How do you search for one thing in a jumble of items? Try the visual searches in , following instructions in the caption. Which of the two was harder? What made one harder than the other? Write down a few ideas and then click to continue.
Visual search is a common theme in studies of attention. By studying how people search through a simple scene, we can learn about how objects are recognized. Anne Treisman performed a series of experiments in the 1980s that led her to suggest that recognition is not immediate, but occurs in two stages. First, basic features are detected, then those features are integrated into a single percept (Treisman, 1988). In the first example image, the dime is distinguishable from the pennies by a simple basic feature, its color, so only one stage of recognition was required. In the second example, you had to integrate several basic features to distinguish the Canadian penny from the American ones, so both stages were required.
You can do two of the experiments that led to this conclusion. , is Experiment 1A and , is Experiment 1B of Treisman and Gelade (1980). Do both experiments; each takes about 5 minutes. After doing both experiments, click to continue.
What can reaction times tell us about visual search? There are two ways a search could proceed in these simple scenes. First, it could proceed in parallel: the entire scene is analyzed at once and the target is identified immediately. In that case, the time taken to react should be the same regardless of the number of items in the scene and whether or not the target is present. The second possibility is a serial search: the scene is analyzed item-by-item until the target is found. In that case, reaction time should increase as the number of items increases. Furthermore, it should take longer to answer when the target is absent (because all items must be analyzed) than when it is present (because the search stops when the target is found).
Your results should have looked something like those in . Results of Experiment A (disjunct search) imply a parallel search. Set size and target presence or absence have little effect on reaction time. However, results of Experiment B (conjunct Search), suggest a serial search. Reaction time increases with set size and it takes longer to answer when the target is absent than when it is present. Why the difference between these two experiments? The targets used in Experiment 1A differ from the other items by a single feature (shape or color) not found in any of the background items. In Experiment 1B, however, the target was defined by conjunction of features (shape and color) where the shape and color were both present in the background items.
Recall that Treisman (1988) proposed two stages to recognition, feature detection and feature integration. Experiment 1A required only the detection stage, which seems to be done automatically, in parallel, by independent feature-detection modules. Experiment 1B required detection (of shape and color) followed by integration (determining whether the target shape and color occur in the same item). It is the integration stage that seems to be done serially rather than in parallel. When doing the experiments, you probably noticed that the second one was harder and may have been aware of scanning each item to find the target.
If your results do not quite match those shown in , there are many possible causes. Subjects in Treisman and Gelade’s study practiced these tasks until they operated at peak efficiency and made few errors. If you made many errors (sacrificing accuracy for speed), your results will differ.
Design and carry out your own visual search experiments. See , for instructions.