Samantha M. W. Wood and Justin N. Wood

Controlled-rearing studies provide the unique opportunity to examine which psychological mechanisms are present at birth and which mechanisms emerge from experience. Here we show that one core component of visual perception—the ability to parse objects from backgrounds—is present when newborn animals see their first object. We reared newborn chicks in strictly controlled environments containing a single object on a single background, then tested the chicks’ object parsing and recognition abilities. We found that chicks can parse objects from natural backgrounds at the onset of vision, allowing chicks to recognize objects equally well across familiar and novel backgrounds. We also found that the development of object parsing requires motion cues, akin to the development of object parsing in human infants and newly sighted blind patients. These results demonstrate that newborn brains are capable of “one-shot object parsing” and show that motion cues scaffold object perception from the earliest stages of learning. We conclude that prenatal developmental programs build brain architectures with an object-based inductive bias, allowing animals to solve object perception tasks immediately without extensive experience with objects. We discuss the implications of this finding for developmental psychology, computational neuroscience, and artificial intelligence.

The experimental design. (A) During the Input Phase, the chicks were raised with a single virtual object moving on a single background. (B) During the Test Phase, the chicks were presented with test trials in which the imprinted object appeared on one display wall and a novel object appeared on the opposite display wall. The timelines show how the virtual objects were presented on the two display walls during sample 4-hour periods during the (C) Input Phase and (D)Test Phase. Across the test trials, the imprinted object and novel object were presented on the familiar background and on novel backgrounds, in a fully crossed design. The object was also presented from familiar viewpoints (0° azimuth rotation) and novel viewpoints (30° & 60° azimuth rotations) on different test trials. This schematic illustrates examples of each background test condition; the actual order of the test trials was randomized across the experiments.

Experiment 1 Results. (A) Recognition performance in the four background conditions. (B) Recognition performance in the three viewpoint conditions. The chicks’ performance (percent time spent with the imprinted object versus novel object) was well above chance level in all test conditions. ***denotes p < .001 (C) Recognition performance in each background condition on each test day. While performance improved across the Test Phase, the chicks successfully recognized their imprinted object on all test days, including Day 1 of testing. (D) Individual subject performance. The graphs show the percent of time each chick spent with the imprinted object versus the novel object. Each chick is represented by a gray marker. The blue boxes indicate the 25th to 50th percentile and the 50th to 75th percentile of performance. Dashed lines indicate chance performance. Error bars show ±1 standard error (SEM).

Experiment 2 Results. (A) Sample frames from the object displays in Experiment 2. The object rotated 360° at a rate of 1°/min. The chicks in Experiment 2 saw all of the same images as the chicks in Experiment 1, but without clear motion cues. (B) Recognition performance in the four background conditions. (C) Recognition performance in the three viewpoint conditions. * denotes p < .05 (D) Recognition performance in each background condition on each test day. Performance remained low and stable across the test days. (E) Individual subject performance. The graphs show the percent of time each chick spent with the imprinted object versus the novel object. Each chick is represented by a gray marker. The blue boxes indicate the 25th to 50th percentile and the 50th to 75th percentile of performance. Dashed lines indicate chance performance. Error bars show ±1 standard error (SEM).

Video Block
Double-click here to add a video by URL or embed code. Learn more

SI Movie 1. Input Phase

SI Movie 2. Test Phas