An eye tracking investigation of developmental change in bottom-up attention orienting to faces in cluttered natural scenes

Abstract

This study examined the contribution of visual salience to bottom-up attention orienting to faces in cluttered natural scenes across development. We eye tracked participants 4 months to 24 years of age as they freely viewed 16 natural scenes, all of which had faces in them. In half, the face was also the winner-take-all salient area in the display as determined by the MATLAB SaliencyToolbox. In the other half, a random location was the winner-take-all salient area in the display and the face was visually non-salient. We found that proportion of attended faces, in the first second of scene viewing, improved after the first year. Visually salient faces attracted bottom-up attention orienting more than non-salient faces reliably and robustly only after infancy. Preliminary data indicate that this shift to use of visual salience to guide bottom-up attention orienting after infancy may be a function of stabilization of visual skills. Moreover, sociodemographic factors including number of siblings in the home and family income were agents of developmental change in orienting to faces in cluttered natural scenes in infancy.

Figure 1. Illustrates images shown to participants and associated saliency heat maps.

Hot colors represent higher salience locations. A) Face is winner-take-all most salient area in display as determined by the MATLAB SaliencyToolbox . B) Non-Salient Face example image where faces are not most salient area in the display. Individuals in photographs have given written informed consent, as outlined in the PLOS consent form, to publication of their photograph.

Figure 2. Illustrates the beta coefficient analytic strategy.

Each input image was divided into a 16x25 matrix. Saliency and visual feature maps were extracted for each matrix grid location (MGL) and then used as predictors in multiple regression analyses, in order to determine their explanatory power for the duration of looking variable in those MGLs, per participant and image.

Figure 3. Depicts proportion attended faces in the Salient and Non-Salient Face conditions.

Average proportions are binned per month in participants less than one year of age (4 month-olds as separate from 5 month-olds), and per year subsequently (3 year-olds are followed by 4 year-olds). Standard error bars reflect variability within age bin.

Figure 4. Illustrates beta coefficients for color, intensity, and orientation maps.

Average beta coefficients are binned per month in participants less than one year of age, and per year subsequently. Standard error bars reflect variability within age bin.