Top Versus Bottom Saliency Bias in Object and Scene Perception

Research has demonstrated observers have a generic bias for top saliency in object identification, such that random shapes appear more similar to ones that share the same tops versus same bottoms (Chambers et al., 1999). These findings are consistent with the idea that in nature, the tops of most important objects and living things tend to be the most informative locations with respect to intentionality and functionality, leading observers to favor attending to top. Yet, such a bias also may imply a generic downward vantage bias, suggesting that unlike natural objects, the more informative aspects of scenes tend to lie below their horizon midpoints. In two experiments, saliency bias was investigated for objects and scenes with both information-balanced and naturalistic stimuli. Experiment 1 replicates and extends the study of the top-saliency effect for information-balanced objects. Here 91 participants made 80 similarity judgments between an information-balanced object and two comparison objects that contain either the same top or the same bottom. Participants also made 80 similarity judgments of information-balanced scenes in which the coordinates of the vertices of the random shapes were replaced with little objects to create a scene. Experiment 2 extends Chambers et al. (1999) by examining top-saliency bias in naturalistic object perception when 91 participants made similarity judgments between a photographed test object and two comparison objects which contain either the same top or the same bottom. Experiment 2 also tests the idea of a downward vantage bias by predicting that naturalistic scenes will be judged more similar when the portions that lie below the horizon are identical versus when the portions above are the same. Results of the two experiments confirm that observers tend to assume a downward vantage when viewing pictures of objects and objects within scenes, which supports that saliency varies as a function of the informative aspect of the visually attended component.