In a study appearing in the Nov. 29 issue of Nature, the researchers describe how increased activity in one brain system is associated with decreased activity in another system during learning of a simple skill. The findings ? which suggest how the brain mediates between the need to store and access a wide range of information and the need for virtually automatic responses in key situations may eventually lead to new strategies for dealing with learning disorders or for diagnosing Alzheimer's disease, Parkinson's disease and other brain disorders.
Previous studies have identified several brain structures that are key to learning and memory and have suggested associations with particular learning tasks. The medial temporal lobe of the brain, which includes a structure called the hippocampus, has been associated with what is called declarative learning, the learning of facts and events. An area called the basal ganglia, deep within the brain, has been associated with nondeclarative learning, learning based on experience that may not involve conscious memory. In the current study, conducted at the Martinos Center for Biomedical Imaging located at the MGH, healthy volunteers were given a simple learning task while undergoing functional MRI scans, which reveal the level of activity in various areas of the brain.
"We have been studying how the brain changes when people acquire skills, which brain systems support particular kinds of learning, and how those systems interact," says Russ Poldrack, PhD, of the MGH Department of Radiology and the Martinos Center, the paper's first author. "The idea that these systems may compete with each other was suggested by animal research, and we wanted to see if this takes place in humans as well."
In the first phase of the current study, volunteer participants learned to associate certain combinations of symbols with particular weather patterns in one of two ways: either they were presented with symbol-bearing cards, asked to indicate which type of weather the cards were associated with and then told the correct answer, or they were simply presented with the cards and told which type of weather they signified. The first version basically a learn by trial-and-error test was designed to test nondeclarative memory, while the second version, in which associations were memorized, was designed to involve declarative memory.
Functional MRI images taken during these tasks revealed increased basal ganglia activity and reduced medial temporal lobe (MTL) activity during the feedback-based, trial-and-error version of the test. The memorization version produced the opposite pattern, with increased MTL activity and lower basal ganglia activity.
A second experiment presented a different group of volunteers with the feedback-based version of the weather prediction test in a way that could follow over time any changes occuring in brain activation. The overall pattern was similar to that seen in the first experiment ? decreased MTL activity and increased basal ganglia activity. But the time-based analysis showed that at the very earliest stages of the experiment the MTL was active and the basal ganglia inactive, with the activation pattern switching as learning proceeded.
Mark Gluck, PhD, of the Center for Molecular and Behavioral Neuroscience at Rutgers-Newark, the study?s last author, developed the weather prediction task with his Rutgers colleagues as a way to study how people learn categorization rules. He explains that traditionally the MTL and especially the hippocampus had been regarded as being involved with declarative memory only. In contrast, Gluck and Catherine Myers, PhD, also a coauthor, have developed an alternative theory: that the hippocampus is involved in all learning and is responsible for determining how new information is encoded by other brain regions. Gluck explains, "The current results provide the first functional neuroimaging data to support our theory. As our models predicted, the hippocampus was activated in the earliest stages of learning, when we expect new encodings to be established, but not in later learning when the encodings are used by other brain structures, such as the basal ganglia." Gluck and Myers? theory is detailed in their recent book Gateway to Memory: An Introduction to Neural Network Modeling of the Hippocampus and Learning.
Gluck also notes that the hippocampus and related structures are damaged early in the process of Alzheimer?s disease and that the current findings may allow development of learning-based tests that could diagnose that disorder in its earliest stages. Poldrack adds that better understanding of how the brain?s memory systems interact could lead to a better understanding of brain plasticity, the ability of brain areas to take on functions usually accomplished by other structures. "We?re most excited about the possibility of using plasticity to solve problems like dyslexia."
The Rutgers components of the study were conducted in Gluck?s lab at the Center for Molecular and Behavioral Neuroscience at Rutgers, the State University of New Jersey, campus at Newark. Gluck is co-director, with Myers, of the Memory Disorders Project at Rutgers-Newark, which promotes increased understanding of human memory and memory impairing disorders through a wide range of research.
Poldrack is principal investigator of the Skill and Language Research Lab at the Athinoula Martinos Center for Biomedical Imaging. A collaboration among the NMR Center at MGH and the Division of Health Sciences and Technology at Harvard Medical School and Massachusetts Institute of Technology, the Martinos Center builds on pioneering MGH imaging research, including the development of high-speed MRI scanning, showing both the function and structure of the brain and other organs.
Other co-authors of the Nature paper are Jill Clark and Juliana Paré-Blagoev, of MGH and the Martinos Center; and Judith Creso Moyano and Daphna Shohamy of Rutgers. The research was supported by grants from the Alafi Family Foundation, the Athinoula Martinos Center for Biomedical Imaging, and the National Science Foundation.