Skip to main content
Dryad logo

Data from: Social and emotional contexts predict the development of gaze following in early infancy


Astor, Kim (2020), Data from: Social and emotional contexts predict the development of gaze following in early infancy, Dryad, Dataset,


The development of gaze following begins in early infancy and its developmental foundation has been under heavy debate. Using a longitudinal design (N = 118), we demonstrate that attachment quality predicts individual differences in the onset of gaze following, at 6 months of age, and that maternal postpartum depression predicts later gaze following, at 10 months. In addition, we report longitudinal stability in gaze following from 6 to 10 months. A full path model (using attachment, maternal depression, and gaze following at 6) accounted for 21 % of variance in gaze following at 10 months. These results suggest an experience-dependent development of gaze following, driven by the infant’s own motivation to interact and engage with others (the social-first perspective).


Gaze data was exported from Tobii Studio and processed in MATLAB version R2017b (9.3.0. 713579) using TimeStudio (version 3.18). See the time series analyses file and visit for user instructions. 

Maternal postpartum depression was assessed using the Edinburgh Postnatal Depression Scale (EPDS) at 6 weeks, 6 months, and 12 months after child delivery.

The SSP was used to assess infant-mother attachment quality when the child was 12 months old. Two certified coders rated attachment according to the ABCD classification, providing a B (secure) vs. ACD (non-secure) score. For more information about the social context measures, including data reduction, see the Supplementary.

Usage Notes

This is the R code used for the path analysis:




GF_data <- read.csv2('BASIC_GF.csv')


Model <- '

GF_10 ~ GF_6 + Dep12mpp + Dep6mpp + Dep6vpp + BvsACD

GF_6 ~ Dep6mpp + Dep6vpp  + BvsACD

Dep12mpp ~ Dep6mpp + BvsACD

Dep6mpp ~ Dep6vpp + BvsACD

Dep6vpp ~ BvsACD


gCD.mod<- genCookDist(Model,data=GF_data,

plot(gCD.mod,pch=19,xlab="observations",ylab="Cook distance")

GF_data.gCDExc <- GF_data[-which(gCD.mod>1),] <- sem(Model, data = GF_data.gCDExc, estimator = 'ML', missing='fiml', std.ov = TRUE)

summary(, fit.measures = TRUE,standardize = TRUE, rsquare = TRUE)