DEPARTMENT OF ANTHROPOLOGY- HUMAN BEHAVIOR RESEARCH | simulation

The representation of self reported affect in body posture and body posture simulation

Karl Grammer, Bernhard Fink, Lisa Oberzaucher, Michaela Atzmüller ,
Ines Blantar and Philipp Mitteroecker

COLLEGIUM ANTHROPOLOGICUM 28 Suppl.2 (2004) 159-173.

It is taken for granted that the non-verbal information we acquire from a person's body posture and position affects our perception of others. However, to date human postures have never been described on an empirical level.

This study is the first approach to tackle the unexplored topic of human postures. We combined two approaches: traditional behaviour observation and modern anthropometric analysis.

Photographs of 100 participants were taken, their body postures were transferred to a three dimensional virtual environment and the occurring body angles were measured. The participants were asked to fill in a questionnaire about their current affective state.

A principal component analysis with the items of the affect questionnaire (Positive Negative Affect Scales, PANAS) revealed five main factors: aversion, openness, irritation, happiness, and self-confidence. The body angles were then regressed on these factors and the respective postures were reconstructed within a virtual environment.

The reconstructed postures from the positive and negative end of the regression were then rated by 50 different subjects. We found the ratings to be valid and accurate in respect to the five factors.

This is a revolutionary method in non-verbal behaviour research, which probably could overcome the stagnation in research in this field, because the method can be applied to any sort of low level data.

Figure 1: (a) depicts the bones system of the avatar used in this study. Note that the joints connecting the head to the body are divided into three joints. In this study only two were used, which were kinematically chained. The same was done for the connection of chest to hip, where the abdominal joint is chained and for the connection of hip to thigh where the upper leg is chained. (b) shows the wire frame model used for the body form calculation.

Figure 2: (a) shows an original captured video frame (single image) with the interviewer on the right side and the subject on the left side. (b) shows the result of the three dimensional transcription with Poser 4.0 from different viewpoints.

Figure 3: These figures visualize the results of the regression analysis of PANAS principal components on body angles as follows from top to bottom: social aversion, social openness, happiness, irritation and self-confidence. On the right of each panel the positive axis is depicted and on the left the negative axis. The figure in the middle shows the average for each panel. Each panel was calculated the following: The second figure from the right is the result when the regression is calculated with the slope and a distance of 1. The figure at the utmost right is a factor of 1.5. This method allows the exaggeration of a posture in either direction. For the rating study of the negative pole the second picture from the right was used, and for the positive pole the second picture from the left was used. This picture is marked with the adjective used in the rating study. The picture on the left is marked with the reconstructed dimension. Note that these postures are reconstructions from the principal components of the PANAS scores. Thus they are completely artificial, but nevertheless reach the correct ratings by observers.

UNIVERSITY OF VIENNA