We learn from our mistakes. The goal of this project is to understand how the errors that drive learning are encoded and processed in the brain. All experiments are done using eyeblink conditioning, a simple form of associative learning that offers a number of advantages: 1) Errors are under experimental control and can be easily manipulated. In eyeblink conditioning, subjects learn to blink and protect the eye from a corneal air-puff that is applied every time after they hear a tone. Error, defined as the discrepancy between the real air-puff and what the subject expected, can be manipulated simply by unexpectedly changing the strength of the air-puff on any given trial. 2) Eyeblink conditioning can be done in mice. This opens up the door for approaches that take advantage of genetic tools to investigate the neural basis of error processing in awake-behaving animals. 3) The neural circuits mediating eyeblink conditioning have been identified previously. Knowledge about the connectivity of these circuits has led to the hypothesis that neurons in the inferior olive generate error signals by comparing information about the real air-puff, presumably conveyed to the inferior olive via an excitatory pathway from the trigeminal nucleus, and the expected air-puff, presumably conveyed via an inhibitory cerebellar pathway. This project uses the 3 advantages listed above to investigate the role of errors in eyeblink conditioning, and to elucidate the coding and processing of error signals by neurons in the inferior olive.
Specific aim 1 challenges the widely-held view that to update expectations (about the strength of future air- puffs, for example), the brain takes into account only the error in the current trial. By unexpectedly changing the strength of the air-puff trial-by-trial, this aim will define the role of history and statistics of prior errors in updating expectations about future events.
Specific aim 2 is about deciphering the neural code used by neurons in the inferior olive to represent errors of different sizes. By doing eyeblink conditioning in mice, while recording single-units and imaging populations of neurons in the inferior olive, this aim will reveal whether synchrony provides information about the size of the error in single trials.
Specific aim 3 takes advantage of genetic tools to ask if synchrony in the inferior olive is necessary and/or sufficient to represent the error signal and drive eyeblink conditioning. Whether synchrony is necessary will be addressed by assessing eyeblink conditioning in connexin36 mice, whose activity in the inferior olive is normal except for reduced levels of synchrony due to loss of electrotonic coupling. Whether synchrony is sufficient will be addressed using optogenetic stimulation in mice expressing channelrhodopsin in the inferior olive, and assessing the efficacy with which different spatio-temporal patterns of neural activation can signal error to drive eyeblink conditioning.
Making errors is typically considered a nuisance, but in fact, error information is essential for driving adaptation, improving performance, and updating expectations about future events in an uncertain and ever-changing world. The goal of this proposal is to understand the neural coding and processing of error signals during a simple form of associative learning. This research will uncover fundamental principles of learning that are particularly relevant for neurological disorders, like ADHD, autism, and schizophrenia, whose symptoms have been linked to failures in monitoring action outcomes and recognizing errors as such.
Showing the most recent 10 out of 16 publications
