When the eyes of a person are damaged this causes complete blindness. Likewise, when Wernicke’s and Broca’s areas of neocortex are damaged this causes complete aphasia, losing the ability to comprehend language as well as the ability to produce speech (Penfield and Roberts 1966). In the absence of the eyes, one can deliver electricity to various regions of the visual system such as the lateral geniculate nucleus, the visual cortex, or the superior colliculus, for instance, to evoke fragments of visual perception (Schiller and Tehovnik 2015; Tehovnik and Slocum 2013), but complete vision has yet to be achieved using such a method during blindness. In the absence of Wernicke’s and Broca’s areas no one has yet tried to recover language by activating subcortical sites that participate in language functions such as the thalamus and cerebellum, for example (Penfield and Roberts 1966; Schmahmann 1997; Tehovnik, Patel, Tolias et al. 2021; but see Ojemann 1991).
The neocortex contains the complete declarative code for language (Corkin 2002; Kimura 1993; Penfield and Roberts 1966). This compelled Pereira, Fedorenko et al. (2018) to use fMRI to collect signals focused on the language areas of neocortex (i.e., 50,000 voxels including the entire neocortex) as sixteen subjects were tested on 180 mental concepts contained in various sentences. It was found that the fMRI signal could predict the correct stimulus sentence at a rank accuracy of 74% correctness, on average (p < 0.01, Fig. 4b of Pereira, Fedorenko et al. 2018). Note that there was some variability in the ‘selected’ 5,000 (out of 50,000) most effective voxels per subject over time. fMRI does not have single-neuron resolution spatially and temporally, but it is now believed that spanning minutes to years the composition of neocortical neurons mediating behavior can vary such that a percentage of cells always remains tuned to a task but that the composition of that percentage fluctuates (Chen and Wise 1995ab; Gallego et al. 2020; Rokni, Bizzi et al. 2007). However, delivering a speech requires great precision of word order. This precision is maintained by neocortical-cerebellar loops that are instrumental in converting the declarative code of neocortex into an explicit motor code via the cerebellum (Gao et al. 2018; Guo, Hantman et al. 2021; Hasanbegović 2024; Zhu, Hasanbegović et al. 2023; Mariën et al. 2017; Ojemann 1983, 1991; Thach et al. 1992), which stores the efference-copy representation for automatic performance (Bell et al. 1997; Chen 2019; Cullen 2015; De Zeeuw 2021; Fukutomi and Carlson 2020; Loyola et al. 2019; Miles and Lisberger 1981; Noda et al. 1991; Shadmehr 2020; Tehovnik, Patel, Tolias et al. 2021; Wang et al. 2023). Patients with damage to Boca’s area can perform movement sequences (of the upper extremities) that have been overlearned but are impaired at learning new sequences (Kimura 1993), which means under such conditions the remaining islands of neocortical and cerebellar connectivity are sufficient to generate previously learned movements. Thus, to learn new sequences (especially as it pertains to language) requires that Broca’s area be intact.