because attention may be delayed in returning to fixation. We know that the inhibition effect lasts a rather long time (up to 2 sec) and it might be possible to obtain a pure measure of the inhibition effect if the time interval for targets is extended beyond those used in the current experiments.
At the very least, these effects seem to suggest a qualification to the conclusions made by Jonides based on his 1981 memory target experiment. In unpublished studies in which a secondary memory task was used, we obtained little evidence of interference with the overall RTs found in the cued spatial orienting task. It appears that memory tasks are not as effective in producing interference as concurrent tasks such as used in Experiments III and IV.
We began this chapter with a brief consideration of two major approaches to the problems of sustained concentration. One position emphasizes a passive filtering of the input information prior to cortical processing. According to this approach, information outside of the attended source is attenuated or filtered at an early stage while attended information is unfiltered. The other position emphasizes active selection of the attended channel by an effortful alignment of attention with signals from the selected source.
There have been two major empirical approaches to the study of sustained attention. One has involved presenting two streams of information with subjects required either to process both (divided attention) or to concentrate on only one (focused attention). The results of this paradigm have shown that focused attention is better than divided only when events on the attended channel must be actively processed, such as in shadowing each event, or when one examines the unattended channel at the moment a target is selected on the attended channel ( Duncan, 1980). It is also important that successive events follow quickly, otherwise the evidence of selectivity obtained from behavioral or EEG result often is lost. The second major paradigm for the study of sustained concentration has involved vigilance. Here only a single source of information is presented. Evidence from high event rate vigilance tasks suggest that when information is rapid, there is a loss in sensitivity over time even with only a single channel of input ( Parasuraman, 1979). Thus, conditions for maximum selectivity are also conditions where the input from the selected channel tends to be reduced in quality.
These empirical observations tend to fit rather poorly with a concept of early primarily sensory and passive filtering of attended information. If such filtering accounted for selectivity, it would not seem to depend so heavily on active processing of each attended signal as it so frequently does. Nor do our results fit with the ability to form and maintain a passive bias toward the selected channel. Instead we find that the selected and unselected information alike is subject to an