Abstract
Brady and Kersten (2003) showed that some adults could learn to recognize a novel object paired with a sound, at least when instructed to look for it, when it was shown repeatedly against varying static backgrounds (texture and shading of object and background were identical). Can infants segment a novel object paired with a sound when it is shown repeatedly on static backgrounds? If so, must the backgrounds vary and/or differ in texture and shading from the novel object? We measured infants' preferences for two novel 3D objects during a pre-test and then showed them one of the novel objects repeatedly against varying backgrounds. During familiarization the object was accompanied by a labeling phrase. At post-test, infants viewed the familiarization object and the other object on a white background while hearing the familiarization object's phrase. If infants segmented the object from the background during familiarization, their post-test looking times at the familiarization object should be longer then at pre-test. In Experiment 1, we showed a 3D object on varying 2D background patterns. Infants looked longer at the familiarization object, but not the other object, post-test minus pre-test, p <0.006, indicating they had segmented the object from the background during familiarization. In Experiment 2, we showed the same 3D object on 3D versions of the backgrounds with identical texture and shading as the object; the background differed on every trial. Evidence of segmentation was not observed at post-test, p > 0.44. In Experiment 3, we showed the same 3D object on a constant 3D background. Again, no evidence of segmentation was observed at post-test, p > 0.36. 5.5 month-old infants can segment without explicit instruction, at least when the background varies and differs in texture and shading from the object. An in-progress experiment attempts to separate these two factors.
Meeting abstract presented at VSS 2014