RESEARCH IN THE MYERS LAB

The Myers Lab conducts research on several topics, all loosely related to the theme of how the human brain organizes and stores information, and how it uses and generalizes that information later.  Some research topics are summarized below.

Computational models of the hippocampal region in learning, memory and generalization

Back in the early 1990s, Gluck & Myers used a computational model to explore what was then a fairly radical idea: that the hippocampal region (which lies in the medial temporal lobes in the human brain) is not just a passive memory store, but helps determine what and how information gets stored, by creating new representations that emphasize important or relevant information, while de-emphasizing irrelevant information.  These new representations then determine how learning can generalize to new situations.

Learning, memory and generalization

To develop this idea, Gluck & Myers developed a neural network model, that included a module representing the hippocampal region, which could interact with other brain structures, such as the cerebral cortex and cerebellum during learning.  In this model, the hippocampal region monitored environmental regularities and established new representations to encode these regularities; this information could then be passed on to other brain areas such as cortex and cerebellum, allowing generalization when irrelevant stimulus features changed, when the context changed, or when familiar stimuli were presented in new ways.  Without the hippocampal region, some simple learning could still occur, but without the hippocampal region, there would be little generalization of this new learning.  These model predictions addressed a range of existing data from animal learning studies, and made several novel predictions that were later confirmed by new empirical research.  They also drove a number of studies to examine learning and generalization in human patients with damage to the hippocampal region (see below).  The Myers Lab went on to develop these ideas further, including examining how the hippocampus interacts with the amygdala during emotional learning.

Individual differences in learning and generalization

More recently, the lab has begun to use model-fitting and error-minimization techniques to "customize" models to fit data obtained from individual human subjects, by determining what parameters (such as learning rate, tendency to explore new options, and tendency to perseverate on familiar responses) best describe each individual's pattern of results; this in turn is providing new insights into how individual differences in learning and memory can give rise to between-group differences in behavior.

Closeup on the dentate gyrus

In the last few years, a recent direction for the modeling work has been an ongoing collaboration with Helen Scharfman, at the Nathan Kline Institute and New York University, to examine how the anatomy and physiology of one hippocampal subregion, the dentate gyrus, could support pattern separation -- helping to develop representations that differentiate superficially similar stimuli.  The dentate gyrus is one of the few areas of the human brain where there is unambiguous evidence for adult neurogenesis (the ability to grow "new" neurons throughout life), and it is also a site of origin for seizure in many patients with epilepsy.  This work is suggesting new ways to understand neurogenesis, and may have implications for monitoring and detecting seizure in vivo.

Application to understanding human disorders of learning and memory

A major effort in our lab is developing simple tasks to assess memory and generalization, both in healthy individuals, and in individuals with memory deficits due to damage to the hippocampus and other brain areas.  Many of these tasks take the form of simple computer "games" in which participants learn how to respond to stimuli on the screen.  In many cases, these tasks test predictions of the computer model, allowing us to assess how well the model encapsulates key ideas about how the brain works. 

Amnesia

Individuals can sustain hippocampal damage in a variety of ways, including hypoxic brain injury (e.g. from carbon monoxide poisoning or cardiac/respiratory arrest), viral encephalitis, or stroke.  In many cases, these individuals develop a form of amnesia, or memory loss, in which older memories are largely spared but there is little or no ability to form new memories.  Surprisingly, these individuals can often still learn to perform well on the computer tasks but -- without hippocampal processing -- they are subsequently impaired at generalization when task demands shift.  Individuals who develop amnesia due to other types of brain damage, that spare the hippocampus, often show other behavioral patterns, which helps us to understand better what these brain areas do and how they interact.

Alzheimer's disease

The hippocampus and nearby structures are among the earliest brain areas to show pathology in Alzheimer's disease (AD).  It's now appreciated that this damage can begin to accumulate years or even decades before clinical symptoms of AD emerge.  Our work has shown that non-demented individuals with this early hippocampal pathology demonstrate generalization deficits on our tasks, and it is possible that this may be an early behavioral marker of incipient (presymptomatic) AD.  Part of our research interest is in determining whether these generalization tasks may be sensitive, inexpensive, and non-invasive tools to predict and track AD, before clinical symptoms emerge.  Such early detection would be very useful for aggressive intervention and for clinical trials as new therapeutic agents become available.

Addiction

Addiction can be defined as occurring when a person continues to seek out and use a drug, despite serious adverse consequences.  We are interested in understanding the brain changes that occur as a person transitions from drug use to drug addiction, and also in the individual differences that make one person more vulnerable to addiction than another.  Currently, a chief interest is in opioid drugs, including both illicit drugs (e.g. heroin) and also prescription painkillers.  These drugs are highly addictive, and carry high risk of relapse even after long periods of abstinence -- yet not everyone who uses these drugs becomes addicted.  Understanding why may help suggest new therapies to help people overcome their addiction.

Understanding post-traumatic stress disorder

Post-traumatic stress disorder (PTSD) is a disorder that can develop in individuals, including veterans, who have been exposed to traumatic events.  Better understanding of how these symptoms develop may lead to better therapy, as well as to possible tools to predict who is at most risk for PTSD, before clinical symptoms emerge.

Avoidance symptoms

A key feature of PTSD is that patients avoid reminders of the trauma; efforts to avoid can negatively impact quality of life if the individual is unable to participate in everyday activities for fear of encountering such reminders.  Our work is suggesting that this pattern of avoidance behavior may not be limited to learning about fear or trauma.  Rather, there may be some individuals who have a cognitive bias to respond to threat by avoidance, rather than by other coping mechanisms.  These individuals learn to avoid faster -- and maintain the behavior longer -- even on computer-based tasks, where there is no explicit reference to trauma or fear.  This in turn suggests new ways to understand avoidance behavior in patients with PTSD, which could lead to better therapeutic approaches, and better ways to assess the success of that therapy.

Re-experiencing

Another key feature of PTSD is re-experiencing, which can include flashbacks, nightmares, and intrusive memories of the traumatic event.  One interpretation of this phenomenon involves generalization, in which hippocampal dysfunction might lead memories to be stored with insufficient contextual tagging, so that the traumatic memories are easily triggered even in situations that should not normally evoke recall.  Again, this interpretation suggests that re-experiencing reflects a general cognitive bias, not limited to learning about fear or trauma.  Our recent studies with computer-based generalization tests suggest that, indeed, individuals with PTSD symptoms are more prone to generalize even in the context of a relatively innocuous computer game.  If this cognitive bias pre-dates exposure, rather than emerging only as a symptom of PTSD, it might provide a way to screen for individuals who are prone to develop PTSD in the wake of exposure to a traumatic event.

Predicting (and preventing) suicide

A tragic number of Americans die by suicide every year, and rates are particularly high among Veterans. Many people have suicidal thoughts or ideation, but only a fraction of those at-risk individuals actually engage in suicidal behavior. An important area of research in our lab is to better understand the cognitive processes by which at-risk individuals transition from ideation to action may suggest warning signs that can be monitored to determine when individuals are likely to attempt suicide, so that clinical resources can be better targeted to those individuals when they need them most. As part of this research, we are developing objective cognitive tests to monitor at-risk individuals, complemented by computational modeling to extract latent cognitive variables that may not be evident from behavior alone.  In addition to our goal of developing objective "warning signs" of upcoming risk for suicidal behavior, this research may also suggest novel routes for therapy to alter cognitive processes that lead to suicidal thoughts and behaviors.