General questions about our research

Why do theoretical neuroscience?

The state of modern neuroscience has often been likened to the state of physics in the 16th century. At the time, the trajectories of stars and planets in the sky were well-documented, but nobody could explain why those numbers were the way they were. It wasn't until Kepler and Newton that it was understood that planetary motion was governed by the gravitational force exerted by the Sun, and equations were found that could not only capture all trajectory data concisely and precisely, but could also be applied to other phenomena. In neuroscience today, the situation is somewhat similar. There is a wealth of data from physiology, psychophysics, and imaging, but conceptual and rigorous frameworks that make sense of them are few and far between. Building those frameworks is the task of theoretical neuroscience.
 

Why use the term "computation"?

Computation as in "computational neuroscience" is used in roughly two meanings. One is the use of analytical and numerical techniques to analyze experimental data. This is done extensively by some of the other laboratories in the Computational Psychiatry Unit. In the other meaning, the brain itself is regarded as a computational device that we are trying to understand the workings of. For example, one of the most basic functions of the brain is to process sensory information in order to produce motor actions. If you are driving on the freeway and want to change lanes, you need to collect many pieces of information: how far away you are from the car in front of you, how fast you are going, whether there is a car beside you or approaching you from behind, etc. All pieces are combined to make the appropriate decision. This is an example of neural computation: the manipulation of (sensory, internal, reward) information in the brain, aimed at generating behavior. Because of the double meaning of the term "computation", some people who work on the second meaning prefer the term "theoretical neuroscience" for what they are doing.
 

Why work at the systems/cognitive level?

Neuroscience has many fascinating levels of study, and which ones are of greatest interest is to a large extent a matter of taste. However, the big questions of the field have commonly revolved around the interplay between behavior (including perception, cognition, and social behavior) and neural activity. Is it possible to explain perception, memory, learning, movement, etc. in terms of the concerted activity of many neurons? Addressing such questions rigorously requires a formalization of human behavior as well as a theory of neural coding. It often also requires narrowing down "behavior" to a specific type of well-defined and experimentally controllable behavior. In my laboratory, we try to apply our theories to as wide a range of cognitive phenomena as possible. Currently, we study multisensory perception, decision-making, visual search, and visual short-term memory. The ultimate goal is not only to explain perceptual behaviors, but also to explain or even predict the neural mechanisms that underlie these behaviors.
 

Why study uncertainty?

Only rarely does the information that humans or animals need to extract from their environment come in clear-cut and ready-made pieces. The generic situation is that sensory information is imperfect, scenes are full of distractors, and the same input can be interpreted in multiple ways. Think for instance about detecting a predator in a cluttered scene, passing a basketball to your teammate while all players are in motion, or recognizing a friend from a distance. In each case, the necessary data come with a lot of uncertainty ("big error bars") and the brain can only infer, rather than know precisely, the state of the world. The idea that perception is unconscious probabilistic inference dates back to Helmholtz, but has only relatively recently begun being investigated in earnest. The ways in which the brain deals with uncertainty have the potential to teach us about the neural mechanisms of perception. In domains like multisensory perception or visual search, there is an abundance of psychophysical data that, when combined with theories of neural coding and physiological data, greatly contribute to our understanding of neural computation.
 

Why join this lab?

My lab works on the interface between systems neuroscience and behavioral studies of perception. We attack traditional problems in cognition from a new perspective: that of neural coding, in particular population coding. Therefore, you will work from a big-picture view of the field, while acquiring specific skills in visual psychophysics, behavioral modeling, and neural modeling. You also will be part of a greater community of enthusiastic researchers. If you have an interest in neuroimaging or investigating psychiatric disease, opportunities are literally next door. Collaborations also exist with laboratories elsewhere that study sensory perception and neural computation. To join the lab, a background in quantitative thinking is strongly preferred.
 

Why come to BCM for computational cognitive neuroscience?

In BCM's Department of Neuroscience, computational work is cutting-edge and diverse. It includes  investigating the neural mechanisms of time perception, modeling reward responses in psychiatric patients, studying the dynamics of decision-making in groups, advanced analysis of neuroimaging data, and developing probabilistic theories of human perception. Each of the computationally oriented neuroscience laboratories also conducts human behavioral or neuroimaging studies, thereby allowing for a direct and fruitful interplay between theory and experiment. This is aided by state-of-the-art imaging and computing facilities, as well as active collaborations with neurophysiological laboratories and clinical divisions. Students of this new approach will acquire the tools and training that will position them uniquely for groundbreaking interdisciplinary research after the completion of their program, as well as for many other career paths.