As we struggle to pick out a familiar face in a busy crowd, identify objects on an X-ray image, or search for berries in a montage of leaves, it can sometimes feel a bit like playing Where’s Waldo? But, as in the game, the ability to identify a particular stimulus—seemingly invisible against a complex backdrop—usually becomes easier with experience or practice. This phenomenon is known as perceptual learning.
Previous studies exploring the neural mechanisms underlying perceptual learning have frequently thrown up conflicting results, with some researchers reporting increased brain activation within the primary visual cortex (V1) while others report no change. Now research from the lab of HMS assistant professor of radiology Yuka Sasaki, of Massachusetts General Hospital, has shed new light on these apparently contradictory findings—implying that previous studies could have been tapping into different stages of the same process.
Sasaki’s team reports in the March 27 Neuron the surprising finding that activity in the human V1 actually exhibits biphasic activity over the course of perceptual learning—increasing during the early phase then strangely disappearing, independent of behavior that continues to improve.
The team discovered this unusual effect by assessing brain activity at specific intervals over a prolonged period of training using functional magnetic resonance imaging (fMRI) while participants engaged in computer-based perceptual-learning experiments. The tasks required subjects to identify target stimuli (in a central or peripheral position) on a computer screen that was filled with horizontal bars. This test screen was repeatedly presented for a short period immediately followed by a blank and then a “masking” screen, composed of randomly oriented V shapes, which made perceiving the target stimuli on the test screen very difficult. Yet, with training, subjects soon became quicker at identifying the targets.
“It’s been known for some time that sensory-perception ability changes with practice, but the underlying neural mechanisms were not really known,” said Sasaki.
To investigate changes in brain responses, all participants were first scanned while performing the experiment, prior to any training, to get a measure of baseline V1 activity. Subjects were then scanned four times over a period of two weeks, with additional training outside the scanner in between scanning sessions.
By doing this, the researchers found that brain activity in V1 increased with initial learning, but then returned to the original level. Behavior, on the other hand, continued to improve even after changes in V1 had subsided.
“If neurons increased their firing more and more on repeated exposure to stimuli, eventually they would explode,” joked Sasaki. “Probably what happens is that during early perceptual learning, there are increased synapses between neurons, but then at some point there is a reorganization and a reduction in the number of inefficient synapses. Because fMRI cannot look at neural activity at the molecular level, this is still a hypothesis.”
