Unlearning Learned Fear: Testing The Prediction Error Hypothesis
Given that extinction training is widely regarded as a learning experience, extinction of fear may be viewed as the acquisition of “learned safety”. The learning of safety is governed, at least theoretically, by the disparity between what is expected to occur (shock) and what actually occurs (absence of shock) during extinction training. Faster and/or stronger learning occurs with a larger disparity or “prediction error” between expectation and actuality. The present experiment tested the prediction-error hypothesis as it applies to acquisition of learned safety. Male Long-Evans rats (240-280 g) underwent contextual fear conditioning followed 24 hr later by reactivation of fear memory; the next day the animals received a retention test. Contextual fear conditioning consisted of placing rats in a dark compartment for 120 sec followed by a single footshock (0.8 mA, 0.5 sec); reactivation of fear memory consisted of confinement of the animals to the dark compartment for either 30 or 120 sec in the absence of shock or no confinement (non reactivation control); the retention test consisted of returning the animals to the dark compartment for 180 sec in the absence of shock. Freezing behavior, defined by the cessation of movement other than that associated with respiration, was monitored and served as a measure of fear during the reactivation and the retention test. If prediction error determines the strength of learned safety, then as expectation of shock (or equivalently fear) increases in the absence of shock with increasing levels of confinement during reactivation, we predict 1) acquisition of learned safety should increase with increasing levels of confinement during reactivation and 2) retention of fear should decrease on the subsequent test. The prediction-error hypothesis was not confirmed. To the contrary, the results indicated that although fear during reactivation increased with increasing levels of confinement, retention of fear during the subsequent test decreased, relative to non reactivation controls, with shorter (30 sec) rather than longer (120 sec) confinement. We take this result as evidence that 30 sec reactivation of fear (or more accurately 30 sec reactivation of weak fear) indeed serves as a learning experience but in this case the learning is not governed by prediction error it is governed by absence of prediction error, that is, by absence of a disparity between what is expected to occur (weak fear) and what actually occurs (absence of shock). Thus, the animals acquire weak fear during reactivation and as a result the animals exhibit retention of weak fear during the subsequent test.
Society For Neuroscience 2013 Annual Meeting
November 9-13, 2013
San Diego, CA
Allen M. Schneider; P. E. Simson; E. H. Park; Patrick K. Monari , '13; Daniel J. Kalamarides , '14; Jennifer L. Walsh , '15; Corinne P. Sommi , '14; Rebecca Teng , '14; Greer E. Prettyman , '15; Nicholas L. Gettino , '13; and L. G. Kirby.
"Unlearning Learned Fear: Testing The Prediction Error Hypothesis".