Attention shifts in the visual field were investigated using a task in which two numerals were flashed one by one, the duration of each numeral being 33 msec. Both numerals were followed by a mask pattern (100 msec in duration). The observer was asked to report the two numerals. If the observer correctly recognized both numerals, it implies that he was able to perform a simple act of attention shift from one item to the next. In the foveal experiment, both numerals were presented at fixation--the second numeral on top of the first. In the extrafoveal experiment, the numerals were also flashed sequentially but in different locations: the first numeral was at fixation and the second at 7 deg eccentricity. The results showed that the observers could not shift attention away from fixation to an extrafoveal position as efficiently as they could maintain attention at fixation.