Ideomotor theory, prospection and hippocampal goal-directed control

Older works treat the hippocampus as anatomy and as a lesion target. Pavlov (1927) reports experiments by Zavadsky on dogs after complete bilateral removal of the gyri pyriformi and the adjacent part of the hippocampus. The aim was to test a claimed link between the gyrus pyriformis and olfaction. Olfactory reflexes returned first, and by the third or fourth day the dogs could pick out bags containing meat or sausage.¹ Rádl (1930) recounts a separate dispute. Owen treated the hippocampus minor as characteristic of Man. Huxley replied that it is among the cerebral characters Man shares most clearly with the apes.²

Pezzulo, Rigoli and Friston (2015) argue that action selection and initiation depend on multimodal prediction errors, not only on stimuli. They say this agrees with a large body of ideomotor evidence, in which even button presses are planned and selected by their anticipated effects.³ They propose adding interoceptive prediction errors to this framework. That could link bodily states to action selection, a link they call under-explored in the animal learning literature. They add two caveats: interoceptive events are largely anticipatory, and the tie between an action and its interoceptive result can be distal.³

Pezzulo and colleagues describe the hippocampus as a structure with a primary role in goal-directed control. Its role has been studied mainly in spatial navigation. They add that evidence points to other abilities that need prospection, such as imagining fictive scenarios.⁴ In associative learning terms they link it to a process called episodic control. On this account, drawn from Lengyel and Dayan (2008), the hippocampus stores state-action sequences that reached a goal and replays them when one of their states recurs.⁴ They read the generation of fictive state sequences as reactivation within generative models.⁴

Fields and Levin (2022) contrast brain time travel with what they call preneural bioelectric time travel. In the second, pattern memories serve as future patterns toward which remodeling and morphogenesis can strive.⁵ They describe the mammalian hippocampus as functioning in part as a spacetime reference frame at the scale of the whole organism.⁶ Other authors use the hippocampus as a benchmark for memory. Gershman and colleagues cite early work linking learning to histone acetylation in the rat hippocampus.⁷ Baluška and Levin (2016) note that glutamate is a key transmitter there, while discussing memory in bone.⁸