Visual search within a web page depends of different strategies, based on semantic and spatial factors. Semantic aspects refer to the similarity between user's goal and label meaning; spatial aspects to label positions expected on the basis of scanning habits. We ran three visual search experiments to show that another strategy is involved; namely, layout following. Users exploit knowledge about web layout conventions and search for items in positions consistent with the categorization of their goals, expecting to find labels that refer to basic level categories in the navigation bar and labels that refer to subordinate level categories in the canvas. In Experiment 1 observers searched for a basic/subordinate target word within either a standard (navigation bar on top and canvas on bottom) or non standard hierarchical page layout. In positive trials labels matching the target were displayed in one of 6 possible positions [3 in the navigation bar and 3 in the canvas]. In Experiment 2 the same visual search was performed in the absence of a web page layout [6 candidate labels on a white background]. In Experiment 3 target non-words were searched in the same standard hierarchical layout of Experiment 1. Experiment 1 revealed a strong layout following effect that cannot be explained by either an independent or a conjoined estimation of the relevance of semantic and spatial labels. Searching within the navigation bar was faster for a basic target than for a subordinate target; while the opposite occurred when labels were displayed within the canvas, regardless of the type of web page layout. The effect was eliminated by removing the layout (Experiment 2) or the target meaning (Experiment 3). Results are consistent with a composite measure of label relevance, defined as the weighted linear combination of semantic, spatial, and layout following aspects.
