jiyeongha / midrb

[Behavioral] The effects of spatiotemporal continuity and mid-level perceptual features of objects on Repetition Blindness

midRB

This repository includes codes for experiments and analyze midRB data (Sorry, the figures are described in Korean!). The texform stimuli are created using a ventral pRF model (Freeman & Simoncelli, 2011; Freeman et al., 2013) in a similar way Long et al (2017) did.
Find the abstract below the figures.

Fig 1. Procedure

Fig 2. Experimental design

Fig 3. Stimulus creating process

Fig 4. Stimulus comparison between Long et al.(2017) and Ha & Yi (2017)

The effects of spatiotemporal continuity and mid-level perceptual features of objects on Repetition Blindness

The visual system has little trouble recognizing the same object even though the retinal image of that object keeps changing due to its orientation, illumination, distance, etc. One of the strategies to maintain such stable object representations is considering spatiotemporal continuity of the object with its visual properties. Spatiotemporally connected objects are likely to form a stable representation as the same object despite having different physical appearances. Conversely, various representations an object has can also help recognizing the same object even when one perceives spatiotemporally separated objects. We sought to investigate whether the effects of spatiotemporal continuity on object recognition is modulated by the dimensions of representations an object has using Repetition Blindness paradigm. Repetition Blindness(RB), one of visual phenomenon related to object recognition refers to difficulty in recognizing a repeated object in rapid serial visual presentation(RSVP). Previous study (Kanwisher, 1987) indicates RB can occur when the observer fails to tokenize for an object presented twice. Adding to this, we attempted to specify the visual properties interacting with spatiotemporal continuity by manipulating mid-level perceptual similarity between two categories (Long, Störmer, & Alvarez, 2017). In this study, we used objects with semantic information as stimuli with various representations. Mid-level perceptual feature and Pattern stimuli, generated based on V2 and V1 receptive field models respectively (Freeman et al., 2013), were used as pseudo-objects having only visual representation. In experiment 1, we replicated a previous study (Long et al., 2017) to assure Mid-level perceptual feature stimuli we used have similar perceptual features within categories. As a result, only in visual search tasks using stimuli containing mid-level perceptual feautures we observed higher visual search efficiency in different-category condition than in same-category condition. In experiment 2, we compared the effect of spatiotemporal continuity on RB size of semantic objects to that of pseudo-objects and tested whether RB also occurs for mid-level perceptual feature stimuli in same-category condition. Results showed that in pseudo-obect condition, RB size was significantly bigger when targets were spatio-temporally presented than when targets were not. On the other hand, RB sizes for semantic objects were not different when targets were spatio-temporally presented and when targets were not. We could not observe significant RB in same-category condition both for mid-level perceptual feature and pattern stimuli. In conclusion, this study suggests objects with small variance in dimensions of representations tend to require spatiotemporal continuity in order to be recognized as the same object while objects with various representations can be easily recognized as the same object without spatiotemporal continuity.