Summary: A new UCLA-led study shows that repetitive practice not only sharpens skills but also drives lasting changes in the brain’s working memory circuits. By training mice to remember sequences of odors and tracking brain activity over days, researchers observed that initially unstable memory patterns gradually stabilize into persistent representations in the secondary motor cortex.
The study improves our understanding of how practice refines neural representations and may inform approaches to memory-related disorders. The team used a custom-built, high-throughput microscope to monitor activity across tens of thousands of neurons simultaneously.
Key Facts:
- Memory Solidification: Repeated practice leads working-memory patterns to stabilize over time.
- Advanced Imaging: Researchers performed volumetric calcium imaging of up to 73,307 neurons (reported approximately as 73,000) to follow population-level changes.
- Clinical Relevance: Understanding how representations crystallize with practice could guide strategies for treating memory impairments.
Source: UCLA
Overview
Researchers from UCLA Health, in collaboration with Rockefeller University, investigated how working memory evolves with practice. Working memory is the brain’s ability to briefly hold and manipulate information, and it underlies many cognitive tasks. The team published their findings in the journal Nature, studying how repeated training alters neural dynamics associated with memory.
In the experiment, head-fixed mice learned an olfactory delayed-association task in which they had to identify and remember the sequence of two odors separated by a five-second delay. The researchers tracked how neural activity changed as the animals practiced the task over roughly two weeks, moving from naïve performance to expert levels.
The team used a custom volumetric calcium microscope capable of imaging cellular activity across large cortical volumes. This approach allowed simultaneous recording from very large neuronal populations, revealing population-level reorganization that single-plane imaging can miss.
As the mice learned, memory-related activity in the secondary motor cortex (M2) changed markedly. During early learning, working-memory representations in the late-delay period were volatile and drifted across days. With continued practice and the transition to expert performance, these late-delay representations became increasingly consistent—what the authors describe as a crystallization of working-memory patterns.
The causal importance of M2 for this task was supported by optogenetic perturbation: inhibiting secondary motor neurons during late-delay and choice periods impaired task performance, indicating that M2 activity is necessary for maintaining and using the remembered information during the decision phase.
Not all regions showed the same transformation. While M2 developed stable late-delay representations with practice, primary motor cortex (M1) and retrosplenial cortex (RSA) did not show comparable improvements in decoding accuracy for working-memory content. Stimulus and choice signals stabilized earlier in training, whereas delay-period representations required continued practice to become robust.
Dr. Peyman Golshani, corresponding author and UCLA Health neurologist, described the change using a musical metaphor: individual neurons are like notes, and the neural pattern—the melody—became more refined and consistent as animals practiced, explaining why behavior becomes more accurate and automatic with repetition.
About this memory and neuroscience research news
Author: Will Houston ([email protected])
Source: UCLA
Contact: Will Houston – UCLA
Image: Credit to Neuroscience News
Original Research: Open access. “Volatile working memory representations crystallize with practice” by Peyman Golshani et al., published in Nature.
Abstract (condensed)
Working memory transiently maintains and manipulates information and is vital for many cognitive functions. The mechanisms that produce and reshape working-memory representations across populations of neurons over days are not fully understood. To address this, head-fixed mice were trained on an olfactory delayed-association task requiring decisions based on the sequential identity of two odors separated by a five-second delay. Optogenetic inhibition of secondary motor cortex during late-delay and choice epochs strongly impaired performance, underscoring M2’s role. Mesoscopic calcium imaging across M2, retrosplenial cortex, and M1 revealed that many late-delay-selective neurons emerged in M2 during learning. Working-memory decoding accuracy during the late delay improved substantially in M2 but not in M1 or RSA as animals became experts. During the early expert phase, late-delay representations still drifted across days even though stimulus and choice signals had stabilized; however, volumetric imaging of up to 73,307 M2 neurons—including deep superficial populations—showed that continued practice ultimately stabilized the late-delay working-memory representations. In short, delay- and choice-related neural activities essential for working-memory performance drift across learning and only crystallize after several days of expert practice.