Even though the human visual system possesses sophisticated machinery for processing color, the brain effortlessly recognizes objects in black-and-white images. A new study from MIT provides a compelling explanation for this ability, suggesting its roots lie in early development.
The researchers combined experimental data and computational modeling to demonstrate that during infancy, when color information is limited, the brain learns to distinguish objects primarily through luminance, the intensity of light they emit. As the visual system matures and the retina and cortex become adept at processing colors, the brain incorporates this new information. However, the early proficiency in luminance-based recognition is retained, allowing for robust object identification even without color cues.
This finding aligns with previous research highlighting the unexpected benefits of initially degraded visual and auditory input for the developing perceptual systems. Professor Pawan Sinha, senior author of the study, explains, “This general idea, that there is something important about the initial limitations that we have in our perceptual system, transcends color vision and visual acuity. Some of the work that our lab has done in the context of audition also suggests that there’s something important about placing limits on the richness of information that the neonatal system is initially exposed to.”
The study, published in Science, also sheds light on why children who regain sight later in life, such as those treated for congenital cataracts, struggle to identify objects in black and white. These children, immediately exposed to a world of color, may develop an over-reliance on color, making them less adaptable to its absence.
This insight emerged from Project Prakash, an initiative in India founded by Sinha in 2005 to treat children with reversible vision loss, including those with congenital cataracts. Observations of these children revealed a significant drop in object recognition accuracy when presented with black-and-white images, unlike children with typical visual development.
To test their hypothesis, the researchers used AlexNet, a convolutional neural network, as a model of the visual system. They trained the network to recognize objects using different input strategies. One strategy mimicked typical development, starting with grayscale images and gradually introducing color, while another reflected the experience of Project Prakash children, receiving full-color input from the start.
The results were striking. The developmentally inspired model excelled at recognizing objects in both color and grayscale, demonstrating resilience to color manipulations. Conversely, the model trained solely on color images struggled with grayscale or hue-altered images. Lead author Lukas Vogelsang notes, “What happens is that this Prakash-like model is very good with colored images, but it’s very poor with anything else. When not starting out with initially color-degraded training, these models just don’t generalize, perhaps because of their over-reliance on specific color cues.”
Crucially, the order of visual input mattered. A model trained on color images before grayscale did not achieve the same proficiency in black-and-white object recognition. Sinha emphasizes, “It’s not just the steps of the developmental choreography that are important, but also the order in which they are played out.”
Analysis of the models revealed that early exposure to grayscale forced them to rely on luminance for object identification, a strategy that remained effective even with the introduction of color. Models trained initially on color struggled to adapt when presented with grayscale.
This suggests that the human brain, highly plastic in early life, benefits from the initial limitations in color vision. As retired neurologist and study author Sidney Diamond observes,
As a newborn, the normally sighted child is deprived, in a certain sense, of color vision. And that turns out to be an advantage.
Sinha’s lab has identified similar advantages of limited sensory input in other areas, such as auditory processing. Their research, funded by the National Eye Institute and the Intelligence Advanced Research Projects Activity, suggests that early limitations may be crucial for developing robust and adaptable perceptual systems, a concept they plan to explore further in areas like language acquisition.
Responses (0 )