In a complex visual image there are many locations that might be important to observe. We can only direct our foveas to one of them at a time, and efficient processing of the image requires us to move our eyes to such important locations as smoothly as possible. We can indeed target our saccadic eye movements to prominent sites within the image, and this targeting is thought to depend on a retinotopic salience map that represents the positions of such sites. Where in the visual system is this map located? Reporting in Neuron, Mazer and Gallant make a compelling case in favour of extrastriate area V4, and show that the map is more complex than previously envisioned.

To investigate the existence of the salience map, the authors trained monkeys in a free-viewing visual search task. The animals were shown a target stimulus (a circular image) on a screen and subsequently asked to identify the target in an array of possible matches. The experimenters recorded the eye movements of the monkeys and, simultaneously, the activity of neurons in the V4 area. They found that neurons showed enhanced visual activity, which preceded saccadic eye movements that were directed to the receptive field of the active neurons. In fact, this retinotopically organized, visually driven activity was predictive of the direction of subsequent saccades. Moreover, the authors found evidence that activity in V4 was not the reflection of an oculomotor signal, but was solely dependent on visual input.

A key finding from this study was that the activity of some V4 neurons could be modulated by the identity of the target. As the original proposal of a salience map did not include a role for top–down modulation of this kind, the results add an extra level of complexity to salience computation. The data also imply that there is no single region that performs this computation, but it depends on the interactions between several areas during natural visual search. Ultimately, the interplay between this network and regions that control target selection and oculomotor planning are responsible for the ability to explore complex visual images in an efficient way.