Table 1.

Random variable names used in our APIC model.

t	Discrete time-step index
ot	Sensory observations (external and visceral)
$s_t^{(1:L)}$	Unobserved states constructed by the internal model
${\rho }_t^{(0:L-1)}$	Parameters to a reference distribution
a t	Control actions emitted by peripheral reflexes
x t	A complete model state for time t
p θ	Feedforward generative model with parameters θ
qϕ	Feedback generative model with parameters ϕ