Visual Search

Studies of visual search investigate the process of looking for and identifying the presence or absence of a specific visual stimulus (a target) embedded among other items (distracters). To date, most of the research questions that have been asked in this area have been related to the very interesting finding that visual searches for some features are much easier and faster than searches for other features. Two natural questions emerge from these findings: Which visual features are found quickly and easily and which visual features are not? And, of course, why are some visual searches easier than others? The answers to both questions and an understanding of the mechanisms involved in visual search will contribute to our understanding of the active nature of visual information processing and to the perceptual organization of our visual world.

How does this perceptual mechanism work? Although low-level stimulus features such as color or size play a major role in determining the efficiency (speed) of a visual search (Geisler & Chou, 1995), other factors like familiarity also play a role (Lubow & Kaplan, 1997; Wang, Cavanagh, & Green, 1994).

As mentioned above, one fascinating aspect of our ability to perform a visual search is that some physical characteristics of stimuli allow for easy and efficient searches, whereas other stimuli result in difficult and time-consuming searches. This property is often referred to as salience, and objects that have high salience are perceived to “pop-out” from their surroundings. It appears that visual scenes can be processed in parallel, that is, simultaneously, and pre-attentively for these items. Low salience objects, on the other hand, require lengthy searches. A search for low salience objects seems to take place in a serial format where item-by-item processing is required. Sometimes simply changing which stimulus is the target and which stimuli are the distracters changes the quality of the search from one mode to another. For example, Triesman & Souther (1985) found that searching for an “Q” among “O”’s was an easy (parallel) task but that searching for a “O” among “Q”’s was a more difficult (serial) task and we demonstrate this with the demonstrations below.

Figure 14. This is an example of searching for a "Q" among a small array of "O's". Clicking on the button labeled "Start" will present the stimulus array for a brief 100msec. This is an example of a parallel search: the "Q" pops out and you have no difficulty seeing it despite the very short duration of the display. Now try the demonstration below:

Figure 15. Same as above, but now you are searching for a "Q" in an array of 36 other items. Again, click on "Start" to flash the array. Here, despite the fact that there is a 9-fold increase in the number of search items, you find the "Q" quickly. There is little if any influence of the number of items. O.k., now try the demonstrations below. In the first example you will be searching for a "O" among "Q's" in an array of 4 items.

Figure 16. Again, you have no difficulty identifying the presence of the target, an "O" in this case, when the array size is small. Now try the following:

Figure 17. Ouch! Were you lucky enough to see it? Not likely. In an array of items such as this the length of presentation must be considerably longer in order to allow people enough time to identify the presence or absence of a target "O". It does not pop-out.

Some of the additional visual features that pop-out are brightness (Gilchrist, Humphreys, Riddock, & Neumann, 1997), color (D’Zmura, 1991), and motion (Nothdurft, 1993). Quite often, though, we are searching for objects that can only be identified by the simultaneous presence of two or more stimulus features. These so-called conjunction searches have been studied for color and form (D’Zmura, Lennie, & Tiana, 1997), motion and form (Muller & Found, 1996), color and orientation (Friedman-Hill & Wolfe, 1995), and for the conjunction of two colors or two sizes (Wolfe, 1992). Some conjunction searches are very difficult and require serial processing. A good example is the children’s game “Where’s Waldo.” In this popular game, children search drawings of crowded social events for one particular individual (Waldo) who is characteristically dressed in a red and white-stripped sweater and cap, glasses, and dark hair. Waldo must be distinguished from a crowd of “distractor” individuals who possess some but not all of these features.

Although there has been extensive research on the topic of visual search over the last decade, it is evident that there is still much to be learned about the basic processes involved. The results of dozens of visual search experiments (studying many different types of visual features and feature conjunctions) has shown us that a sharp distinction between serial and parallel processing may be too simplistic (Wolfe, 1998). The allocation of attention in visual search probably lies along a continuum, where stimulus features and context determine search efficiency. This raises interesting questions for future research. For example, what situations result in optimal search efficiency? How could the search context for a specific target be manipulated to maximize search efficiency?

The answers to these questions have a very practical value in applied settings. That is, outside the laboratory in the real world. What we have learned about visual search has quickly been applied to many real world situations such as air traffic control (Vortac, Edwards, Fuller, & Manning, 1993), driving (Lajunen, Hakkarainen, & Summala, 1996; Summala, Pasanen, Rasanen, & Sievanen, 1996), visual display design (Fisher & Tanner, 1992) and how visual displays are monitored in the workplace (Liu, 1996). Today, questions regarding sensation and perception are increasingly being applied to problems outside the laboratory.