An automated approach for measuring infant head orientation in a face-to-face interaction

Head orientation in face-to-face interactions between mothers and infants is an important component of their communicative processes. Manual coding, however, is laborious. Obtaining inter-observer reliability is difficult, with disagreements mostly being related to the on- and offsets of a limited number of orientation categories. We used a motion capture system and developed an automated method for the quantitative measurement of infant head orientation in mother--infant face-to-face-interactions. Automated motion capture systems have the potential to objectively document not only the on- and offset of behaviors, but also continuous changes. Infants wore a cap with three reflecting markers, and eight infrared cameras captured the positions of the markers. Analytic algorithms generated continuous three-dimensional descriptions of the infants' head movements. We report here on an initial feasibility study of four infants. To evaluate the effectiveness of the automated approach, we compared it to standard manual categorical coding of six infant head orientations. We found that the central reliability issue was disagreement at the boundaries of the coding categories identified by continuous automated coding versus manual coding. The automated method was both more feasible and more precise in capturing continuous small changes. The study provides evidence for the usefulness of automated measurement of infant head orientation when infants interact in relational space.