Home

[Teaching | Lab wiki]

What are we interested in?  Here’s an example: you’re in a cheese store, trying to decide which cheese to get. You’re offered a piece of cheese, and you like it, perhaps you might buy that one. But you’re not sure yet. After a couple of seconds you try a piece of another. You compare the two and decide to buy one of them. What happened in your brain as you went through all this? What are the neural mechanisms that allow you to remember, for a few seconds, how much you liked the first one; to compare the two; to make a decision; to apply the rules of behavior appropriate for the context you’re in (here, a cheese store)? In other words, what are the neural mechanisms underlying our cognitive abilities?

What methods do we use in our research?
We use a combination of computational, behavioral, electrophysiological, pharmacological, and optogenetic techniques. We train rats and mice to perform tasks that require cognitive components that we’re interested in studying. For example, we train them to remember a stimulus for a few seconds, and to then make a behavior based on their memory of the stimulus. We can then study neural responses during this behavior, and observe the neural correlates of short-term memory. To help us understand the mechanisms behind our findings, we build computational models of networks of spiking neurons, with which we explore the circuit architectures and mechanistic hypotheses that could explain the experimental results. The models both give us greater insight into potential mechanisms, and help us decide what are the best next experiments to test and distinguish between hypotheses.

Who are we? Personnel in the lab range from purely computational to purely experimental. We try to minimize the barriers in going from one end of this spectrum to the other: all researchers in the lab are encouraged to flow easily and freely within the computational/experimental spectrum, according to their own interests and needs at any given point in time, and to talk frequently with people at other points of the spectrum.

The specific tasks we are studying. Here are three examples of tasks that we use, all studied from the above combined computational/experimental approach.

The first example focuses on decision-making. Behavior in many different kinds of decisions (including value-guided decisions, memory-guided decisions, and sensory-guided decisions) is well fit by a model in which evidence for or against each option in the decision gradually accumulates over time, until it either reaches an internal threshold or the decision-making moment is imposed externally. How does this accumulation happen, i.e., what are the biophysical circuit mechanisms underlying evidence accumulation? The task we developed to study this in rats is described in Brunton et al. 2013 in the Publications list.

The second example focuses on executive/cognitive control. In this task, we study how it is that cognitive state can flexibly determine appropriate rules of behavior. Rats experience two types of trials, “Pro” and “Anti.” In Pro trials, they must orient towards a light to obtain a water reward. In Anti trials, they must orient away from a light to obtain a water reward. Which type of trial they will be in is indicated to them by a sound before the trial starts; this sets the cognitive stage. Then, when the light comes, the rats must use their knowledge of whether it is a Pro or Anti trial to select which of two opposite sensorimotor rules is the appropriate one to follow. Details regarding the ProAnti task-switching behavior can be found in Duan et al. 2015 in the Publications list.

The third and final example focuses on working memory, and is a task in which rats are presented with a first stimulus; then there is a pause; then they are presented with a second stimulus; and the rats must compare the two stimuli and make a binary decision based on the comparison. (This is our rodent analogue of deciding between the two DVDs in the music store.) We use this task to study short-term memory and decision-making. (See Akrami et al., 2018 in the Publications list).