Abstract
When observing a simple visual scene, such as an array of dots, observers can easily and automatically extract their number. How does the visual system accomplish this feat? We tested the degree to which number perception relies on real-world experience by testing how well participants estimate the number of objects embedded within real scenes, such as those in Figure 1 (e.g., "How many chairs are on this beach?"). In Experiment 1 (N = 25), participants estimated either the number of black dots on a white background or the number of objects shown within a mix of indoor and outdoor scenes (Figure 1). Surprisingly, we found that participants were severely impaired at estimating the number of objects in real-world scenes, showing four times the estimation error compared to the dot stimuli. In Experiment 2 (N = 48), we investigated whether this impairment is simply caused by the irrelevant background information drawing attention away from the objects. Participants estimated the number of objects in scenes whose background was either faded to grayscale or entirely removed, thereby allowing participants to easily locate the objects (Figure 1). Although participants performed marginally better when the background was entirely eliminated, we still found that number perception is severely impaired in scenes compared to dot stimuli in either condition. Finally, in Experiment 3 (N = 50), we found that participants showed no impairment when estimating the number of objects that were randomly positioned on the screen, suggesting that the impairment in Experiments 1 and 2 stems entirely from the actual distribution of objects in the real-world scenes, not from the identity of the objects themselves. In conclusion, we find that – despite having decades of experience in scene perception - observers are surprisingly poor at estimating the number of objects embedded within real world scenes.
Meeting abstract presented at VSS 2018