The hippocampal formation is distinctly different from the sensory maps of the neocortex. First, rather than being topographically organized to encode the spatial characteristics of a single sense such as vision, audition, or somatosensation (the topographic senses), it is devoid of any topography for the hippocampus is a pathway that transmits sensory-motor information to the neocortex for long-term storage (Corkin 2002; Corrigan et al. 2023; Knecht 2004; Morrison and Hof 1997; Munoz-Lopez et al. 2010; Rolls 2013; Roux et al. 2021; Scoville and Milner 1957; Squire et al. 2001; Xu et al. 2016). Second, topographic maps are designed for the sharing of information such that conditioning by electrical stimulation of one part of a map generalizes immediately to another part of the map without additional training, but the same is not true of the hippocampus (Bartlett, Doty et al. 2005; Doty 1969; Knight 1964; Tehovnik and Slocum 2009). Each neuron within the hippocampal formation is independent such that the sharing of information between the activated fibres does not occur (Knight 1964). This independence is critical when conveying learned information to the neocortex so that all items present in the environment during learning—regardless of the nature of the sensation—can be transmitted independently (but as a unit) so that all objects are stored along with their context at the single cell level in the neocortex (Lu and Golomb 2023). Third, the neocortical location where information is stored varies according to whether it is allocentric or egocentric. Allocentric information (i.e., information not anchored to any body part) is stored in the temporal and orbital cortices, and egocentric information is stored in the parietal cortex and in the medial regions of the frontal cortex, such as in the supplementary motor area (Merker 2005), which encodes body posture (Chen and Tehovnik 2007; Fukushima et al. 2011; Schiller and Tehovnik 2015; Wiesendanger 2006). And this information is conveyed to the cerebellum such that the allocentric information is transmitted to the mediolateral lobe (composed of Crus I/II) and the egocentric information is transmitted to the posterior lobe (Tehovnik, Patel, Tolias et al. 2021).

The neuro-mechanics of how objects embedded within a sensory context are conveyed to the neocortex in rodents and primates has been studied in detail (Berger, Deadwyler et al. 2011; Deadwyler, Berger et al. 2016). Recordings were made of CA3 and CA1 hippocampal neurons to predict the firing pattern of the CA1 neurons as animals performed either a delayed-nonmatch-to-sample memory task for rodents (or a delayed-match-to-sample memory task for primates) using eight implanted electrodes in CA3 and eight implanted electrodes in CA1. Once the pattern of activity was documented, electric currents resembling the pattern of activity were injected through the electrodes in CA1 to see whether activating the hippocampus would potentiate an animal’s performance following a delay period. The stimulation was delivered for about 1 to 3 second during or immediately after the sample period but before the start of the delay; the current, pulse duration, and frequency were restricted, respectively, to 100 μA (or less), 0.5 ms, and 20 Hz (or less). The memory/delay periods tested ranged between 1 and 60 seconds for both rodents and primates. Electrical activation of the hippocampus potentiated the memory recall (see Fig. 1); silencing the hippocampus abolished the potentiation and diminished the memory recall (but see Footnote 1). The stimulation results prove that the information conveyed via hippocampal fibres about the sensory environment enhances memory.

Under normal circumstances it takes up to one month of training to consolidate the declarative aspects of a new task at the level of neocortex by way of the hippocampus, but of course the precise period of consolidation depends on the complexity of the task since simple classical conditioning tasks can be consolidated immediately (Corkin 2002; Corrigan et al. 2023; Knecht 2004; Morrison and Hof 1997; Munoz-Lopez et al. 2010; Pavlov 1927; Rolls 2013; Roux et al. 2021; Scoville and Milner 1957; Squire et al. 2001; Swain, Thompson et al. 2011; Takahara et al. 2003; Xu et al. 2016). Each region of the neocortex is believed to send projections to the cerebellum such that they occupy a broad region of the cerebellar cortex, and the return projections from the broad region of the cerebellar cortex converge onto a small region of the neocortex (see Fig. 2, from Hasanbegović 2024). This projection system converts declarative information into executable code for movement (Tehovnik, Hasanbegović, Chen 2024). When optogenetics is used to disable a delayed licking response in mice as mediated by the anterolateral motor cortex at the level of the cerebellum, silencing the regions of the input/output overlap in cerebellar cortex interrupts the performance of the task (Hasanbegović 2024), even though a large part of the cerebellar cortex is innervated by the anterolateral motor cortex. This over-innervation indicates the importance of the cerebellum in combining information from all aspects of the neocortex as well as the spinal cord and vestibular nuclei during task execution (Thach et al. 1992). Most importantly, brain regions not directly involved in a task are interconnected at the level of the cerebellum to mediate the shifts of an animal’s body during a task, which is subserved by neurons in the anterior and posterior lobes of the cerebellum. Incidentally, movement-proprioception-and-balance are often considered unconscious sensations (Eccles et al. 1967), which explains why the cerebellum has been associated with the execution of unconscious acts, unlike the neocortex (Tononi et al. 2008ab).

In the case of the neocortex, even for something as basic as a shock to the wrist for wrist/auditory tone conditioning, large portions of the neocortex were recruited in humans, as verified with PET (Hugdahl 1998). Posterior regions of neocortex activated included the extrastriate, somatosensory, and temporal cortices, and anterior regions activated included the orbital frontal, dorsolateral prefrontal, and inferior and superior frontal areas. This supports the observations of Pavlov (1929) that the neocortex is necessary for classical conditioning based on cortical ablation. For wrist-auditory associations, the critical sites of activation are the somatosensory and superior frontal cortices (for the shock and wrist response) and the temporal cortex (for the conditioned tone response); nevertheless, additional regions were activated. Indeed, using wide-field two-photon calcium imaging, much of the neocortex is engaged during ‘generic’ task execution (Musall et al. 2019), which agrees with the thinking of Lashley (1929) that the neocortex is a sensory integrator which distinguishes it from sensory receptors by using its stored information to modify ongoing behavior (Froudarakis et al. 2019).

In summary, for conscious events to be stored in the neocortex and then transmitted to the motor system at some future time, the following structures are critical: (1) the hippocampus for memory consolidation, (2) the neocortex for the long-term storage of declarative information, and (3) the cerebellum for the storage of the executable code that merges consciousness with the motor system.

Footnote 1: Since only one hippocampal hemisphere was silenced, it is possible that the silencing produced a position habit thereby contributing to the memory deficit.

Figure 1: Mean percent-correct performance is plotted as a function of delay in seconds. The blue curve represents performance in the absence of stimulation and the red curve represents performance with stimulation of CA1 at the sample period. The green curve represents performance for a scrambled signal delivered to the hippocampus. The top panel illustrates the delayed non-match-to-sample task performed by the rats. For other details see figure 3 of Deadwyler, Berger et al. (2011).

Figure 2: Top panel: anatomical transsynaptic (anterograde) tracer was injected into the anterolateral motor cortex of the mouse and traced to the cerebellar cortex. Notice the bilateral innervation of the mediolateral lobe of the cerebellum (i.e., mossy fibre density). Bottom panel: anatomical transsynaptic (retrograde) tracer was injected into the ventrolateral thalamus, which innervates the anterolateral motor cortex. Notice the strong retrograde innervation of the medial regions of the cerebellar cortex (i.e., Purkinje cell label). To interconnect the (lateral) input with the (medial) output, signals would need to be interlinked via the parallel fibres of the granular neurons, which are the most plentiful neurons of the brain (Herculano-Houzel 2009; Hueng 2008) and are central to the computational power of the cerebellum. The data are from Fig. 1.1 of Hasanbegović (2024).