Philipp Haueis (Berlin School of Mind and Brain)
Neural structures with multiple functions make it unclear when we have successfully described what a structure does when it works. Several recent accounts attempt to tackle this problem of multifunctionality differently. Rathkopf (2013) proposes an intrinsic function concept to describe what a structure does whenever it works, whereas Burnston (2016a) argues for context-sensitive descriptions to tackle multifunctionality. McCaffrey (2015) proposes a middle road by indexing invariant or context-sensitive descriptions to the mechanistic organization of a multifunctional structure.
In this paper, I argue that these accounts underestimate the problem of multifunctionality. Because they implicitly assume that “multifunctional” means “contributing to multiple cognitive functions”, they overlook other types within the purview of their accounts: circuit switching in central pattern generators and gain control in cortical microcircuits. Central pattern generators are multifunctional because they can switch between rhythmic motor outputs (Briggmann and Kristan 2008). Cortical microcircuits are multifunctional because some circuit elements process sensory information, whereas others prevent damage by controlling circuit gain (Merker 2013).
These circuit functions are not operative in cognitive processing but instead enable such processing to occur at all. Yet they exhibit exactly the features that philosophical accounts recruit to handle (cognitive) multifunctionality. Similar to Rathkopf’s intrinsic function account, circuit switching and gain control can be analysed without reference to the behavior of the organism. Yet, they do not replace but complement task-based functional analyses of multifunctional structures, thus questioning the plausibility of the intrinsic function account. Circuit switching and gain control also show that Burnston’s and McCaffrey’s accounts are incomplete. Because he focuses on cognitive contexts, Burnston’s contextualism fails to capture how circuit switching and gain control change with biochemical and physiological contexts, respectively. These contexts make the problem of multifunctionality harder than Burnston acknowledges, because different context types cross-classify the response of multifunctional structures. Similarly, McCaffrey’s typology of mechanistic organization to classify multifunctional structures fails to capture how circuit switching or gain control are mechanistically organized. Because central pattern generators can switch rhythmic outputs independently of sensory inputs, they are mechanistically decoupled from cognitive functions that process those inputs. In contrast, gain control is essentially coupled to cognitive functions because it is only necessary to prevent damage when a cortical microcircuit processes sensory information.
My analysis shows that existing philosophical accounts have underestimated the problem of multifunctionality because they overlooked circuit functions that are not operative in, but instead enable cognitive functions. An adequate account of multifunctionality should capture all types of multifunctionality, regardless of whether they are cognitive or not.
