Psychologists and neuroscientists who study vision use the word “salience” all the time, in a rather confusing way. It seems to me the word takes on 3 different meanings, sometimes in the course of one paper. All of these are useful and would deserve their own word.
First, there is salience as a visual attribute. When we call a visual object “salient”, we are saying that it stands out, that it exerts a pull on our attention. It’s an interesting visual attribute because unlike others, like “red”, or “blurry”, calling something “salient” may be partly a *meta-perceptual* statement: a statement about what the object does to our visual system. The red dot in the classical pop-out display above is red, symmetrical and round. These are all direct visual properties of the dot. The red dot is salient, because it attracts our attention. This makes the property of symmetry or redness primary (perceptual), and that of saliency secondary (meta-perceptual).Second, there is the idea of a saliency map as imagined by Koch & Ullman (1985). A saliency map is a hypothetical computational mechanism that the brain may use to decide where to move the eyes next. According to that theory, each visual location is attributed a certain intensity, and the eyes move to the location where that intensity is highest. Neurophysiological theories are more specific and assume for example that the map is implemented in a set of neurons, with each neuron representing some spatial location, and its firing rate the intensity at that location.
I’ve used the word intensity here to avoid using the word salience again. As everybody is well-aware (this includes of course Koch & Ullmann), many factors come into play in the selection of eye movements, of which visual salience is only one. If I give you the task to look only at the green dots above, you will be able to do so, despite the salient red dot. This implies that in the context of the task , if there is indeed a map in the brain that determines eye movements, then its intensity at the location of the salient red dot is low. That’s the problem pointed out by Fecteau & Munoz (2006), and they suggest using the word “priority map”. This would make a lot of sense, but the history of science suggests that bad formulations tend to stick around regardless (eg., statistical “significance”), so I’m not too hopeful.
Third, there is empirical salience. That’s used essentially when someone runs an experiment where eye movements are measured, finds that subjects were very interested in a certain location/object, and calls that location “salient”. In this case, salience does not describe a visual attribute, nor a theoretical construct, but an empirical finding. It’s less cumbersome than saying that the location has a high density of fixations, but it’s a potential source of confusion because a location can be empirically salient without being visually salient, and vice-versa.
In conlusion: we should probably talk about a priority map from now on. I haven’t got a good proposal for another word to use “empirical salience”, unfortunately. In the meanwhile maybe the right policy is to always prefix salience with “empirical” or “visual”, so that it’s clear which one we mean.
References
dahtah 