“Aunt Amanda’s pink buggy is over there!” your small child shouts as you are standing on the corner of a huge parkfield. How can he discriminate it? How can anyone find a particular object in a world filled with thousands of other objects? In particular, how does the search for a looked-for thing proceed when one knows its colour or location beforehand?

Two main questions have to be addressed: “what is selected?” and “how is it selected?”. The first question is concerned with where visual information processing is directed, and there are two possible answers: either locations or objects are selected. The second question is concerned with the how of selection, in other words, what cues are able to trigger and mediate it. A common finding has been that location is a superior cue to other stimulus features (e.g., colour, shape, size) in selecting among simultaneous stimuli. I am mainly concerned with the latter question in this paper.

Several theories can account for the advantage of spatial cues over other kinds of cues. In what follows, five theoretical views that propose different predictions for the effects of pre-knowledge of location and other stimulus attributes are presented. The first three are early-selection theories, and the latter two are late-selection theories. According to the early-selection approach, since sensory processes are limited, they require attention to select the stimuli that are further processed. Attentional selection should thus occur early, before the identification of the presented stimuli, but after the preliminary analysis of the physical stimulus features, and after the division of the visual field into objects and a background. The late-selection approach, in turn, claims that all the stimuli can be processed automatically in parallel, without any need for selection. Selection should thus occur late, after the semantic analysis of all the stimuli in order to recruit the effectors in an appropriate way.

First, the spotlight theory describes the focus of attention as a mental spotlight. At each moment, the attentional spotlight is directed on a restricted area in the visual field, so that the quality of perception is enhanced within that area. As a result, valid pre-knowledge of a target’s location improves performance compared with trials in which the target appears at an unattended location. In this theory, no pre-attentive processing occurs outside the attentional focus, whereas within the focus, bottom-up signals can automatically guide attention to the cued location. It is also possible to voluntarily deploy attention to a particular location in a stimulus field. Top-down commands to attend to a target’s particular feature have no effect, however. For example, pre-knowledge of a stimulus colour cannot improve reaction-time performance, because it cannot help subjects to know where to attend to next.

Second, an alternative version of the space-based account claims that the size of the attentional focus can be adjusted like a zoom lens. Attention can thus vary from a uniform distribution over the stimulus field to a highly focused concentration on a particular location. Even though pre-knowledge can facilitate the processing of stimuli within the attentional focus, there is a gradient on the borders of the focus so that the facilitation drops progressively from the focus centre to the border. When a location cue is shown, attention is concentrated on a restricted area of the visual display. When no spatial information is provided, attention is distributed over the entire stimulus area. Advance information about a particular stimulus feature (other than its location) could, however, prime the processing of the critical feature value. Hence, for example, a colour cue as a prime can facilitate the detection of the target. Since priming effects are an example of automatic pre-attentive processing, primes can facilitate target processing, but they cannot inhibit the processing of other stimuli.

Third, the second group of early-selection theories claims that pre-knowledge of stimulus attributes other than their location can produce both costs and benefits. According to these theories, even though attentional selection is always based on location, different types of cues can be used to guide attention to a target’s location. Therefore, these views might be called mixed early-selection theories, according to which the selection of a target can be thought of as a two-stage operation. First, there is full-field pre-focal selection based on location information or information about other stimulus attributes, and then, attention is deployed to the critical location. In addition to the bottom-up activity at the pre-attentive level, top-down commands are thus also capable of guiding attention. If the target is not salient enough, that is if the bottom-up activity does not make it possible to discriminate the target from a distractor, top-down knowledge is critical. Thus, valid colour cues, for example, as well as location cues would produce benefits, and invalid colour cues would incur costs.

Fourth, if only two elements are presented, one target and one distractor, late selection should be possible, so that both elements are identified before selection. According to the post-categorical filter theory, pre-knowledge of the target’s location, via the location module, elicits additional activation in the target’s representation in the input module. Since stimulus selection occurs late, various cues can be used to trigger this feedback loop, even though selection is always mediated by location. Selection by abrupt onset location cues can be thought of as a one-step operation, and selection by other attributes as a two-step operation running via a representation of visual space. In a similar way, symbolic location cues, like arrows, should be effective, but they cannot be used immediately in the same manner as peripheral location cues. There must be a top-down signal from higher centres that can interpret the symbolic cue in terms of the rule, and then transform this interpretation in such a way that stimulus processing is enhanced on the cued direction in the array-based representation.

In this theory, facilitating the processing of a target by repeating a target-defining feature over trials can influence the selection process. If, for example, the target’s shape remains the same from trial to trial, the processing of the target can be facilitated to a degree that the target is more quickly detected and identified. Hence, even though feature cues, such as information about the target’s colour presented in advance of the target as a precue, should produce cue-validity effects, priming for shape might be capable of regulating the effect of colour cues.

Fifth, the all-features-are-equal theories also claim that stimulus selection occurs late. What the selection system has to do is to match a template to the stimulus representations and to assign weights to the input proportional to the degree of match. The higher the weighting, the more probable it is that the representation is passed through a limited capacity system to the next level. Different criteria can be used in setting these weights; spatial information is in no way special. Spatial cues may, however, provide more discriminative power. Information on stimulus location may be encoded, for example, more rapidly or accurately than other stimulus features. Therefore, spatial cues may be more useful than other feature cues in distinguishing a target from distractors.