Users of augmented reality (AR) must direct their attention toward real world as well as artificial information. The authors investigated some aspects of interference between the 2 sources of information that affect performance in completing a visual search task. The search task was carried out under 3 different conditions, 2 of them as found in AR in mobile systems. Participants were asked to detect a target that was superimposed on a background. Target and background were presented on a screen subtending a rectangular area of 55degrees x 43degrees (horizontal x vertical). The target appeared at 6 different locations on the screen. A video recording of a car drive served as the background. In 1 condition, the recording was replayed continuously. Static images of the record were sampled at 5-sec intervals and replayed as background in another condition. A uniform gray background served as a baseline.
Detectability (d') of the target was highest in the baseline condition. A reduced detectability was found in the presence of static images. Lowest detectability was found in the condition with continuous playback of the video recording. A deterioration of reaction time was found to increase with the same order of conditions as listed earlier. Participants were more efficient in completing the detection task when the targets were presented in the lower part of the screen than in the upper part. The authors concluded that performance in detecting artificial information depends not only on spatial characteristics but also on temporal variations of the background on which the artificial information is superimposed. Determination of suitability of AR systems used in mobile applications therefore requires the characterization of temporal aspects of the presented visual information.
They also concluded that the presentation of artificial information in the upper field of vision is a practicable alternative in the case that the lower field is overloaded. However, this statement is true only in the absence of motion. It can be assumed that a visual task involving real world information may be impeded by the adding of artificial information. Artificial information in AR systems should, therefore, be avoided whenever it is not needed.
Because of the particular material used in the experiment, the outcome of the search task might have depended on the participants' driving experience. The results obtained, however, indicate that the total amount of kilometers driven is not correlated with performance on the task.