This thesis reports on the development and validation of a computer tool created to facilitate the input of three-dimensional human postural information. The computer program under development attempts to provide a high resolution technique that is easy to use and not overly time-consuming. The method incorporates two- and three-dimensional graphics and allows the representations to be manipulated to the same perspective as the subject being recorded. A mouse input system is used to allow users to select and manipulate limb postures from one of six different views (left, right, front, back, top, and three-dimensional). The separate views are coordinated to force a consistent representation for later analysis or storage. Human factors concepts have been incorporated into the program to increase the spatial compatibility of the task and to streamline the data input in an attempt to increase both recording accuracy and speed.

A validation experiment was performed using 30 subjects, 15 male and 15 female, who were pre-tested for spatial ability. Subjects were asked to input postural information for five postures that were presented on videotape. The postures represented five different positions ('sitting,' 'pushing,' 'lifting,' 'reaching,' and 'crouching') that spanned the range of positions a worker is likely to adopt in the workplace. Subjects viewed each posture from one of three different viewing angles (the 'right' side in the sagittal plane, the 'front' side in the frontal plane, and an 'oblique' angle between the other two). Kappa coefficients were calculated to compare subjects' responses, and it was demonstrated that the method was reliable between subjects (p < .001). An analysis-of-covariance (ANCOVA) was performed on the speed of response data, with the spatial ability scores representing the covariate variable. Those subjects viewing the posture from the 'right' viewing angle had significantly lower times than those subjects viewing the postures from the 'front' or 'oblique' perspectives. Also, subjects were significantly slower when recording the pushing posture than when recording any of the other four postures. The average time required to record a posture ranged from 2.55 to 5.98 minutes, with an overall average of 4.04 minutes. It was also demonstrated that spatial ability had no effect on the speed of the subjects' responses.

PI coefficients and Gamma Statistics were calculated to determine the accuracy of the method by comparing the subjects’ responses to accurate measurements of each posture. It was shown that the recordings were similar to the expert measurements (p̲ < .05) in all cases, except for the condition where the subjects were recording the frontal plane of the ‘crouching’ posture. The accuracy of the method was also evaluated using a analysis-of-covariance (ANCOVA) to analyze the angular deviations of the subjects responses from the expert measurements. Twelve link segments on the stick figure were used as the dependent measures, and the spatial ability scores were used as the covariate variable. The results of the ANCOVA indicated that subjects viewing the postures from the 'right' viewing angle were significantly more accurate than those subjects viewing the postures from either the ‘front’ or ‘oblique’ perspectives. Also subjects were significantly less accurate when recording the ‘pushing’ posture than when recording any of the other four postures. Finally, it was demonstrated that spatial ability had no effect on the accuracy of the subjects’ responses.