| Summary |
Imagine yourself in the corner of a dark room trying to see a dim flickering light at the other end of the room. Although you would probably perceive the light most of the time, you may occasionally make mistakes. We will, of course, blame our brains for our imperfect performance. However, the reason why we sometimes make errors, even in the simplest tasks, could stand behind a fundamental principle of neuronal information processing. Thus, the brain is not a passive camera-like device that records every event in the environment, but is an active device interpreting rather than simply reporting external events. Understanding this interpretative act requires knowledge about what happens in the brain in the absence of stimulation. Indeed, the brain is never at rest. Even in the absence of sensory inputs, brain cells spontaneously fire electrical impulses to make neuronal networks wander through different states. It is these intrinsic fluctuations that filter our perception of the world to make us see things differently depending on the brain's internal context. For instance, if a stimulus is presented at the 'wrong' time, when neuronal fluctuations are very large, the stimulus may not be distinguished from internal noise, hence it will most likely go unnoticed.
The idea that the brain has an internal state, or context, is not new. About 60 years ago, when digital computers were not available, psychologists and physiologists were striving to measure variations in brain rhythms while humans were performing specific behavioral tasks. The motivating question was whether the brain's internal state is able to shape sensory processing. One example is the alpha rhythm, which is measured over the occipital region. This rhythm is fairly regular, but waxing and waning, and has a dominant frequency around 10 Hz when the person is awake and relaxed. Although attempts to view the alpha rhythm as a cycle of neuronal excitability that allows us to see better in the 'high' state and to see worse in the 'low' state led to inconsistent results, they opened up the intriguing possibility that the brain's internal state could in fact influence perception. However, the technique available at the time, i.e., electroencephalography (EEG), although allowed to record voltage signals from the scalp, was unable to explain the origin of the recorded signals. Furthermore, EEG signals represent average changes in electrical field over large brain regions, hence they lack the degree of spatial resolution required to truly understand the relationship between neuronal internal states and behavior. The recent development of the technique of multiple electrode electrophysiological recording in behaving animals has made possible experiments addressing the relationship between neuronal internal states and behavioral performance. Using this technique we plan to study how the internal state of neuronal networks in the monkey brain influences how individual cells encode information to control behavior.
The focus of our studies is the visual cortex, which is the brain region that allows us to perceive the world. When we look at a visual scene, cells in the visual cortex respond to information streaming in from millions of "wires" that carry a pixilated image of the world to construct an internal representation. At the first stages of cortical processing, primary visual areas create a fragmented picture of the world, dominated, for instance, by small oriented lines highlighting edges. This representation is subsequently passed on to a higher functioning level of the visual cortex, where neurons typically respond to more complex image features, such as shapes and objects. One of the strengths of this hierarchical model of visual processing is the focus on the spatial representation of images. Clearly, it is the hierarchical model that helped realize that images have a distributed representation across many cortical areas, and that these representations need to interact among each other in order to create a crystal-clear percept of the world. However, one problem with the hierarchical model is the fact that it ignores the temporal aspects of visual processing, despite overwhelming evidence that visual perception is dynamic, not static. Hence, constructing a cortical representation of the world requires not only knowledge about the physical aspect of incoming stimuli, but also information about the state of neuronal networks involved in stimulus processing. It is the interaction between the neuronal response to a stimulus and the neuronal state or context what defines a real-time cortical representation. Understanding how neuronal context interacts with incoming stimuli to produce behavioral responses is essential to uncovering the basic principles of cortical function.
There are at least two ways in which the state of the cortex can change: (1) by intrinsic fluctuations in spontaneous, or ongoing, activity, and (2) by fluctuations in the behavioral state of the organism, possibly due to changes in the level of alertness, arousal, or attention. Whether and how these types of statedependent changes in neuronal activity have any functional significance is unknown. We plan to address these issues by combining visual behavior and physiology. That is, we will directly measure neuronal activity in visual cortex when monkeys perform specific visual tasks that allow to control both the visual stimuli and the behavioral state of the animal. The choice of rhesus monkeys for the proposed research is based on several important advantages offered by this species, such as visual performance that rivals that of humans in many respects. There is a growing body of evidence that the visual cortical machinery of rhesus monkey closely parallels that of humans, hence the results of our research will be directly relevant to humans.
We will pursue two lines of experiments that may bring us closer to understanding what happens to our brain as we experience the world. First, do individual neurons in the visual cortex encode similar stimuli differently depending on the level of spontaneous, or ongoing, activity? If they do, how does statedependent coding of visual information affect perceptual discriminations? Second, how does neuronal context change when the behavioral state of the animal is altered? Behavioral state can possibly change ongoing activity in visual cortex through feedback connections. These connections are believed to convey global information about behavioral context and state from cognitive areas down to the visual cortex. At the beginning, we will study changes in internal state induced by attention and task demands. Subsequently, we will explore a completely new phenomenon: How does the expectation of reward induce state-dependent changes in neuronal activity, and how do the changes influence neuronal and behavioral performance? In the cerebral cortex, reward effects have been exclusively examined in the higher areas. Therefore, understanding whether and how the expectation of reward affects visual processing would be a groundbreaking example of how even basic cortical processes underlying low to mid-level vision could be modulated by the internal state of the animal. The results will contribute to a revision in our understanding of how visual inputs are combined with non-visual (reward) signals to encode behaviorally-relevant visual features, and are likely to have an impact on central coding in other sensory systems.
Understanding what happens in the brain in the absence of sensory inputs constitutes a new direction of my lab's research. Whys is this direction important is pursue? One key function of the brain is to construct and dynamically modify representations of the environment, and then use these representations to control behavior. By taking advantage of neuronal context and state, individual neurons can adapt rapidly to changes in the environment to efficiently encode relevant information. We believe that this information processing principle is by no means restricted to the visual cortex. Its simplicity makes it a strong candidate for a general principle that may be implemented throughout the brain. We suggest the possibility that neurons in higher-level cortical areas could exhibit emotional and cognitive state dependency. For instance, listening to the same song every day could elicit different emotional reactions depending on the state of neuronal networks involved in song processing. This dependency could have a slower time course and may be more persistent than state-dependent effects in visual cortex, yet it may obey the same rules. Identifying general principles of information processing in the brain is the major goal of systems and cognitive neuroscience. We believe that our proposal makes several steps in that direction. |
