p = .672. We carried out a repeated measures ANOVA on gaze-following scores with the within subjects factor of test session (1,2,3,4). There was no significant change in gaze-following across sessions, F(3,333) = 1.03, p = .378. Together, these results suggest that there were no detectable systematic changes in infants' performance across sessions in either task.  Figure S2. Both low-socialization (top graph) and high-socialization (bottom graph) rearing groups showed a positive relationship between neonatal imitation and gaze following. Infants' overall imitation index predicted their gaze following, F(1,116) = 5.76, p = .004, with a correlation coefficient of .30, indicating that approximately 9% of the variability in gaze following scores can be accounted for by infants' imitative capacity and early rearing environment.

S10
Supplemental Table   Table S1. One-sample t tests comparing gaze following performance to chance (.50) for each group of infants (imitators and non-imitators, high-and low-socialization infants) within each condition (head-turn and head+torso-turn). Means (M), standard deviations (SD), and effect sizes (Cohen's d) are also reported.