Prediction of Information Transfer from Computer-Generated Dot Matrix Displays

Abstract

Two separate experiments derived and validated predictive metrics (or equations) of information transfer from dot matrix displays. The initial experiment involved three levels of dot size, three levels of dot shape, three levels of dot spacing, and two levels of ambient illuminance. The predictor variable pool was obtained by Fourier analysis of vertical and horizontal microphotometric scans of each experimental display combination.

Multiple regression techniques were used to derive predictive equations for each task. Substantial proportions of experimental variance were accounted for by these equations. These proportions ranged from R² = .47 for the menu search model to R² = .57 for the reading task model.

An external validation study was conducted using dot parameters equivalent to three commercially available displays. In addition, the matrix size was varied in this experiment. Except for the random search performance, the predictions derived in the first study correlated well with the meassured performance in this study (Spearman r = .73). The equations were found to be very sensitive to predictor variables which were outside the range of the original regression.

The photometric scan and Fourier analysis methodology was found to be accurate and very repeatable in this research.

This research has demonstrated that it is possible to account for large proportions of experimental variance on visual performance tasks with relatively simple display-related parameters. The proportion of variance accounted for by the derived models ranged from .47 for the menu search model to .57 for the Tinker SOR model. The terms used in the metrics inherently contain information about many display parameters usually treated as isolated from one another. The values predicted from the metrics have been shown to be well correlated with actual performance when the predictor variables are within the range of the original variables from which the metrics were derived.

The methodology developed during this research has been shown to be valid and the photometric procedures and mathematical treatment have proved stable in a real experimental setting.