Point-of-gaze, EEG and ECG measures of graphical/keyboard interfaces

H. David1, R. Mollard2, P. Cabon2 and B. Farbos2

1Eurocontrol Experimental Centre, Bretigny-sur-Orge, France
2Laboratoire de l'Anthropologie Appliqué, Paris, France

Eight experienced but not currently practising Air Traffic Controllers carried out simulation exercises using the TRACON II Autonomous Air Traffic Control (ATC) simulator at Eurocontrol Experimental Centre, (EEC), Bretigny, France. Each controller undertook four exercises, controlling 15 and 30 aircraft entering in 30 minutes, using graphic (trackball/pointer/windows) and coded keyboard input methods. The two lighter loaded exercises were carried out on one afternoon, following some preliminary training exercises, and the more heavily loaded exercises were carried out on the following afternoon, to minimise circadian rhythm problems.

A preliminary analysis was carried out using The Observer (Noldus Information Technology) on-line, to record major shifts of attention. Significant events during the exercise were noted and subsequently included in the record. The point of gaze was recorded using a SensoMotoric Instruments iView head-mounted eye-tracking device. Eye-movement analysis was confined to the busiest 20 minutes of each simulation. Initially, the location, duration and frequency of eye-movements were analysed on a minute-by-minute basis. The mean number of fixations per minute was approximately 15 per minute for the graphic mode, and 25 per minute for the keyboard mode. In keyboard mode, the controllers switched frequently between the keyboard, the active strip area and the radar. Controllers using the graphic interface spent about 57% of their time looking at the radar, 17% looking at the active strips and 17% looking at 'pop-ups'. Controllers using the keyboard interface spent about 47% of their time looking at the radar, 20% looking at active strips and 20% looking at the keyboard. Surprisingly, the traffic load made no significant differences in the duration or number of fixations, for either control mode. The controllers' left/right frontal electro-encephalogram (EEG) and electrocardiogram (ECG) were recorded using a Vitaport psychophysiological recorder. The estimated theta-rhythm power rose for higher traffic load in the keyboard mode, as might be expected. In the graphic mode, however, it fell. For both modes, the variability of heart rate (sinusarythmia) fell for higher task loads. There was a significant negative correlation between sinusarythmia and the number of fixations for radar, active strips, and keyboard and between sinusarythmia and the time spent looking at the radar.

There were significantly higher scores for the mental, physical and temporal demand and effort sub-scales of the NASA Task Load Index for heavier traffic load. The overall TRACON score was higher for higher traffic load, although the deliberate overloading of controllers in the high traffic load produced variable scores. Specific error frequencies showed a more complex pattern. There were more separation losses in the keyboard mode, suggesting less situational awareness. There were more handover errors in heavy task load conditions, suggesting that controllers may decide to 'shed' this task under time stress, and more missed approaches in the graphic control mode, which may be attributed to the lower precision of the graphic methods for height and speed allocation.

This was an initial feasibility study, which should be repeated with larger numbers of subjects. The observed results can only be regarded as tentative, but are indicative.


Paper presented at Measuring Behavior 2000, 3rd International Conference on Methods and Techniques in Behavioral Research, 15-18 August 2000, Nijmegen, The Netherlands

© 2000 Noldus Information Technology b.v.