The parallel fibres of the cerebellar cortex interconnect the mossy fibres (the carriers of sensory input) with the Purkinje neurons, which are critical for adaptation during the learning of new tasks (Bell et al. 1997; De Zeeuw 2021; Gallistel et al. 2022; Ito and Miyashita 1975; Loyola et al. 2019; Robinson 1976; Shadmehr 2020; Tehovnik et al. 2021; Wang et al. 2023). Sultan and Heck (2003) have indicated that the length of a parallel fibre within the cerebellar cortex is 7 mm in cats (Brand et al. 1976) and 6 mm in primates (Thach et al. 1991) with a conduction velocity of 0.3 to 0.5 m/s (or 0.3 to 0.5 mm/ms). It would therefore take 12 to 20 ms to travel 6 mm along a parallel fibre in primates, for example. Performing a behavior at a latency of 12 to 20 milliseconds or greater is too slow for the execution of some critical behaviors, especially the vestibulo-ocular reflex, which can occur at a latency of 5 to 6 ms after full adaptation (Hunter and Cullen 2002). The slow conduction time of the parallel fibres indicates that the cerebellum is not used for task execution, but only for the learning of a task, which fits with the results of cerebellar-cortex-only lesion experiments for a range of behaviors including the vestibulo-ocular reflex, eye-blink conditioning, and object-response associations (Ito et al. 1974a; Kassardjian et al. 2005; Sendhilnathan, Goldberg 2020b; Swain, Thompson et al. 2011; Takahara et al. 2003; Takemori and Cohen 1974), and it coincides with the thinking of Miles and Lisberger (1981). In these experiments, the learning of a task is abolished by destruction of the Purkinje neurons while sparing the cerebellar nuclei. Thus, once a task is learned, its performance is not interrupted by removal of the cerebellar cortex.
But interruption of the cerebellar cortex by optogenetic activation of the Purkinje neurons always interferes with the performance of highly over-trained mice required to remember the location of a tactile stimulus to obtain a reward, as reported by Gao and colleagues (see Fig. 1 of Gao, De Zeeuw et al. 2018; also see Fig. 4 of Zhu, Hasanbegović et al. 2023). One detail differentiates the cerebellar interruption experiments of the other studies (Ito et al. 1974a; Kassardjian et al. 2005; Sendhilnathan, Goldberg 2020b; Takemori and Cohen 1974) and those of Guo, De Zeeuw et al. (2018): the latter disabled the cerebellum by actively inhibiting the cerebellar nuclei, which are important for the mediation of all automated behaviors (Miles and Lisberger 1981; Takahara et al. 2003); hence, the engram or the executable code following over-training is stored in circuits connected to the cerebellar nuclei independent of the cerebellar cortex above.
Nevertheless, it has been shown that once animals are overtrained, the activity of the cerebellar cortex continues unabated as though to suggest that it still somehow participates in the execution process (Gao, De Zeeuw et al. 2018; Zhu, Hasanbegović et al. 2023) even though it is not necessary. We would suggest that the nervous system is very conservative when it comes to learning such that it is programmed to always check the status of adaptation as it relates to a given behavior. For example, when you brush your teeth in the morning from day to day this activity is typically performed automatically and unconsciously, but if you had just returned from a dental cleaning, you might be more apt to brush certain teeth with more vigor due to recommendations by your dentist. In short, the adaptation process is always subjected to updates to optimize behavioral performance.
Indeed, Lisberger (1984) has documented that the latencies of the vestibulo-ocular reflex oscillate across trials such that for fast trials the signal bypass the cerebellar cortex to execute a response using the vestibular nerve-vestibular nuclei-abducens pathway, and for slow trials the signal utilizes the vestibular nerve-mossy fibre-parallel fibre-Purkinje neuron-vestibular nuclei-abducens pathway. The same process is likely in play for oscillations between fast and slow eye-blink conditioning trials (Gallistel et al. 2022) and fast and slow saccadic eye movement and pursuit trials (Miles, Optican et al. 1986; Schiller and Tehovnik 2015). So, the next time you watch Usain Bolt charging down a 200-meter track to secure a gold medal be assured that on that trial the cerebellar cortex was not being used during task execution.