Working memory is the cornerstoneof cognition, the ability to retain information for short periods of time. Webuilt anhigh-throughput semi-automatized setup that trains mice toperform a task to explore this function. Analyzing data that arose from such systemusing Generalized Linear Models, we found that errors could ariseboth from problems in memory but also biases related to pervious sequences of responses. We found that the blockageNMDA receptorswith systemic injection of antagonists could modulate the proportion oferrors, with animals showing more predominance of non-memory errors under the effects of NMDA antagonists. We developed a mathematical model based on Hidden Markov Models that is able to account for both types of errors, both for standard behaviorand that observed during pharmacological manipulations. Finally, we performed extracellular recordings in the cortex while animals were doing the task and recovered neural traces that could explain errors due to deficits in working memory.
© 2001-2025 Fundación Dialnet · Todos los derechos reservados