Fig. 1: Schematic representation of the task and RL results for all four association schemes, highly predictive anticorrelated (HA), highly predictive correlated (HC), variable predictive (VP) and non-predictive (NP).
From: Timing along the cardiac cycle modulates neural signals of reward-based learning
a Representation of the task and cardiac-related neural signals (HEP) recorded at outcome for 4 s - enough to record on average 3.5 HEPs. b Example of association between cues and predicted colour for one version of the task (anti-correlated blocks – see “Methods” section). c Prediction strengths for each block. d Model prediction of choices. e Learning rates from the best fitting model. f, g Reaction times and performance (mean reward across the four blocks) h left panel Definition of absolute and signed PE. h Right panel Each coloured pattern illustrates a different way in which activity in a neural structure may manifest if it is sensitive to different aspects of outcomes and their associated PE (locked at time of feedback). The blue pattern illustrates activity as a function of outcome valence – in a categorical manner as either positive or negative. The red pattern shows a monotonically increasing response profile consistent with a continually varying PE representation. The black pattern is continually varying as a function of the difference between the outcome and the expectation in an unsigned fashion. i Signed PEs are illustrated after averaging across all participants and all blocks, for the example of a task like that illustrated in panel (c). j Absolute PEs are similarly illustrated after averaging across participants and across blocks (k, l) and using the same format as for panels (i, j).
