Overview of Current Research Program

Mark A. Gluck is a Professor of Neuroscience at Rutgers University-Newark, co-director of the Rutgers Memory Disorders Project, and publisher of the public health newsletter, Memory Loss & the Brain. For the last twenty-five years he has worked at the interface of psychology, neuroscience, and computer science, where he has been developing and testing computational neural-network models of animal and human learning.

Gluck’s research focuses on two main problems: How are new associations formed? And How do we apply past learning to novel situations? The work in Gluck’s lab spans three interdisciplinary axes, integrating across (1) animal and human learning, (2) brain and behavior, and (3) experimental and clinical perspectives. Gluck’s research starts, conceptually, at the behavioral level, where he and his collaborators map mathematical theories of classical conditioning to cognitive models of human learning. At the same time, they have shown how many of these same theories of animal learning provide a framework for interpreting and informing empirical studies of the neural bases of classical conditioning, especially the contributions of the cerebellum, basal ganglia, and hippocampus. These two interdisciplinary links—from animal to human learning and from behavioral theories to neural substrates—converge in providing Gluck and colleagues with the tools to study, and formally model, how different brain regions interact during learning. In collaboration with numerous colleagues and collaborators, Gluck utilizes a diverse array of scientific techniques including behavioral studies of clinical patients, structural and functional brain imaging in healthy normal individuals, and human behavioral genetics.

The clinical studies fall into two categories. The first focus on patients with damage to their fronto-striatal circuits due to Parkinson’s disease, dystonia, frontal strokes, or fronto-temporal dementia (FTD), also known as Pick’s disease. These clinical studies have led to novel insights into how people learn from the rewards (and punishments) that provide error-correcting feedback about recent behavioral choices and decisions, and how the neuromodulator dopamine is critical to this process. In turn, this research has provided potentially important clinical insights into how medications for Parkinson’s disease alter cognitive function. Other work has suggested novel behavioral training programs that can remediate some cognitive deficits in Parkinson’s disease by encouraging the recruitment of brain regions not damaged by the disease.

The second group of patients studied by Gluck and his long-time collaborator, Catherine Myers, are characterized by dysfunction to their medial temporal lobe, especially the entorhinal cortex and hippocampus. These include hypoxic (oxygen deprivation) patients and others with global anterograde amnesia, non-demented elderly with mild hippocampal atrophy, and patients with amnestic forms of Mild Cognitive Impairment (MCI), a likely early stage of Alzheimer’s disease. Studies in these patient groups have elucidated the role of the medial temporal lobe in forming appropriate stimulus representations during learning. Gluck and Myers have argued that hippocampal-dependent changes in representation are key for the effective transfer (that is, generalization) of past learning to future novel task demands. This research has also provided some clinically important outcomes, including novel behavioral assessment tools that may aid in the early diagnosis and detection of Alzheimer’s disease as well as in the assessment of the efficacy of new therapeutic treatments.

In both the fronto-striatal and medial-temporal-lobe programs of research, the findings from Gluck’s clinical studies are supported by converging parallel studies using functional brain imaging in healthy normal adults. Other converging data come from studies of the behavioral implications of individual variability in genes associated with different neuromodulators.

While the focus of Gluck lab is on human learning research, parallel animal studies provide important comparisons to the human research. These animal studies include lesion, drug, and electrophysiological studies in rats as well as behavioral and pharmacological studies of transgenic mice. In general, these conditioning tasks are logically equivalent in structure and solution to the discrimination and categorization tasks studied by Gluck and colleagues in their human research. This animal research may also lead to clinically relevant outcomes, especially the development of novel pre-clinical animal tests to assist in the discovery and evaluation of new therapeutic compounds for Alzheimer’s disease and Parkinson’s disease.
The computational models developed through Gluck and Myers’s program of integrated behavioral and neuroscience research were primarily intended as tools to understand how the brain works. Nevertheless, these models may lead towards novel biologically-inspired cognitive architectures for artificial intelligence, machine learning, and cognition. With support from the Office of Naval Research (ONR) and the Defense Advanced Research Projects Administration (DARPA), Gluck and colleagues have applied their brain models to practical problems including sonar classification, the detection of mechanical faults and other anomalies in helicopter gear boxes and submarine pumps, and the development of new integrated architectures for autonomous robots.