Semester of Graduation

Spring 2024


Master of Arts (MA)



Document Type



The hippocampus is understood to play a key role in the formation of episodic memories. Prominent theories suggest this is made possible through use of sparse encoding schemes, in which a small number of neurons encode each episodic event using a distinct pattern of neuronal activity. However, more recent computational models suggest that the hippocampus may also be capable of representing semantic structure, a notion previously thought to be exclusive to the neocortex. In our previous line of work, we trained multivariate classifiers to predict the semantic content, specifically affect, of computer-generated faces given spiking data taken from individual neurons located in the hippocampus and other surrounding limbic structures. Our models were unable to successfully decode affect even when provided with neurons that had previously shown increased activity from baseline for valenced faces. The results of our analysis suggested that the spiking activity of individual neurons within the hippocampus may not convey reliably decodable representations of affect. A more distributed viewpoint would argue that encoding the information necessary to perform higher-order cognitive functions, such as affective evaluation, would require differential levels of activity across much broader neuronal populations. To test this viewpoint, we trained multivariate classifiers to predict affect from a time series analysis of the local field potentials recorded from the previously analyzed dataset. Across varying levels of model complexity, we were unable to decode affect from the spectral information contained within the recorded local field potentials. In tandem with our previous results, we provide support for existing theoretical models that suggest that the hippocampus is primarily responsible for encoding episodic events and that complex semantic structure is represented elsewhere in the brain.



Committee Chair

Cox, Christopher