| LeDoux Lab 2012 SfN Abstracts |
|---|
| Program#/Poster#: |
|
||
| Title: | Causality and its neural underpinnings in active and passive aversive learning | ||
| Location: | Halls B-H | ||
| Presentation Time: | Tuesday, Oct 16, 2012, 9:00 AM -10:00 AM | ||
| Authors: | *T. MADARASZ1, O. AKHAND1, J. P. JOHANSEN2, J. E. LEDOUX3; 1Ctr. For Neural Science, New York Univ., New York, NY; 2RIKEN Brain Sci. Inst., Wako, Japan; 3Ctr. for Neural Sci., New York Univ., New York, NY |
||
| Abstract: | Learning causal structure is a pivotal function of human cognition, as well as an important challenge in machine learning and the analysis of big data. Yet little is known with certainty about the computations and the neural mechanisms that enable the brain to learn about causality, or about the extent to which this ability generalizes to nun-human animals. Previous studies have suggested that both rats and humans could employ a normative statistical model and learn the structure of so-called Bayesian Networks to infer dependencies and causal relationships between variables in their environment, in ways that other associative learning models fail to correctly predict. Here we explore in the framework of Pavlovian and instrumental aversive conditioning, a preeminent battleground of different learning theories, how rats learn about the relationships between auditory cues, footshocks and their own actions, while we also probe the underlying neural circuitry. In particular, we build on our previous findings regarding how rats learn changing tone-shock contingencies that arise by the delivery of unsignalled shocks during and auditory fear conditioning paradigm. We previously demonstrated that depolarization of amygdala pyramidal neurons, a necessary teaching signal for auditory fear conditioning, is also required to keep track of such changes in contingencies. Here we show that the dorsal hippocampus on the other hand is not required for this learning, and that animals are sensitive to degraded tone-shock contingencies even if they cannot from contextual fear memories. We further show that modeling the learning process as structure learning in a Bayesian Network provides a better fit than competing learning models. Finally, we develop an instrumental paradigm to test if rats are using causal reasoning when deciding whether to perform interventions during active avoidance learning, and probe the necessary brain regions for such causal models to arise. Taken together, these findings enable us to make forays into understanding the neural mechanisms that give rise to the ability to reason with causal models. |
||
| Support: | Sevier Laboratories | ||