Neural network mechanisms of working memory interference
- Autores: Joao Moura Barbosa
- Directores de la Tesis: Albert Compte Braquets (dir. tes.)
- Lectura: En la Universitat de Barcelona ( España ) en 2019
- Idioma: español
- Tribunal Calificador de la Tesis: Mark Stokes (presid.), Roland Alexander Charle (secret.), Gabriela Mochol (voc.)
- Programa de doctorado: Programa de Doctorado en Biomedicina por la Universidad de Barcelona
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
Combining humans and monkey neurophysiological experiments, we found that after memory-guided reports, past stimulus information ceased to be represented in PFC neurons. Surprisingly, this information was again represented by such neurons before furthcoming stimulus onset. Because this was reminiscent of ‘activity-silent’ mechanisms, we called this phenomenon reactivation. By introducing short-term plasticity dynamics in the bump-attractor model, we explained such reactivation dynamics and derived novel predictions that we then validated, linking behavior and neurophysiology. First, we found enhanced functional connectivity between PFC simultaneously recorded neurons that were involved in memory maintenanced, compared with trials that these same neurons were not involved in memory maintenanced. Second, in line with the simulations, we found that during trials undergoing higher information reactivation, monkeys and humans were prone to stronger serial biases. Together, these results point an activity-silent mechanism, such as short-term plasticity, as the underlying mechanism of interference from previous memories in working memory (i.e. serial biases).
Finally, we described for the first time serial biases in color working memory. In particular, we found that attractive serial biases build up in the course of an experimental session, an effect that cannot be explained solely by short-term plasticity. Rather, to account for such effect, we tentatively speculate that metaplasticity mechanisms need to be included in existing theoretical model for working memory interference.
