Estimations of object frequency are frequently overestimated

Abstract

Real-world scenes are complex but lawful: blenders are more likely to be found in kitchens than beaches, and elephants are not generally found inside homes. Research over the past 40years has demonstrated that contextual associations influence object recognition, change eye movement distributions, and modulate brain activity. However, the majority of these studies choose object-scene pairs from experimenters’ intuitions because the statistical relationships between objects and scenes had yet to be systematically quantified. How do intuitive estimations compare to actual object frequencies? Across six experiments, observers estimated the frequency with which an object is found in a particular environment, such as the frequency of “mug” in an office. Estimated frequencies were compared to observed frequencies in two fully labeled scene databases (Greene, 2013). Although inter-observer similarity was high, observers systematically overestimated object frequency by an average of 32% across experiments. Altogether, these results speak to the richness of scene schemata and to the necessity of measuring object frequencies.