Purchase this article with an account.
Colin S Flowers, Rachel M Skocypec, Mary A Peterson; Does Semantic Activation Affect Human Object Detection in Natural Scenes?. Journal of Vision 2019;19(10):58a. doi: https://doi.org/10.1167/19.10.58a.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We showed (VSS 2018) that human participants can detect whether a flickering colored dot probe near an object’s border is located “on” or “off” the object in masked 100-ms exposures of color photographs (mean d’ = 0.97); performance was better for central than peripheral locations. The current study paired a subset of those photographs with flickering black/white probes (to improve peripheral probe visibility) and tested whether semantic activation from a word prime shown before the photographs could improve d’. Before each photograph, an unmasked prime was presented for 140 ms. The photograph followed after a 100 ms ISI. Three prime conditions were tested: Neutral (‘XXXXXXX’), Match (the basic level name of the object [e.g., ‘zebra’]), and Mismatch (the name of an object from a different category [natural/artificial; e.g., ’bowl’ was the mismatch prime for zebra]). As before, “on” responses for dots located on an object were considered HITs; “on” responses for dots located off the object were considered FAs. Detection sensitivity was again greater for central than peripheral objects (d’ = 1.242 vs 0.851; p < 0.001). In addition sensitivity was greater for photographs with natural than artificial objects (d’ = 1.374 vs 0.719; p < 0.001), consistent with evidence that natural scenes are processed faster than artificial scenes (Rousselet, et al, 2005). Prime type did not significantly affect d’ scores, but did affect criterion, which was more liberal following Match than Mismatch primes (−0.227 vs 0.008; p = 0.036). We take this criterion difference to reflect guesses based on the feasibility of the object in the scene. Thus, semantic priming does not affect object detection in natural images under these conditions. We plan to analyze the photographs individually to assess how object vs. scene-wide characteristics (e.g., object size, amount of crowding in the scene) influence performance in this task.
This PDF is available to Subscribers Only