Abstract
Past research demonstrated a top saliency bias in object identification, with random shapes appearing more similar when they share the same top vs same bottom. This is consistent with tops of natural objects and living things tending to be the most informative locations of intentionality and functionality, which leads observers to favor attending to tops. However, this bias may also reflect a generic downward vantage tendency, with more informative aspects of scenes tending to lie below the horizon. Here, two experiments test for a top saliency bias with objects and bottom saliency bias with scenes. Participants observed picture triptychs and judged if the central object or scene appeared more similar to comparison objects or scenes that contain the same top vs the same bottom. Experiment 1 stimuli were images of impoverished, information-balanced polygons or scenes comprised of polygons. Experiment 2 extended the triptych stimuli to images of naturalistic objects or scenes. We took a Bayesian approach to model the beta-binominal distribution of the top saliency responses. Results support a vertical saliency bias for object tops and scene bottoms that varies as a function of informative aspects of visually attended stimuli, with the bias strengthened for more ecologically valid stimuli. However, for information-balanced stimuli, this pattern held for objects but not scenes, supporting that vertical information imbalance drives a generic downward vantage bias.