Ruth Rosenholtz believes that visual attention is in crisis. Her argument makes several valid points. There is a tendency to reify “attention” as a single thing when, in fact, attention refers to a large family of selective processes in the nervous system. If I ask you to “pay attention” to your toes in your shoes, that is a different “attention” than the attention you would use to look for the letter “x” in the previous sentence. Turning to my favored topic of visual search, Rosenholtz is undoubtedly correct that several puzzles in the visual search literature might be understood as peripheral vision/crowding/clutter phenomena. Her work on these effects makes important contributions to the search literature. But Ruth would like to go further. Her initial hope was to ban “attention,” at least as a term in discussion in her lab. However, as the bulk of her article shows, there is a range of tasks that appear to involve selective processes we could call “attentional.” I want to argue that visual search is one of those tasks that involves a central role for selective visual attention, even if the properties of peripheral vision do play their part in search.

The synthesized images in Figure 4 in Rosenholtz’s article are a nice illustration of how peripheral vision might make chopped salad out of the preattentive features in a display. Indeed, the figures evoke Anne Treisman’s claim that, in preattentive vision, features are unbound, floating free from their objects (Treisman & Gelade, Reference Treisman and Gelade1980). However, the existence of this jumble of features fails to explain how it is that observers can find and identify targets in peripheral vision if those observers successfully deploy attention to that object. Could the target emerge from the jumble without attention? Perhaps features just recombine randomly, allowing you to successfully detect the target if you wait. Suppose, however, that we flash a ring of letters around fixation and ask if there is a “T” present. Let us assume that the letters are big enough and widely separated enough that there is no acuity/crowding effect blocking the identification of the letter in the periphery. With a short flash, say – 100 msec, performance will be poor. However, if we cue the potential location of the T, perhaps 100 msec before the onset of ring of letters, then performance will be near perfect. The task has not changed, but the elimination of spatial uncertainty changed appearance of the T. In that 100 msec, something has been deployed to the cued location, allowing the jumble of features to become a recognizable T. I wrote that sentence without the use of the terms “binding” or “attention,” but that is a mere word game. The standard way to describe this situation is to say that the exogenous cue (Klein, Reference Klein2009) directed your selective attention to the subsequently presented object. Attention enabled the binding of the object’s features into a recognizable letter, “T.” In the absence of the cue, either attention is deployed to the wrong item, by chance, or it is diffused over multiple items, and, most of the time, you fail to locate the T.

Figure 1.

What do you see when you look at the black cross? How can what you see change without a change in the stimulus?.

If you do not like the idea of covert attention, a different objection can be raised to the notion that peripheral vision fully accounts for the difficulty of some search tasks, even if we agree, peripheral factors are important. In most real-world search tasks, the eyes move around because, as Rosenholtz notes, object recognition is not very good in the periphery. If eye movements are allowed and observers search for that letter T, they will typically miss 5–10% of those targets. Moreover, if we track their eye movements, we will find that, in many cases, those observers fixated on or were very close to the missed target. That is, the target fell in the fovea and was missed, nevertheless (Wu & Wolfe, Reference Wu and Wolfe2022) (Wolfe et al., Reference Wolfe, Wu, Li and Suresh2021). Why are fixated targets missed on some trials but not on others (Li, Wolfe, & Hulleman, Reference Li, Wolfe and Hulleman2024)? It seems unlikely that basic visual processes changed. This seems more like a failure that involves something that we might call “attention” (Wolfe, Kosovicheva, & Wolfe, Reference Wolfe, Kosovicheva and Wolfe2022).

If you look at the black fixation cross in the figure, you can get an intuitive feel for the processes discussed here and in the main Rosenholtz article. You immediately see a ring of two-color plusses. Call this the “gist” of the scene; the product of what we have labeled as a “non-selective pathway” (Wolfe, Vo, Evans, & Greene, Reference Wolfe, Vo, Evans and Greene2011). You might also sense some instability as implied in the illustration of peripheral summary statistics shown in Figure 4 of Rosenholtz’s target article. Now, looking at the plusses presented in this comment, without moving your eyes, search for purple-vertical, green-horizontal plusses. You seem to do this by “attending” (for want of a better word) from item to item or location to location, binding features until you find the target. While still fixating, direct your attention to the bottom item. If you found the target at the upper left, you might not have noticed that there are two targets. Now, with no change in the stimulus, you are aware that bottom item looks like a target, too. You could say that you shrank your Functional Visual Field around that item, allowing you to limit processing to a small set of features (Hulleman & Olivers, Reference Hulleman and Olivers2017), but that sounds quite a lot like attending to that item.

In sum, Ruth Rosenholtz does not much care for the term “attention.” As Miranda says of the “monster,” Caliban, in Shakespeare’s Tempest, Ruth might say, “’Tis a villain, sir, I do not love to look on.” (Tempest 1:2:370–371). Like Prospero, we might respond “But, as ’tis, we cannot miss him. He … serves in offices that profit us” (Tempest 1:2:372–375). There is much work to be done on visual search and visual attention, more generally, but visual attention is not in crisis, and no model of visual search will be plausible without a role for visual selective attention, whatever one may call it.