Academic journal article Canadian Journal of Experimental Psychology

Top-Down and Bottom-Up Aspects of Active Search in a Real-World Environment

Academic journal article Canadian Journal of Experimental Psychology

Top-Down and Bottom-Up Aspects of Active Search in a Real-World Environment

Article excerpt

Human visual attention has been studied extensively by asking participants to search for a target on a computer screen. When the target is dramatically different from the other items in the display then the target is easy to find, and thus the properties of the stimulus are believed to drive the attention of the observer. In a different but related line of work, investigations of gaze allocation in complex scenes have tended to focus on the fixations made when people look at static pictures. Much of this work tests the principle that image salience guides where people look, and thus visually distinctive or surprising locations are preferentially fixated.

In both search and scene viewing, therefore, the principle of feature-driven or bottom-up attentional selection has been derived from performance in the restricted conditions of computer-based experiments. The aim of the present research was to determine the extent to which this principle generalizes to gaze allocation in a real-world search task. We begin by describing how visual search has been studied, both in the labouratory and in the context of a natural scene. We then outline some of the reasons why it is important to study the generalisation of cognitive principles to natural behaviour, and consider the implications for visual search and gaze allocation.

Simple Visual Search in the Lab

Visual search-the behaviours and mechanisms that allow us to find visual objects-is possibly the most investigated task in cognitive science. Over a decade ago, more than a million trials had been analysed and thousands of scientific articles published on this topic (see Wolfe, 1998). A standard visual search experiment consists of a target surrounded by several distractors, which are differentiated on the basis of one or two simple visual dimensions (e.g., a "Q" amongst "0"s, a "T" amongst "L"s, or a red horizontal line amongst green horizontal lines and red vertical lines). These items are typically presented on a computer screen that lies completely within the participant's visual field. One of the key findings from this paradigm is that the slope of the function relating search time with the number of to-be-searched items varies with different types of target. When a target can be detected on the basis of a single feature, it is found efficiently, and there is little or no cost of increasing the number of distractors surrounding the target. The target is said to "pop out" and be automatically selected by parallel processing of all the items in the visual field. When targets are defined by multiple, conjoined features, they are more difficult to locate, and search is thought to proceed with the serial allocation of visual attention (Treisman & Gelade, 1980).

A related concept to the case of a pop-out target is bottom-up attentional selection. This is defined as selection that is determined by properties of a stimulus (e.g., its contrast with the distractors). Bottom-up attention can be distinguished from top-down attention, with selection in the latter case being controlled according to the knowledge of the observer. In a search task, the searcher has a certain amount of top-down knowledge about the target that can guide attention. For example, if the target is known to be a certain colour, attention can be guided toward items of that colour, resulting in a more efficient search. However, in some situations, the most conspicuous item in the display appears to capture attention-bottom-up-regardless of the observer's task (Theeuwes, 2004).

Search in the Real World

If we now consider examples of search from our everyday experience, as articles on this subject often do, it becomes clear just how different they are from visual search in the labouratory. Think of the last time you located your keys on your desk or your car in the parking lot. How do these tasks differ from the model tasks used in cognitive science labouratories? First, the target is often not in the visual field at the onset of search. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.