Ultrafast Cortical Gain Adaptation in the Human Brain by Trial-To-Trial Changes of Associative Strength in Fear Learning.

Abstract

In fear conditioning, more efficient sensory processing of a stimulus (the conditioned stimulus, CS) that has acquired motivational relevance by being paired with an aversive event (the unconditioned stimulus, US) has been associated with increased cortical gain in early sensory brain areas (Miskovic and Keil, 2012). Further, this sensory gain modulation related to short-term plasticity changes occurs independently of aware cognitive anticipation of the aversive US, pointing toward implicit learning mechanisms (Moratti and Keil, 2009). However, it is unknown how quickly the implicit learning of CS-US associations results in the adaptation of cortical gain. Here, using steady-state visually evoked fields derived from human Magnetoencephalography (MEG) recordings in two experiments (N = 33, 17 females and 16 males), we show that stimulus-driven neuromagnetic oscillatory activity increases and decreases quickly as a function of associative strength within three or four trials, as predicted by a computationally implemented Rescorla-Wagner model with the highest learning rate. These ultrafast cortical gain adaptations are restricted to early visual cortex using a delay fear conditioning procedure. Short interval (500 ms) trace conditioning resulted in the same ultrafast activity modulations by associative strength, but in a complex occipito-parieto-temporo-frontal network. Granger causal analysis revealed that reverberating top-down and bottom-up influences between anterior and posterior brain regions during trace conditioning characterized this network. Critically, in both delay and trace conditioning, ultrafast cortical gain modulations as a function of associative strength occurred independently of conscious US anticipation.SIGNIFICANCE STATEMENT In ever-changing environments, learned associations between a cue and an aversive consequence must change under new stimulus-consequence contingencies to be adaptive. What predicts potential dangers now might be meaningless in the next situation. Predictive cues are prioritized, as reflected by increased sensory cortex activity for these cues. However, this modulation also must adapt to altered stimulus-consequence contingencies. Here, we show that human visual cortex activity can be modulated quickly according to ultrafast contingency changes within a few learning trials. This finding extends to frontal brain regions when the cue and the aversive event are separated in time. Critically, this ultrafast updating process occurs orthogonally to aware aversive outcome anticipation and therefore relies on unconscious implicit learning mechanisms.