This work is the result of a fantastic multi-institution collaboration, led by Han Yi, with @bchandra_pitt, Human Brain Research Lab @uiowa, @NeurosurgUCSF, @ChangLabUCSF! 11/11

Some neural populations change (though not all in the same way), while others hold onto their existing patterns. This provides the balance that allows learning without catastrophic forgetting. 10/11

Together, these results illustrate a key mechanism of what’s known as the stability-plasticity tradeoff – when learning something new, don’t change things too much so that you forget the old stuff. 9/11

i.e., neural populations that change activity as a function of learning have some intrinsic properties that make them amenable to change. 8/11

We also found that we could predict which populations would be learning populations based on passive listening to Mandarin sounds *before* training. 7/11

Strikingly, learning populations didn’t all just show increases in activity as learning progressed. Instead, we saw a diverse set of stimulus-specific changes, including increases and decreases. 6/11

These “learning populations” (red) were distinct from populations that changed their activity only based on the number of exposures to a given stimulus (independent of behavior; blue). 5/11

We found neural populations throughout temporal and frontal cortex that tracked *trial-by-trial* performance as participants learned. 4/11

While native English listeners learned to identify these sounds in Mandarin, we recorded activity directly from their brains using implanted electrode arrays. 3/11

Sound on for a brief explanation of the problem: 2/11

Why is it so challenging for adults to learn a new language? Excited to announce our latest paper on neural changes during speech learning, out now in @PNASNews. https://t.co/Mm0AJvJxoN 1/11