Matthieu M. de Wit
Heath E. Matheson- 1Department of Neuroscience, Muhlenberg College, Allentown, PA, United States
- 2Department of Psychology, University of Northern British Columbia, Prince George, BC, Canada
Mainstream cognitive neuroscience aims to build mechanistic explanations of behavior by mapping abilities described at the organismal level via the subpersonal level of computation onto specific brain networks. We provide an integrative review of these commitments and their mismatch with empirical research findings. Context-dependent neural tuning, neural reuse, degeneracy, plasticity, functional recovery, and the neural correlates of enculturated skills each show that there is a lack of stable mappings between organismal, computational, and neural levels of analysis. We furthermore highlight recent research suggesting that task context at the organismal level determines the dynamic parcellation of functional components at the neural level. Such instability prevents the establishment of specific computational descriptions of neural function, which remains a central goal of many brain mappers – including those who are sympathetic to the notion of many-to-many mappings between organismal and neural levels. This between-level instability presents a deep epistemological challenge and requires a reorientation of methodological and theoretical commitments within cognitive neuroscience. We demonstrate the need for change to brain mapping efforts in the face of instability if cognitive neuroscience is to maintain its central goal of constructing computational mechanistic explanations of behavior; we show that such explanations must be contextual at all levels.
Introduction
A brief history of the origins of modern functional brain mapping
The historical neuropsychological literature is full of case-studies of individuals with specific behavioral impairments following damage to local regions of the cerebral cortex. Perhaps most famously, Broca (1861) described a patient who, after damage to the left inferior frontal gyrus (IFG), permanently lost almost all of his expressive language abilities while his receptive abilities remained largely intact. Other patients may exhibit fluent (but incoherent) expressive language paired with receptive impairments, typically following lesions in the left superior temporal gyrus (STG). From these observations, Wernicke (1874/1969) developed one of the earliest neurocognitive models of language, which centered on the specific functions of IFG and STG and their connection via association fibers. Wernicke argued that complex functions such as language are the result of the interaction of multiple, simpler, sensory, motor, and association processes, thereby effectively ending the search for localized higher-level “faculties” in the brain that had dominated the study of brain-behavior relationships up to that time (Fancher, 1996; Bergeron, 2007). Since then, researchers have continued to grapple with understanding the functions of brain parts and their relationship to behavior, and one of the main goals of cognitive neuroscience remains mapping functional descriptions onto neural activity. But what does successful mapping look like?
Functional brain mapping as the analysis, decomposition, and localization of function
The goal of mainstream cognitive neuroscience1 is to “investigate brain-behavior interactions” and to “address both descriptions of function and underlying brain events” (Journal of Cognitive Neuroscience, 2021); it seeks “an understanding how the functions of the physical brain can yield the thoughts, ideas, and beliefs” of the mind (Gazzaniga et al., 2019, p. 4). In the mainstream cognitive neuroscience research literature, the term function is critical to the enterprise. Importantly, function is often implicitly and interchangeably used at multiple, typically three, distinct levels of analysis (Marr, 1982/2010; Craver, 2014; Krakauer et al., 2017; Zednik, 2018; see Garson, 2016, for discussion of conceptualizations of biological function). The first level concerns that of observable behavior or its disorders. Here, cognitive neuroscientists distinguish between various complex behavioral or cognitive categories described at the personal or organismal level; that is, at the level of the behaving human being. For example, we talk about the abilities of (or impairments of) “paying attention,” “using a tool,” or “speaking.” These are the phenomena of which the discipline ultimately seeks a mechanistic explanation, in terms of a description of the parts and interactions of a system that gives rise to the phenomenon (Miłkowski, 2013; Craver and Tabery, 2015; Zednik, 2019).2
To this end, at the second, subpersonal, level the phenomenon described at the first level is decomposed into components that perform specific computational operations over representations and are organized together in a specific way (Bechtel and Abrahamsen, 2010). These components are therefore identified by their functions (Piccinini and Craver, 2011), the descriptions of which tend to be “human-interpretable” (Hasson et al., 2020). For instance, computational processes are hypothesized that implement, e.g., “attentional orienting,” “action selection,” or “linguistic retrieval,” which are subcapacities of the organismal-level capacity of paying attention, using a tool, and speaking, respectively. In other words, and following Wernicke’s lead, at the second level multiple latent, interacting, domain-specific or domain-general, functional components are postulated that explain the production of a particular complex behavior or cognitive ability observed at the first level. In general terms, it is widely accepted that computational operations are thought to involve inputs which feed into the manipulation of internal, typically – but not necessarily (Piccinini and Scarantino, 2011) – representational states to produce outputs that are then used downstream in further computational processes (Shea, 2018).
The third level at which function is invoked is that of the brain. Thus, following the characterization of a componential computational architecture at the second level (and sometimes before this characterization has taken place; see Agis and Hillis, 2017), an attempt is made to map the components onto the activity of the brain, effectively reifying that architecture; that is, the researcher tries to describe the coordinated computational operations in terms of physical neural processes (Piccinini and Shagrir, 2014; Burnston, 2021). There are at least two possible characterizations of this final step. Sometimes, the assumption is that a computational operation will directly map onto a specific brain “part,” where a part is broadly construed as a cell, localized assembly of cells or a distributed network of cells (Glasser et al., 2016). For instance, using lesion-symptom mapping or functional magnetic resonance imaging (fMRI), cognitive neuroscientists may search for a neural region or network of regions whose function it is to implement the computations that instantiate attentional orienting (while sometimes considering the biological constraints of the part such as, e.g., neural response latencies or receptor types; e.g., Kravitz et al., 2013), and another region or network whose function it is to implement the computations for action selection. However, sometimes the assumption is that a computational operation requires additional decomposition before it can be mapped onto the brain. For instance, using computational modeling of different types of cells or cell networks, attentional orienting is broken down further into component operations (e.g., a number of different computations in a connectionist model). Either way, the goal is to mechanistically explain organismal-level phenomena, via organized, interacting computational components, in terms of functionally specialized brain cells, assemblies, or networks that combine into interacting brain parts.3 Though cognitive neuroscience has moved on from structure-function mapping understood in the modular sense described earlier (e.g., expressive language is due to Broca’s area), many cognitive neuroscientists – implicitly or explicitly – still seek to discover the brain basis of our ability to pay attention, use a tool, or speak (Cabeza and Nyberg, 2000; see Zerilli, 2019 for a historical overview of this approach).
Given this, brain mapping continues to be a major effort, which can also be gleaned from its influence on the infrastructure of the field (e.g., the UCLA Brain Mapping Center; the journal Human Brain Mapping). Indeed, the idea of functional brain mapping is so deeply engrained in current thinking about brain-behavior relationships that it is implicit in our nomenclature and pervades textbooks and day-to-day public discourse about neuroscience. Thus, there is talk of the primary visual cortex or the dorsal attention network, which are believed to implement specialized computations underlying visual and attentional behavior, respectively. In a sense these efforts are clearly successful. For instance, it is possible to manipulate the activity of cortical regions using transcranial magnetic stimulation (TMS) and observe predictable effects on behavior, lesions to early visual regions result in predictable visual field deficits, and sophisticated computational models exist that impressively replicate various properties of real-world brain networks.
However, at the same time, the practice of functional brain mapping as defined above has met serious empirical and theoretical challenges, particularly in recent years (Uttal, 2001; Pessoa, 2008; Anderson’s, 2010, Anderson, 2014; Klein, 2012; Burnston, 2016b; Khalidi, 2017; Stanley et al., 2019; Viola, 2020; Zerilli, 2020). We are certainly not the first to note challenges in this domain: Previous neuroscientists critical of the functional brain mapping enterprise have called for a more effective integration of research findings at each level of analysis (Krakauer et al., 2017), have argued for a redescription of function at the level of local brain regions (Price and Friston, 2005; Poldrack, 2010), or have pointed out general limitations of current mapping efforts (Poldrack, 2006; Genon et al., 2018), among other criticisms. However, with some important exceptions to be discussed below, several of these researchers still aim to identify the componential computational operations of well-defined brain parts and in that sense maintain the central premises of functional brain mapping. For example, Shine et al. (2016, p. 26) state that “[i]n considering the computational capacities of independent brain regions, we will make the argument that computational specialization is not only abundant in the brain, but also that it would be difficult to imagine a working brain that did not contain such specialization.” However, the viability of this approach is in question.
Outline of the review and further analysis
In contrast to previous critiques, we will suggest that the core problem cognitive neuroscience faces is an epistemological one: The present paper will integrate empirical and philosophical literature to show that the goal of giving any specific, computational description of context-independently defined brain parts is not possible, and therefore that the explanatory strategy of mainstream cognitive neuroscience is in need of revision. The section “Challenges to the Practice of Structure-Function Brain Mapping: An Integrative Review” provides an integrative review of the empirical literature on flexible neural tuning (section “Neural Tuning and Functional Brain Mapping”), plasticity and recovery of organismal-level function following brain lesions (section “Lesion Studies, Plasticity, and Functional Brain Mapping”), inter and intra-individual neural degeneracy (section “Degeneracy and Functional Brain Mapping”), neural reuse (section “Neural Reuse and Functional Brain Mapping”), and the neuroscience of enculturated skills such as reading and mathematics (section “Enculturated Skills and Functional Brain Mapping”). These areas are increasingly being investigated by researchers within cognitive neuroscience, and we summarize one consequence of these challenges that has been recognized in the field as “weak contextualism,” which denotes variable mappings relative to the organismal level of description (section “Consequences of Weak Contextualism”). However, section “Strong Contextualism, Instability of Mapping, and Indeterminate Part Ontology for Cognitive Neuroscience” describes the fundamental incommensurability of the goals of mechanistic explanation and any context-independent – even weak contextual – descriptions of brain function, and highlights a consequence not yet widely recognized within the field, a type of instability in functional mapping that reflects a “strong contextualism.” Most significantly, we review recent research that forces a reconsideration of what constitutes a relevant neural “part” to begin with and show that the parcellation of functional components shifts depending on the task context we choose to study at the organismal level. Finally, the section “Discussion: Implications for Cognitive Neuroscience” provides a brief methodological and theoretical sketch of a cognitive neuroscience that can maintain its central goal of constructing robust computational mechanistic explanations of behavior by being sensitive to the fact that such explanations must be contextual at all levels.
Challenges to the practice of structure-function brain mapping: An integrative review
Below we provide an integrated review of research that present challenges to structure-function mapping (see Ames and Fiske, 2010; Anderson, 2014; Seifert et al., 2016; Hartwigsen, 2018; Rule et al., 2019 for previous focused reviews of each of the topics discussed in this section). Afterward, we will highlight the consequences this literature has had on the current state of the art in cognitive neuroscience.
Neural tuning and functional brain mapping
Neuroscience has a long history of characterizing the neural tuning of brain regions, where manipulations of stimulus parameters of popular interest (e.g., line orientation, face-ness of an object, etc.) are correlated with changes in neural activity (firing rate, BOLD signal, etc.; see Buzsáki, 2020, for a brief description of this history). However, neuroimaging results show that many neural regions are not statically tuned to particular types of stimuli in a stable manner (Bair, 2005; Clopath et al., 2017). While flexibility in neural tuning at the single neuron level has been shown for some time (e.g., Miller, 2000), recent results show that the neural tuning of most brain regions appears capable of changing rapidly between different cognitive tasks. For instance, Çukur et al. (2013) had participants watch videos while lying in the scanner. They were instructed to attend to different features of the videos, specifically vehicles or humans. As they did so, it was shown that the tuning characteristics of almost every region in cortex shifted depending on the goal of the observer (i.e., voxel activity was better explained by responding to humans when searching for humans and vice versa for vehicles). Other research has demonstrated more specific effects. For instance, ventral cortical regions, typically implicated in identifying objects, do not do so in an all-or-none fashion but shift their response tuning to object identity depending on the exact task participants are performing (Harel et al., 2014). In addition, recent advancements in multi-voxel pattern analysis (MVPA; Kriegeskorte et al., 2008) suggest that the “representational geometry” (i.e., the abstract, multidimensional space of neural activity patterns; see Kriegeskorte and Diedrichsen, 2019) of different network nodes is dynamically defined by context. For instance, geometry of areas in the dorsal stream is better described by action similarity in a task where participants make judgments about action, but category similarity when participants make judgments about category (see also Anderson and Oates, 2010; Gallivan and Culham, 2015; Bracci et al., 2017). Similarly, patterns of activity in some areas of the cortex are implicitly tuned to the category of objects depending on cues that are available in the environment (Matheson et al., 2021). The important thing to note here is that the tuning of single cells determines the representational geometry that is measured in these studies (Kriegeskorte and Wei, 2021). Thus, changes in neural representational geometry in different tasks reflect changes in neural tuning across most of the cortex depending on the context.
Overall, these neural tuning findings challenge functional mapping efforts because they suggest that the target – the response properties of a neural region to particular stimulus information – is a moving one that is shaped by task context.
Lesion studies, plasticity, and functional brain mapping
Second, findings of neural plasticity and recovery have long complicated the functional brain mapping literature. After (sometimes extensive, e.g., García et al., 2017; Bowren et al., 2021) brain damage, patients may be able to partially or fully recover the behavioral or cognitive ability that was lost (Kolb and Gibb, 2013). This trajectory of recovery can continue for years (Hartwigsen and Saur, 2019) and has even been observed following lesions in early sensory areas (in adult cats; see Jiang et al., 2021). One account of this phenomenon, consistent with the assumption that local neural regions perform specialized computational operations, is that the regained organismal-level function is supported by different computational processes. Perhaps a compensatory mechanism, involving different cognitive strategies, restores function at the organismal level (Dixon et al., 2008; see De Brigard, 2017 for a related discussion focusing on brain and cognitive strategy changes associated with healthy aging).
Another more commonly offered account of recovery that is consistent with functional brain mapping is that, as a result of plasticity, the remaining undamaged brain regions are able to reorganize themselves to accommodate new functions; that is, neural parts are repurposed to perform new computational operations. However, if recovery is indeed a matter of repurposing (i.e., the specific computational operation formerly realized by the damaged area is now performed by another area), one would expect to lose the function that the newly colonized area was responsible for. Alternatively, it is commonly postulated that the remaining intact tissue is utilized more efficiently after recovery, allowing for the former and the new computational operation to be implemented alongside each other, but this notion of redundancy (Friston and Price, 2003) begs the question why the colonized tissue was utilized less efficiently before, given that neural tissue is notoriously expensive to maintain. To our knowledge, observations of loss of one function accompanying recovery of another function appear to be largely absent from the patient literature.
On the contrary, several recent treatment studies have reported gains in the language domain as a result of upper extremity movement therapy in stroke patients (Harnish et al., 2014; Primaßin et al., 2015; see Anderlini et al., 2019 for review; and Stoll et al., 2021 for related work in limb apraxia). Furthermore, brain damage is often associated with multiple co-occurring deficits (e.g., patients presenting with both limb apraxia and aphasia following left hemisphere lesions), similarly suggesting that the impacted area supports diverse behavioral domains (Behrmann and Plaut, 2014; Goldenberg and Randerath, 2015). So called “crossed” cases of classical clusters of deficits have moreover been reported in patients with atypical lateralization of function (e.g., limb apraxia presenting together with aphasia following a right hemisphere lesion; Raymer, 1999; see Vingerhoets et al., 2013 for findings of co-lateralization in neurologically intact individuals). Each of these findings is hard to explain under the assumption that impacted functions are implemented by specialized neural parts, and may be interpreted as further evidence that the functional significance of any given region is sort of a chameleon which does not permit a context-free definition (Price et al., 2016; Price, 2018).4
Degeneracy and functional brain mapping
It is increasingly recognized that the phenomenon of neural degeneracy – the notion that structurally different neural processes can produce equivalent behaviors at the organismal level – plays an important role in the brain (Edelman and Gally, 2001; Price and Friston, 2002; Noppeney et al., 2004, 2006; Figdor, 2010; Bateson and Gluckman, 2011; Sporns, 2011; Marder et al., 2015; Anderson, 2016; Seifert et al., 2016; De Brigard, 2017; Viola, 2020). Both inter- and intra-individual cases of degeneracy have been observed (Anderson, 2016). In the domain of numerical cognition, Tang et al. (2006) reported markedly different patterns of activation during simple arithmetic between native Chinese and native English speaking individuals, with left perisylvian activation in the former cultural group, and a network of “visual” and “premotor” regions in the latter, despite equivalent stimuli (Hindu-Arabic numerals) and equivalent performance at the behavioral level. In the language domain, Biduła et al. (2017) reported many different variants and degrees of language lateralization in neurologically intact individuals with normal language abilities, involving, in addition to more or less typical and atypical lateralization of classic language areas, right hemisphere components of the “default mode network” as well as an atypical role for the cerebellum. Finally, Merabet et al. (2008) provide evidence for intra-individual degeneracy in neurologically intact individuals by showing the existence of multiple, different neural substrates for braille reading (see De Brigard, 2017 for additional examples related to healthy aging). Degeneracy is clearly widespread, both in the intact and the lesioned brain (Fotopoulou, 2014; Price, 2018).5 Indeed, we would argue that functional recovery following brain damage can be considered a prime example of both inter and intra-individual degeneracy (see also Mogensen, 2011; Abrevaya et al., 2017; Hartwigsen and Saur, 2019).
Neural reuse and functional brain mapping
Fourth, there is extensive evidence from neurologically intact individuals showing that some, if not most, brain regions are implicated in many different behaviors, suggesting that they can be reused in different contexts. That is, regions are typically capable of participating in a diverse array of functions. The above-mentioned study by Merabet et al. (2008) showed, using TMS, that the occipital (i.e., “visual”) cortex of blindfolded sighted participants became causally involved in tactile perception following an intense 5-day braille reading training program, providing evidence for its functional perceptual capacity beyond the visual modality (see also Murray et al., 2015). Similarly, in congenitally blind individuals who have never possessed sight, the occipital cortex has been shown to be sensitive to non-perceptual stimulus attributes such as the grammatical structure of spoken sentences and the difficulty of math equations (Bedny, 2017). Much evidence has been marshaled in the last decade or so showing that even gross functional distinctions at the organismal level, for instance the difference between emotional processes and cognitive ones (Pessoa, 2008) or between perception and action (Cisek, 2007), do not hold at the neural level, due (in part) to reuse.
The consistency of reports of functional heterogeneity suggests that reuse is not a curiosity but a general feature of the nervous system (Anderson, 2010, 2014). Importantly, despite empirical advancements showing that some regions can be functionally further subdivided (e.g., Broca’s area; Fedorenko and Blank, 2020) reuse is observed regardless of the level of granularity at which these analyses are performed (i.e., whether the brain was parsed into, say, 10 or a 1,000 regions); thus, reuse may be reduced (Poldrack, 2006) but does not go away at an increased spatial resolution (Anderson et al., 2013; Uddin et al., 2014). For instance, it is clear that, at the resolution of the entire brain, the entire brain is reused to support different behaviors, but neural reuse can be observed even at the resolution of single neurons, with neurons involved in either sensory or motoric functions depending on the behavioral and concomitant neural context in the roundworm C. elegans (Bargmann, 2012). The fact that reuse phenomena do not disappear at smaller resolutions presents a major challenge to determining structure-function mappings.
Enculturated skills and functional brain mapping
Fifth and finally, Dehaene and Cohen (2007); (see Menary, 2015; Jones, 2018 for discussion) suggest that cognitive tasks that require enculturation and formal schooling in order to be displayed (like reading, writing, and mathematics) are supported by the neuronal “recycling” of neural regions that originally evolved for other purposes, yet have the right structure to implement those tasks – again a type of reuse, though reuse in this case is defined at longer timescales (see Borra and Luppino, 2018 for additional examples). The visual word form area in occipitotemporal cortex is a good example of such a region. The recycling account is convincing because these abilities have arguably emerged too recently (i.e., within the last few thousand years) for evolution to have generated specialized cortical regions to support them.
Consequences of weak contextualism
This brief review integrates some of the most significant challenges facing cognitive neuroscience. Within the field, it is increasingly recognized that all of these phenomena suggest that the functional role of a brain region is context-dependent. However, while context-dependence is recognized by the field, we argue that this is recognition of a type of “weak” contextualism. By weak contextualism we mean that most researchers accept that behavioral context shapes a region/network’s functional role in organizing the organismal-level behavioral phenomenon, but still maintain that the functioning of the part itself is context-independent. That is, it is thought that functions of brain parts are not stable when defined relative to the organismal level, but functions defined at the computational and neural level are (compare to Burnston’s, 2016a,2019 “absolutism”). Researchers sympathize with weak contextualism when they make continued calls for context-independent computational descriptions of brain regions while recognizing that this computation implements a cognitive subcapacity that contributes to many different organismal-level behaviors. For example, Vingerhoets et al. (2013) report evidence in neurologically intact individuals that skilled action and expressive language involve strongly overlapping neural components. To account for this overlap, they postulate that these components implement the production of complex (i.e., precise, articulated, coordinated, nested) learned movement, an operation that is common to both tool use and speech, explaining its involvement in each of those contexts (see Knops et al., 2009; de Wit et al., 2012; Parkinson et al., 2014 for similar conceptualizations of shared informational or computational demands across different behavioral domains). Thus, with this approach, the functional description of neural parts is context-independent and is abstract enough to account for their contextual involvement in a wide range of tasks (Price and Friston, 2005; Shine et al., 2016; Humphreys et al., 2021; see also Anderson’s, 2010 early “working” vs. “use” conceptualization of neural reuse). Note that this type of contextualism, though now often accepted in the field, is already quite far removed from the traditional structure-function accounts described in the section “Introduction,” which have typically characterized the function of brain regions relative to the organismal level (e.g., the fusiform face area is important for face identification) or at a minimum in terms of cognitive subcapacities directly related to specific organismal-level abilities (e.g., a region involved in “attentional selection”). Regardless, an acceptance of weak contextualism would still allow us to find the computational function of a brain part (as was suggested in the case of Vingerhoets et al., 2013).
Strong contextualism, instability of mapping, and indeterminate part ontology for cognitive neuroscience
While there may be sympathies toward weak contextualism within cognitive neuroscience, we argue that there is, inescapably, a form of “strong” contextualism that is not widely acknowledged (and therefore one that is far from accepted). Our argument builds on arguments within the philosophy of science regarding the consequences of seeking mechanistic explanation. Again, by mechanistic explanation, we mean the goal of providing a description of the parts and interactions of a system that give rise to a phenomenon (e.g., Craver, 2014), and in cognitive neuroscience the typical approach is to seek a mapping of behavior to computation to brain. Indeed, some philosophers have argued that a type of strong contextualism is an unavoidable consequence of seeking mechanistic explanations in general (i.e., not just a problem for cognitive neuroscience; see Lee and Dewhurst, 2021 for discussion) and therefore it is an unavoidable consequence of the explanatory goals of cognitive neuroscience. Here, we highlight two critical arguments to demonstrate strong contextualism; one regarding the functional mapping of the computational level to the brain level, and one regarding the parcellation of brain parts. We hope to show that these two epistemological issues demand methodological and theoretical re-orientation within the field that is much more significant than the demands of weak contextualism.
Dynamic functional mappings
First, our integrative review leads to the conclusion that behavioral context not only determines the contribution of a brain part to the organization of behavior (i.e., weak contextualism), but that context determines which computational operation that brain part implements in support of the behavior. That is, the computation a region performs is not a specialization of that region, but rather is determined by the behavioral and neural context in which the region finds itself, and can shift when the context changes (Sporns, 2011; Klein, 2012; Anderson, 2014, 2015a; Burnston, 2016b,2019, 2021; Khalidi, 2017; and see Mesulam, 1990; McIntosh, 1999, 2000, 2004 for early arguments in this direction). Klein (2012) illustrates the idea clearly with a discussion of the function of pistons in trucks with engine brakes: “Most of the time, [pistons] compress a fuel-air mixture to the point of detonation and transmit the generated power to the crankshaft. On trucks equipped with engine brakes, the pistons also have a second function: when the engine brake is engaged, the pistons use power from the wheels to compress air in the cylinder, slowing the truck. Which function the piston performs depends on things external to it: whether it is powering or slowing the truck depends on the ignition system and the valve timing” (p. 955, italics added; in this metaphor, this is the neural context). Notice that the function of the piston depends on whether we are interested in explaining the “going” or “stopping” action of the truck (the behavioral context); specifically, it is causally transmitting explosive force to the crankshaft in one instance and is reacting to a vacuum in the cylinder in the other. Here, we have one part (the piston) that is not simply performing an abstract function useful to both stopping and going (cf. Vingerhoets et al., 2013), but that has a functional description that is dependent on our explanatory goals. Thus, under the strong contextualism view, there is nothing specialized about the functional role of a piston – there is no single computational operation performed by the part that plays a role across phenomena. Thus, the mapping of the part to the computation is unstable. We argue that this conclusion holds for brain parts. The empirical evidence collected in the last section is consistent with the idea that, when seeking mechanistic explanations, brain regions are best modeled with different computations in different tasks (regardless of whether the computations are described mathematically or verbally). For example, in one context a brain region’s activity might be best modeled as multiplying an input signal whereas in another it is best modeled with addition; in one context a network might be best described as an “integrator,” while in another it is a “filter,” etc. Thus, functions of brain parts are not stable regardless of whether they are defined relative to the organismal, or computational, or neural level.
Note that we are not denying that neurons (and neural networks) have physiological, morphological and other neuroanatomical (e.g., topological) characteristics that ensure they can do some things and not others, in the same way we wouldn’t deny a piston’s physical characteristics that allow it to do some things and not others. Our point is that a structure’s participation in any given behavior, while obviously – and importantly – constrained by its properties, is not determined by those properties, and we cannot describe a part’s functional contribution in a context-independent way at any level within our mechanistic explanation.
Importantly, not even extreme abstraction of the putative computational function will allow us to recover a stable structure-function mapping for a part. This is the case given that we can choose to study an infinite number of phenomena at the behavioral level and that we will evolve skills in the future that we have not characterized yet. For instance, while the “production of complex learned movement” attributed to a neural network may apply to explanations of skilled action and expressive language (Vingerhoets et al., 2013), we are unable to rule out that the same neural network supports some behavior that does not require “complex learned movement,” simply because we haven’t studied all existing behavioral phenomena, nor can we know what phenomena will arise in the future. Thus, unlike weak contextualism, strong contextualism reveals an instability in structure-function mapping that prevents us from ever finding the computational function of brain parts.
Dynamic part ontologies
The second component of our argument relates to brain parts themselves. Because the identification of parts plays a central role in mechanistic explanation (Kaiser, 2018), identifying brain parts (defined as networks, anatomical regions, subregions, circuits, or single cells) is a cornerstone of cognitive neuroscience. Thus, a central issue in a cognitive neuroscience that takes mechanistic explanation seriously is determining the right “part ontology” (cf. Stanley et al., 2013; Viola and Zanin, 2017). However, strong contextualism shows that context determines not only a part’s computational role in any given behavior, but also determines what we should even consider to be a relevant part in the first place – that is, context determines the appropriate ontology of parts (cf. Poldrack and Yarkoni, 2016; Genon et al., 2018; Uddin et al., 2019). Strong contextualism challenges the idea that there is any context-independent parcellation of the brain that cognitive neuroscience can use in its mapping efforts. That is, changing our explanatory goals (i.e., which behavioral phenomenon we seek to explain) results in a redefining of the causally relevant parts that give rise to the phenomenon (see Craver and Kaplan, 2020). Indeed, it follows from the strong contextualism described here that (1) the boundaries of functionally relevant brain parts can shift every time we identify and want to explain a new behavioral phenomenon of interest and map the requisite computations onto brain parts, (2) these boundaries need not follow any obvious structural boundaries, and (3) they might shift even within participants, depending on context. These are not mere speculations. Consistent with point (2), functional boundaries within the brain need not follow any obvious neuroanatomical boundaries. For example, King et al. (2019) reported that functional subdivisions of the cerebellum did not coincide with lobular boundaries. Further, parcellation requires thresholding and clustering approaches, and it is well-known that different approaches will lead to different neuroanatomical maps (Sporns, 2012). Additionally, if, as we have argued above, the computational operation of a part is not context-independently predetermined by its material properties alone, then there is no reason to assume that its size, shape, or position would be context-independently definable for neural explanations of behavior. To see this, consider the following instructive metaphor in which a fictional researcher is interested in mapping function to structure in the extraneural human body. In describing high-fiving, the researcher identifies the hand as a whole as a functional part, while in describing the feeling of soft materials for pleasure, (parts of) individual finger segments constitute the relevant parts. In line with this analysis, and consistent with (1) and (3), Salehi et al. (2020) recently found that the boundaries of functional brain parts shift depending on the behavioral state of the participant. More specifically, they “demonstrate that the parcels are indeed consistent for a given condition, but reproducibly reconfigure across conditions, even when starting with the same initial atlas each time” (p. 2). This evidence shows that the foundational assumptions of cognitive neuroscience that it can map the brain parts to the computations that support organismal level behavior are not tenable.
Overall, strong contextualism presents an underappreciated characterization of brain-behavior interactions, revealing that context shifts are associated with instability of mapping between organismal, computational, and neural levels. It also leads to the counterintuitive conclusion that context determines part ontology (and not the reverse; see section “Discussion: Implications for Cognitive Neuroscience” below for more details). This epistemological consideration lays bare a unique challenge to cognitive neuroscience’s goal of “an understanding how the functions of the physical brain can yield the thoughts, ideas, and beliefs” of the mind (Gazzaniga et al., 2019, p. 4).
Discussion: Implications for cognitive neuroscience
As discussed above, much of the current infrastructure of cognitive neuroscience has been and continues to be shaped by structure-functional brain mapping, and this has real consequences for how money, time, and physical resources get distributed, which includes how undergraduate and graduate students are introduced to the field and how knowledge is classified and disseminated. In our opinion, strong contextualism should force the field to seriously reconsider its approach to the study of how organismal-level functioning maps onto the computational level and how this maps onto the neural level. Below, we gesture toward a number of implications that researchers should consider when pursuing functional brain mapping.
Strong contextualism forces a shift in the goal of mainstream cognitive neuroscience toward developing relevant part ontologies and seeking computational mechanistic explanation in context. First, we must recognize that by deciding that we are interested in a particular organismal-level phenomenon in a particular situation (say, visual decision making in a dual rather than a single task setting), we “fix” the part ontology which best helps explain the behavior computationally – though of course we don’t know in advance what that part ontology is (in the same way that deciding whether we are interested in going or stopping helps fix the part ontology for the best explanation of truck behavior). Thus, cognitive neuroscience should abandon the aim of describing the context-independent specialized functions of well-defined brain parts. Rather, in specific contexts it should seek to answer the empirical question of what are relatively stable parcellations, likely at the level of the individual or at a minimum of a small subgroup of individuals, that together with context-dependent computational descriptions lead to robust explanations of the behavioral phenomena of interest (Salehi et al., 2020; Viola, 2020). That is, we switch from asking “What does neural part X do, and is it important for behavior Y?” to “Given behavior Y, what neural parts – maybe X, maybe Z – and associated putative computational functions are needed to explain it?”
Second, contrary to previous proposals that attempt to deal with weak contextualism (Price and Friston, 2005; Poldrack, 2010), our analysis suggests that the determination of a part ontology should actually take place at a relatively low (rather than high) level of abstraction, rendering it more useful for explanations in specific contexts (Klein, 2012; Anderson, 2014; Burnston, 2016b,2021). For instance, the hippocampus is a behaviorally promiscuous region that appears in explanations within many areas of cognitive neuroscience, most commonly in accounts of spatial or mnemonic behaviors (Jeffery, 2018, for a brief history) but other types of tasks as well, leading some to suggest there is an impasse in theorizing about hippocampus function (Ekstrom and Ranganath, 2018; see Humphreys et al., 2021 for a similar discussion centering on the angular gyrus). Consistent with weak contextualism, one approach to this staggering complexity has been to offer relatively abstract functional descriptions that could account for the hippocampus’s role in all of these phenomena, for instance pattern separation/completion (Yassa and Stark, 2011), scene construction (see Maguire et al., 2016 for debate), or context equivalency (Maurer and Nadel, 2021). Notice that these are all descriptions of functions that are abstract enough to be equally applicable across many parts of the brain and clearly show conceptual overlap (in at least the verbal accounts of their computational roles); further, it is hard to imagine identifying a behavioral phenomenon that would not implicate these types of computational functions. In contrast, strong contextualism points to why such functional promiscuity exists, and suggests that the hippocampus may be best understood, mechanistically, as a part that implements one type of computational operation that supports some behaviors in some tasks (e.g., memory in memory tasks), but other operations in other tasks (e.g., in navigation tasks). Importantly, the hippocampus, as a brain part defined in a context-independent way, may not be the right “part” for explanations of some other tasks which may require further decomposition into additional parts of the hippocampus itself (e.g., in imagination tasks). Another natural consequence from this analysis is that a functional part may crosscut gross neuroanatomical boundaries. As a hypothetical example, in some contexts but not others subregions of both the angular gyrus and the supramarginal gyrus may together constitute a functional part. Ultimately, the part ontology will depend on whether we are interested in explaining memory, navigation, or imagination.
Note that we are not denying that the hippocampus (and its neural contexts) have anatomical structure and physiological properties that we can characterize – and in fact, in moving away from focusing on context-independent functional descriptions, the description of such constraints takes on an increased importance (see also Anderson, 2015a; Bolt et al., 2017). For instance, there is evidence that, given its physiology, the hippocampus implements sequence generation (Buzsáki and Tingley, 2018). However, the question is about the context-independent relevance of any given anatomical or physiological feature for explanations within cognitive neuroscience. We are suggesting that the degree to which a particular part (e.g., hippocampus) with particular properties will make a stable appearance in computational mechanistic explanations of particular behaviors across contexts, and whether the mechanistic explanation (with a particular part ontology) results in strong predictions for behaviors of interest, is an empirical question to be addressed in cognitive neuroscience under the specter of strong contextualism.
Finally, the implications of strong contextualism extend beyond strategies for empirical work; it affects how we theorize and interpret results. Indeed, it is recognized by many that the behavioral and neural sciences are in a “theory crisis” (Eronen and Bringmann, 2021). Our review suggests that one source of the problem is failing to recognize the instability of mapping “paying attention,” “using a tool,” and “speaking” to computational operations and brain parts. Strong contextualism, then, suggests an additional way to address the theory crisis: By developing formal computational mechanistic explanations in context, cognitive neuroscience can explicitly test effective part ontologies with associated computational operations and functional roles based on the phenomena it seeks to understand, where the phenomena it seeks to understand are open to revision and new classification. Doing so opens up cognitive neuroscience not only to better “part ontologies” but also to more useful cognitive ontologies (Poldrack, 2010). Such an approach may lead to further advancements in which historic ontologies may be dismantled in favor of ones with potentially greater accuracy and hence hopefully wider reaching clinical and scientific impact (e.g., see Renoult et al., 2019, for an example of the dismantling of the episodic vs. semantic distinction and challenges this poses to conceptions of “memory”; see Anderson, 2011 for challenges to the reification of “attention”; see Buzsáki, 2020, for challenges to the distinction between perception and action, etc.). Thus, we foresee an iterative process in which respecting the notion that context fixes the part ontology that best explains behavior allows us to remain open to other functional mappings in other contexts, and searching for empirically useful part ontologies will also shape how we identify and characterize the behavioral phenomena we wish to explain (see also Anderson, 2015b). This is a suggestion that goes beyond stating we need to understand behavior better before we should seek mechanistic explanations (Krakauer et al., 2017); rather, it opens up the possibility of characterizing both part and behavioral ontology differently than we do now, making way for new theoretical approaches and insights into brain-behavior interactions in a way that allows cognitive neuroscience to feed back into the behavioral sciences.
We have encountered resistance to the arguments presented here. Strong contextualism is counterintuitive and there are two common reactions to these arguments that we want to briefly address before we conclude. First, one might accept weak contextualism but conclude the strong version is a step too far. Again, as we have shown, many researchers hold that there is a single abstract computational process that each region performs, and we just have to wait for technology and/or methodology to catch up before we can successfully map that region to a computation that maps to behavior. We have already addressed the limitations of this conclusion but want to highlight here that our arguments regarding strong contextualism are not technological or methodological, but primarily epistemological (and our integrative review provides empirical support for the epistemological issues we have highlighted). Advances in technology or methodology without theoretical advances will not address the problems identified in the sections “Challenges to the Practice of Structure-Function Brain Mapping: An Integrative Review” and “Strong Contextualism, Instability of Mapping, and Indeterminate Part Ontology for Cognitive Neuroscience.”
Second, one might argue that contextualized explanation is already the current state of the art, in which researchers in disconnected (siloed) subfields of cognitive neuroscience identify context-sensitive functions for brain parts, no one is really looking for context-independent explanations in cognitive neuroscience, and everyone already acknowledges that there is no principled way of discovering the mapping between a structure and its computations. If this was indeed the state of the art, there would be no reason for the anatomical parcellations to be similar across silos, as accepting strong contextualism means accepting empirically defined parcellations that will be contextually driven. However, parcellations are similar across silos (e.g., hippocampus is mainly structurally defined in both memory and navigation research), suggesting that researchers do think there is a meaningful context-independent parcellation of the brain that can be mapped to computations and behaviors of interest.
Ultimately, the efforts of cognitive neuroscience will be judged by their utility in fulfilling its goals as stated above (Gazzaniga et al., 2019; Journal of Cognitive Neuroscience, 2021). It is our contention that taking strong contextualism seriously, both when it comes to determining computational operations, and when it comes to determining the size, shape and location of the neural parts that instantiate these operations, will position us better to mechanistically explain the clinical behavioral observations that this article started with, and even to more fully understand what it means for us to pay attention, use a tool, or speak to one another.
Author contributions
All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication. Authors have contributed equally to this manuscript.
Funding
HM acknowledges the support of the Natural Sciences and Engineering Research Council of Canada (NSERC), Discovery Grant RGPIN-2018-06139.
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Acknowledgments
HM would like to thank the students in the lab at UNBC for fruitful discussions of the main thesis.
Footnotes
- ^ While cognitive neuroscience is currently dominated by cognitivist assumptions of representation, computation and mechanism (which we here call the mainstream approach), 4E (embodied, embedded, enactive, extended) as well as ecological approaches to the study of the brain are becoming increasingly prevalent. Further, recent work aims to merge key insights from both traditions (e.g., Piccinini, 2022). In the present article, we have chosen to limit our analysis of research practices as they occur in the mainstream cognitive neuroscience literature. A similar analysis of neuroscientific research practices within 4E and ecological frameworks would be a rather different enterprise and is outside the scope of the current article (for examples of work in this tradition, see Gibson, 1966; Chiel and Beer, 1997; Barrett, 2011; van Orden et al., 2012; Dotov, 2014; Anderson, 2014; Kiverstein and Miller, 2015; de Wit et al., 2017; Hutto et al., 2017; Dewhurst, 2018; Bruineberg and Rietveld, 2019; de Wit and Withagen, 2019; van der Weel et al., 2019; Ryan and Gallagher, 2020; Raja, 2021; Raja and Anderson, 2021).
- ^ The topic of how cognitive neuroscience has come to decide on which explananda are worth pursuing is interesting on its own, and a point we return to later, in the section “Discussion: Implications for Cognitive Neuroscience” (see Fancher, 1996; Kästner, 2017 for discussion). Further, more detailed discussion of the role of mechanism in science is beyond the scope of this manuscript (see Cartwright et al., 2020). Our argument assumes a commitment to understanding cognition mechanistically, that is, being explicable in terms of the joint action of parts. Even if one denies a specific characterization of mechanism as discussed in the philosophy of science, we think it uncontroversial that there is at least an implicit and informal understanding that cognitive neuroscience, like the cognitive psychology it inherits, is often seeking mechanistic explanation.
- ^ Note that different researchers may flesh out the second, computational level to a greater or lesser degree, and describe its connections to the third, neural level in more or less detail. Some researchers build full-fledged sophisticated computational models without paying much attention to the neural level, while other researchers attempt to characterize functions of brain regions using a combination of behavioral and neuroimaging methods, without fleshing out the computational operations of those regions in any detail.
- ^ A common alternative interpretation of the co-occurrence of symptoms is that lesions “do not color within the lines.” That is, that they may impact multiple – smaller – spatially co-located functionally specialized regions. However, this line of reasoning does not seem readily capable of accommodating the above-reported “crossed” cases and the impact of treatment in one behavioral domain on performance in another behavioral domain.
- ^ It is important to exclude the possibility that what seems like degeneracy at the neural level is in fact driven by organismal-level variability between or even within participants in cognitive or behavioral strategies when performing a task (Gardner et al., 2013; Berneiser et al., 2018).
References
Abrevaya, S., Sedeño, L., Fitipaldi, S., Pineda, D., Lopera, F., Buritica, O., et al. (2017). The road less traveled: alternative pathways for action-verb processing in parkinson’s disease. J. Alzheimers Dis. 55, 1429–1435. doi: 10.3233/JAD-160737
Agis, D., and Hillis, A. E. (2017). The cart before the horse: when cognitive neuroscience precedes cognitive neuropsychology. Cogn. Neuropsychol. 34, 420–429. doi: 10.1080/02643294.2017.1314264
Ames, D. L., and Fiske, S. T. (2010). Cultural neuroscience. Asian J. Soc. Psychol. 13, 72–82. doi: 10.1111/j.1467-839X.2010.01301.x
Anderlini, D., Wallis, G., and Marinovic, W. (2019). Language as a predictor of motor recovery: the case for a more global approach to stroke rehabilitation. Neurorehabil. Neural Repair 33, 167–178. doi: 10.1177/1545968319829454
Anderson, B. (2011). There is no such thing as attention. Front. Psychol. 2:246. doi: 10.3389/fpsyg.2011.00246
Anderson, M. L. (2010). Neural reuse: a fundamental organizational principle of the brain. Behav. Brain Sci. 33, 245–66; discussion266–313.
Anderson, M. L. (2014). After Phrenology: Neural Reuse And The Interactive Brain. Cambridge, MA: The MIT Press.
Anderson, M. L. (2015a). Beyond componential constitution in the brain: starburst amacrine cells and enabling constraints. Open MIND 1, 1–13. doi: 10.15502/9783958570429
Anderson, M. L. (2015b). Mining the brain for a new taxonomy of the mind. Philos. Compass 10, 68–77. doi: 10.1111/phc3.12155
Anderson, M. L. (2016). “Neural reuse and in-principle limitations on reproducibility in cognitive neuroscience,” in Reproducibility: Principles, Problems, Practices, and Prospects, eds H. Atmanspacher and S. Maasen (Hoboken, NJ: John Wiley and Sons, Ltd), 341–362. doi: 10.1002/9781118865064.ch16
Anderson, M. L., and Oates, T. (2010). A critique of multi-voxel pattern analysis. Proc. Annu. Meet. Cogn. Sci. Soc. 32, 1511–1516.
Anderson, M. L., Kinnison, J., and Pessoa, L. (2013). Describing functional diversity of brain regions and brain networks. NeuroImage 73, 50–58. doi: 10.1016/j.neuroimage.2013.01.071
Bair, W. (2005). Visual receptive field organization. Curr. Opin. Neurobiol. 15, 459–464. doi: 10.1016/j.conb.2005.07.006
Bargmann, C. I. (2012). Beyond the connectome: how neuromodulators shape neural circuits. BioEssays 34, 458–465. doi: 10.1002/bies.201100185
Barrett, L. (2011). Beyond The Brain: How Body And Environment Shape Animal And Human Minds. Princeton, NJ: Princeton University Press.
Bateson, P. P. G., and Gluckman, P. D. (2011). Plasticity, Robustness, Development and Evolution. Cambridge: Cambridge University Press.
Bechtel, W., and Abrahamsen, A. (2010). Dynamic mechanistic explanation: computational modeling of circadian rhythms as an exemplar for cognitive science. Stud. Hist. Philos. Sci. Part A 41, 321–333. doi: 10.1016/j.shpsa.2010.07.003
Bedny, M. (2017). Evidence from blindness for a cognitively pluripotent cortex. Trends Cogn. Sci. 21, 637–648. doi: 10.1016/j.tics.2017.06.003
Behrmann, M., and Plaut, D. C. (2014). Bilateral hemispheric processing of words and faces: evidence from word impairments in prosopagnosia and face impairments in pure alexia. Cereb. Cortex (New York, N.Y.: 1991) 24, 1102–1118. doi: 10.1093/cercor/bhs390
Bergeron, V. (2007). Anatomical and functional modularity in cognitive science: shifting the focus. Philos. Psychol. 20, 175–195. doi: 10.1080/09515080701197155
Berneiser, J., Jahn, G., Grothe, M., and Lotze, M. (2018). From visual to motor strategies: training in mental rotation of hands. NeuroImage 167, 247–255. doi: 10.1016/j.neuroimage.2016.06.014
Biduła, S. P., Przybylski, Ł, Pawlak, M. A., and Króliczak, G. (2017). Unique neural characteristics of atypical lateralization of language in healthy individuals. Front. Neurosci. 11:525. doi: 10.3389/fnins.2017.00525
Bolt, T., Anderson, M. L., and Uddin, L. Q. (2017). Beyond the evoked/intrinsic neural process dichotomy. Netw. Neurosci. 2, 1–22. doi: 10.1162/NETN_a_00028
Borra, E., and Luppino, G. (2018). Large-scale temporo–parieto–frontal networks for motor and cognitive motor functions in the primate brain. Cortex 118, 19–37. doi: 10.1016/j.cortex.2018.09.024
Bowren, M. D., Tranel, D., and Boes, A. D. (2021). Preserved cognition after right hemispherectomy. Neurol. Clin. Pract. 11, e906–e908.
Bracci, S., Daniels, N., and Op de Beeck, H. (2017). Task context overrules object- and category-related representational content in the human parietal cortex. Cereb. Cortex 27, 310–321. doi: 10.1093/cercor/bhw419
Broca, P. (1861). Remarks on the seat of the faculty of articulated language, following an observation of aphemia (loss of speech). Bull. Soc. Anat. 6, 330–357.
Bruineberg, J., and Rietveld, E. (2019). What’s inside your head once you’ve figured out what your head’s inside of. Ecol. Psychol. 31, 198–217. doi: 10.1080/10407413.2019.1615204
Burnston, D. C. (2016b). A contextualist approach to functional localization in the brain. Biol. Philos. 31, 527–550. doi: 10.1007/s10539-016-9526-2
Burnston, D. C. (2016a). Computational neuroscience and localized neural function. Synthese 193, 3741–3762. doi: 10.1007/s11229-016-1099-8
Burnston, D. C. (2019). Getting over atomism: functional decomposition in complex neural systems. Br. J. Philos. Sci. 72, 743–772. doi: 10.1093/bjps/axz039
Burnston, D. C. (2021). Contents, vehicles, and complex data analysis in neuroscience. Synthese 199, 1617–1639. doi: 10.1016/j.neuroscience.2016.06.014
Buzsáki, G. (2020). The brain–cognitive behavior problem: a retrospective. Eneuro 7:ENEURO.0069-20.2020. doi: 10.1523/ENEURO.0069-20.2020
Buzsáki, G., and Tingley, D. (2018). Space and time: the hippocampus as a sequence generator. Trends Cogn. Sci. 22, 853–869.
Cabeza, R., and Nyberg, L. (2000). Imaging cognition II: an empirical review of 275 pet and fMRI studies. J. Cogn. Neurosci. 12, 1–47. doi: 10.1162/08989290051137585
Cartwright, N., Pemberton, J., and Wieten, S. (2020). Mechanisms, laws and explanation. Eur. J. Philos. Sci. 10, 1–19.
Chiel, H. J., and Beer, R. D. (1997). The brain has a body: adaptive behavior emerges from interactions of nervous system, body and environment. Trends Neurosci. 20, 553–557. doi: 10.1016/S0166-2236(97)01149-1
Cisek, P. (2007). Cortical mechanisms of action selection: the affordance competition hypothesis. Philos. Trans. R. Soc. B 362, 1585–1599. doi: 10.1098/rstb.2007.2054
Clopath, C., Bonhoeffer, T., Hübener, M., and Rose, T. (2017). Variance and invariance of neuronal long-term representations. Philos. Trans. R. Soc. B 372:20160161. doi: 10.1098/rstb.2016.0161
Craver, C. F. (2014). “Levels,” in Open Mind, eds T. Metzinger and J. M. Windt (Frankfurt am Main: MIND Group).
Craver, C. F., and Kaplan, D. M. (2020). Are more details better? On the norms of completeness for mechanistic explanations. Br. J. Philos. Sci. 71, 287–319.
Craver, C. F., and Tabery, J. (2015). Mechanisms In Science. Available online at: https://stanford.library.sydney.edu.au/archives/sum2017/entries/science-mechanisms/ (accessed May 28, 2020).
Çukur, T., Nishimoto, S., Huth, A. G., and Gallant, J. L. (2013). Attention during natural vision warps semantic representation across the human brain. Nat. Neurosci. 16, 763–770. doi: 10.1038/nn.3381
De Brigard, F. (2017). Cognitive systems and the changing brain. Philos. Explor. 20, 224–241. doi: 10.1080/13869795.2017.1312503
de Wit, M. M., and Withagen, R. (2019). What should a “gibsonian neuroscience” look like? Introduction to the special issue. Ecol. Psychol. 31, 147–151. doi: 10.1080/10407413.2019.1615203
de Wit, M. M., de Vries, S., van der Kamp, J., and Withagen, R. (2017). Affordances and neuroscience: steps towards a successful marriage. Neurosci. Biobehav. Rev. 80, 622–629. doi: 10.1016/j.neubiorev.2017.07.008
de Wit, M. M., Van der Kamp, J., and Masters, R. S. W. (2012). Distinct task-independent visual thresholds for egocentric and allocentric information pick up. Conscious. Cogn. 21, 1410–1418. doi: 10.1016/j.concog.2012.07.008
Dehaene, S., and Cohen, L. (2007). Cultural recycling of cortical maps. Neuron 56, 384–398. doi: 10.1016/j.neuron.2007.10.004
Dewhurst, J. (2018). Context-Sensitive ontologies for a non-reductionist cognitive neuroscience. Aust. Philos. Rev. 2, 224–228. doi: 10.1080/24740500.2018.1552102
Dixon, R. A., Garrett, D. D., Bäckman, L., Stuss, D. T., and Winocur, G. (2008). Principles of compensation in cognitive neuroscience and neurorehabilitation. Cogn. Neurorehabil. 2, 22–38. doi: 10.1080/09602011.2014.1003947
Dotov, D. G. (2014). Putting reins on the brain. How the body and environment use it. Front. Hum. Neurosci. 8:795. doi: 10.3389/fnhum.2014.00795
Edelman, G. M., and Gally, J. A. (2001). Degeneracy and complexity in biological systems. Proc. Natl. Acad. Sci. U.S.A. 98, 13763–13768. doi: 10.1073/pnas.231499798
Ekstrom, A. D., and Ranganath, C. (2018). Space, time, and episodic memory: the hippocampus is all over the cognitive map. Hippocampus 28, 680–687.
Eronen, M. I., and Bringmann, L. F. (2021). The theory crisis in psychology: how to move forward. Perspect. Psychol. Sci. 16, 779–788. doi: 10.1177/1745691620970586
Fedorenko, E., and Blank, I. A. (2020). Broca’s area is not a natural kind. Trends Cogn. Sci. 24, 270–284. doi: 10.1016/j.tics.2020.01.001
Figdor, C. (2010). Neuroscience and the multiple realization of cognitive functions. Philos. Sci. 77, 419–456. doi: 10.1086/652964
Fotopoulou, A. (2014). Time to get rid of the ‘modular’ in neuropsychology: a unified theory of anosognosia as aberrant predictive coding. J. Neuropsychol. 8, 1–19. doi: 10.1111/jnp.12010
Friston, K. J., and Price, C. J. (2003). Degeneracy and redundancy in cognitive anatomy. Trends Cogn. Sci. 7, 151–152. doi: 10.1016/S1364-6613(03)00054-8
Gallivan, J. P., and Culham, J. C. (2015). Neural coding within human brain areas involved in actions. Curr. Opin. Neurobiol. 33, 141–149. doi: 10.1016/j.conb.2015.03.012
García, A. M., Sedeño, L., Herrera Murcia, E., Couto, B., and Ibáñez, A. (2017). A lesion-proof brain? Multidimensional sensorimotor, cognitive, and socio-affective preservation despite extensive damage in a stroke patient. Front. Aging Neurosci. 8:335. doi: 10.3389/fnagi.2016.00335
Gardner, M. R., Brazier, M., Edmonds, C. J., and Gronholm, P. C. (2013). Strategy modulates spatial perspective-taking: evidence for dissociable disembodied and embodied routes. Front. Hum. Neurosci. 7:457. doi: 10.3389/fnhum.2013.00457
Garson, J. (2016). A Critical Overview of Biological Functions. Cham: Springer International Publishing, doi: 10.1007/978-3-319-32020-5
Gazzaniga, M. S., Ivry, R. B., and Mangun, G. R. (2019). Cognitive Neuroscience: The Biology Of The Mind, 5th Edn. New York, NY: W.W. Norton and Company.
Genon, S., Reid, A., Langner, R., Amunts, K., and Eickhoff, S. B. (2018). How to characterize the function of a brain region. Trends Cogn. Sci. 22, 350–364. doi: 10.1016/j.tics.2018.01.010
Glasser, M. F., Coalson, T. S., Robinson, E. C., Hacker, C. D., Harwell, J., Yacoub, E., et al. (2016). A multi-modal parcellation of human cerebral cortex. Nature 536, 171–178. doi: 10.1038/nature18933
Goldenberg, G., and Randerath, J. (2015). Shared neural substrates of apraxia and aphasia. Neuropsychologia 75, 40–49. doi: 10.1016/j.neuropsychologia.2015.05.017
Harel, A., Kravitz, D. J. D., and Baker, C. I. C. (2014). Task context impacts visual object processing differentially across the cortex. Proc. Natl. Acad. Sci. U.S.A. 111, E962–E971. doi: 10.1073/pnas.1312567111
Harnish, S., Meinzer, M., Trinastic, J., Fitzgerald, D., and Page, S. (2014). Language changes coincide with motor and fMRI changes following upper extremity motor therapy for hemiparesis: a brief report. Brain Imaging Behav. 8, 370–377. doi: 10.1007/s11682-011-9139-y
Hartwigsen, G. (2018). Flexible redistribution in cognitive networks. Trends Cogn. Sci. 22, 687–698. doi: 10.1016/j.tics.2018.05.008
Hartwigsen, G., and Saur, D. (2019). Neuroimaging of stroke recovery from aphasia – Insights into plasticity of the human language network. NeuroImage 190, 14–31. doi: 10.1016/j.neuroimage.2017.11.056
Hasson, U., Nastase, S. A., and Goldstein, A. (2020). Direct fit to nature: an evolutionary perspective on biological and artificial neural networks. Neuron 105, 416–434. doi: 10.1016/j.neuron.2019.12.002
Humphreys, G. F., Ralph, M. A. L., and Simons, J. S. (2021). A unifying account of angular gyrus contributions to episodic and semantic cognition. Trends Neurosci. 44, 452–463. doi: 10.1016/j.tins.2021.01.006
Hutto, D. D., Peeters, A., and Segundo-Ortin, M. (2017). Cognitive ontology in flux: the possibility of protean brains. Philos. Explor. 9795, 1–17. doi: 10.1080/13869795.2017.1312502
Jeffery, K. J. (2018). The hippocampus: from memory, to map, to memory map. Trends Neurosci. 41, 64–66.
Jiang, H., Stanford, T. R., Rowland, B. A., and Stein, B. E. (2021). Association cortex is essential to reverse hemianopia by multisensory training. Cereb. Cortex 31, 5015–5023. doi: 10.1093/cercor/bhab138
Jones, M. (2018). Numerals and neural reuse. Synthese 197, 3657–3681. doi: 10.1007/s11229-018-01922-y
Journal of Cognitive Neuroscience (2021). Available online at: https://direct.mit.edu/jocn [Accessed May 18, 2021].
Kaiser, M. I. (2018). “The components and boundaries of mechanisms,” in The Routledge Handbook of Mechanisms and Mechanical Philosophy, eds S. Glennan and P. Illari (New York, NY: Routledge).
Kästner, L. (2017). Philosophy of Cognitive Neuroscience: Causal Explanations, Mechanisms and Experimental Manipulations. Berlin: De Gruyter.
Khalidi, M. A. (2017). Crosscutting psycho-neural taxonomies: the case of episodic memory. Philos. Explor. 20, 191–208. doi: 10.1080/13869795.2017.1312501
King, M., Hernandez-Castillo, C. R., Poldrack, R. A., Ivry, R. B., and Diedrichsen, J. (2019). Functional boundaries in the human cerebellum revealed by a multi-domain task battery. Nat. Neurosci. 22, 1371–1378. doi: 10.1038/s41593-019-0436-x
Kiverstein, J. D., and Miller, M. (2015). The embodied brain: towards a radical embodied cognitive neuroscience. Front. Hum. Neurosci. 9:237. doi: 10.3389/fnhum.2015.00237
Klein, C. (2012). Cognitive ontology and region- versus network-oriented analyses. Philos. Sci. 79, 952–960. doi: 10.1086/667843
Knops, A., Thirion, B., Hubbard, E. M., Michel, V., and Dehaene, S. (2009). Recruitment of an area involved in eye movements during mental arithmetic. Science 324, 1583–1585. doi: 10.1126/science.1171599
Kolb, B., and Gibb, R. (2013). Searching for the principles of brain plasticity and behavior. Cortex 58, 251–260.
Krakauer, J. W., Ghazanfar, A. A., Gomez-Marin, A., Maciver, M. A., and Poeppel, D. (2017). Neuroscience needs behavior: correcting a reductionist bias. Neuron 93, 480–490. doi: 10.1016/j.neuron.2016.12.041
Kravitz, D. J., Saleem, K. S., Baker, C. I., Ungerleider, L. G., and Mishkin, M. (2013). The ventral visual pathway: an expanded neural framework for the processing of object quality. Trends Cogn. Sci. 17, 26–49. doi: 10.1016/j.tics.2012.10.011
Kriegeskorte, N., and Diedrichsen, J. (2019). Peeling the onion of brain representations. Annu. Rev. Neurosci. 42, 407–432.
Kriegeskorte, N., and Wei, X. X. (2021). Neural tuning and representational geometry. arXiv [Preprint]. arXiv: 2104.09743.
Kriegeskorte, N., Mur, M., and Bandettini, P. (2008). Representational similarity analysis—connecting the branches of systems neuroscience. Front. Syst. Neurosci. 2:4. doi: 10.3389/neuro.06.004.2008
Maguire, E. A., Intraub, H., and Mullally, S. L. (2016). Scenes, spaces, and memory traces: what does the hippocampus do? Neuroscientist 22, 432–439.
Marder, E., Goeritz, M. L., and Otopalik, A. G. (2015). Robust circuit rhythms in small circuits arise from variable circuit components and mechanisms. Curr. Opin. Neurobiol. 31, 156–163. doi: 10.1016/j.conb.2014.10.012
Marr, D. (1982/2010). Vision: A Computational Investigation Into The Human Representation And Processing Of Visual Information. Cambridge, MA: MIT Press.
Matheson, H. E., Garcea, F. E., and Buxbaum, L. J. (2021). Scene context shapes category representational geometry during processing of tools. Cortex 141, 1–15. doi: 10.1016/j.cortex.2021.03.021
Maurer, A. P., and Nadel, L. (2021). The continuity of context: a role for the hippocampus. Trends Cogn. Sci. 25, 187–199.
McIntosh, A. R. (1999). Mapping cognition to the brain through neural interactions. Memory 7, 523–548. doi: 10.1080/096582199387733
McIntosh, A. R. (2000). Towards a network theory of cognition. Neural Netw. 13, 861–870. doi: 10.1016/S0893-6080(00)00059-9
McIntosh, A. R. (2004). Contexts and catalysts: a resolution of the localization and integration of function in the brain. Neuroinformatics 2, 175–182. doi: 10.1385/NI:2:2:175
Menary, R. (2015). “Mathematical cognition—a case of enculturation,” in Open MIND, eds T. Metzinger and J. M. Windt (Frankfurt am Main: MIND Group), 25. doi: 10.15502/9783958570818
Merabet, L. B., Hamilton, R., Schlaug, G., Swisher, J. D., Kiriakopoulos, E. T., Pitskel, N. B., et al. (2008). Rapid and reversible recruitment of early visual cortex for touch. PLoS One 3:e3046. doi: 10.1371/journal.pone.0003046
Mesulam, M.-M. (1990). Large-scale neurocognitive networks and distributed processing for attention, language, and memory. Ann. Neurol. 28, 597–613. doi: 10.1002/ana.410280502
Miller, E. K. (2000). The prefontral cortex and cognitive control. Nat. Rev. Neurosci. 1, 59–65. doi: 10.1038/35036228
Mogensen, J. (2011). Reorganization of the injured brain: implications for studies of the neural substrate of cognition. Front. Psychol. 2:7. doi: 10.3389/fpsyg.2011.00007
Murray, M. M., Thelen, A., Thut, G., Romei, V., Martuzzi, R., and Matusz, P. J. (2015). The multisensory function of the human primary visual cortex. Neuropsychologia 1, 1–9. doi: 10.1016/j.neuropsychologia.2015.08.011
Noppeney, U., Friston, K. J., and Price, C. J. (2004). Degenerate neuronal systems sustaining cognitive functions. J. Anat. 205, 433–442. doi: 10.1111/j.0021-8782.2004.00343.x
Noppeney, U., Penny, W. D., Price, C. J., Flandin, G., and Friston, K. J. (2006). Identification of degenerate neuronal systems based on intersubject variability. NeuroImage 30, 885–890. doi: 10.1016/j.neuroimage.2005.10.010
Parkinson, C., Liu, S., and Wheatley, T. (2014). A common cortical metric for spatial, temporal, and social distance. J. Neurosci. 34, 1979–1987. doi: 10.1523/JNEUROSCI.2159-13.2014
Pessoa, L. (2008). On the relationship between emotion and cognition. Nat. Rev. Neurosci. 9, 148–158.
Piccinini, G. (2022). Situated neural representations: solving the problems of content. Front. Neurorobotics 16:846979. doi: 10.3389/fnbot.2022.846979
Piccinini, G., and Craver, C. F. (2011). Integrating psychology and neuroscience: functional analyses as mechanism sketches. Synthese 183, 283–311. doi: 10.1007/s11229-011-9898-4
Piccinini, G., and Scarantino, A. (2011). Information processing, computation, and cognition. J. Biol. Phys. 37, 1–38. doi: 10.1007/s10867-010-9195-3
Piccinini, G., and Shagrir, O. (2014). Foundations of computational neuroscience. Curr. Opin. Neurobiol. 25, 25–30. doi: 10.1016/j.conb.2013.10.005
Poldrack, R. A. (2006). Can cognitive processes be inferred from neuroimaging data? Trends Cogn. Sci. 10, 59–63. doi: 10.1016/j.tics.2005.12.004
Poldrack, R. A. (2010). Mapping mental function to brain structure: how can cognitive neuroimaging succeed? Perspect. Psychol. Sci. 5, 753–761. doi: 10.1177/1745691610388777
Poldrack, R. A., and Yarkoni, T. (2016). From brain maps to cognitive ontologies: informatics and the search for mental structure. Annu. Rev. Psychol. 67, 1–26. doi: 10.1146/annurev-psych-122414-033729
Price, C. J. (2018). The evolution of cognitive models: from neuropsychology to neuroimaging and back. Cortex 107, 37–49. doi: 10.1016/j.cortex.2017.12.020
Price, C. J., and Friston, K. J. (2002). Degeneracy and cognitive anatomy. Trends Cogn. Sci. 6, 416–421. doi: 10.1016/S1364-6613(02)01976-9
Price, C. J., and Friston, K. J. (2005). Functional ontologies for cognition: the systematic definition of structure and function. Cogn. Neuropsychol. 22, 262–275. doi: 10.1080/02643290442000095
Price, C. J., Hope, T. T. M., and Seghier, M. L. (2016). Ten problems and solutions when predicting individual outcome from lesion site after stroke. NeuroImage 145, 200–208. doi: 10.1016/j.neuroimage.2016.08.006
Primaßin, A., Scholtes, N., Heim, S., Huber, W., Binkofski, F., and Werner, C. J. (2015). Determinants of concurrent motor and language recovery during intensive therapy in chronic stroke patients: four single-case studies. Front. Neurol. 6:215. doi: 10.3389/fneur.2015.00215
Raja, V. (2021). Resonance and radical embodiment. Synthese 199, 113–141. doi: 10.1007/s11229-020-02610-6
Raja, V., and Anderson, M. L. (2021). “Behavior considered as an enabling constraint,” in Neural Mechanisms, Vol. 17, eds F. Calzavarini and M. Viola (Cham: Springer International Publishing), 209–232. doi: 10.1007/978-3-030-54092-0_10
Raymer, A. M. (1999). Crossed apraxia: implications for handedness. Cortex 35, 183–199. doi: 10.1016/S0010-9452(08)70793-7
Renoult, L., Irish, M., Moscovitch, M., and Rugg, M. D. (2019). From knowing to remembering: the semantic–episodic distinction. Trends Cogn. Sci. 23, 1041–1057.
Rule, M. E., O’Leary, T., and Harvey, C. D. (2019). Causes and consequences of representational drift. Curr. Opin. Neurobiol. 58, 141–147.
Ryan, K. J. J., and Gallagher, S. (2020). Between ecological psychology and enactivism: is there resonance? Front. Psychol. 11:1147. doi: 10.3389/fpsyg.2020.01147
Salehi, M., Greene, A. S., Karbasi, A., Shen, X., Scheinost, D., and Constable, R. T. (2020). There is no single functional atlas even for a single individual: functional parcel definitions change with task. NeuroImage 208:116366. doi: 10.1016/j.neuroimage.2019.116366
Seifert, L., Komar, J., Araújo, D., and Davids, K. (2016). Neurobiological degeneracy: a key property for functional adaptations of perception and action to constraints. Neurosci. Biobehav. Rev. 69, 159–165. doi: 10.1016/j.neubiorev.2016.08.006
Shine, J. M., Eisenberg, I., and Poldrack, R. A. (2016). Computational specificity in the human brain. Behav. Brain Sci. 39:e131. doi: 10.1017/S0140525X1500165X
Stanley, M. L., Gessell, B., and De Brigard, F. (2019). Network modularity as a foundation for neural reuse. Philos. Sci. 86, 23–46. doi: 10.1086/701037
Stanley, M. L., Moussa, M. N., Paolini, B., Lyday, R. G., Burdette, J. H., and Laurienti, P. J. (2013). Defining nodes in complex brain networks. Front. Comput. Neurosci. 7:169. doi: 10.3389/fncom.2013.00169
Stoll, H., de Wit, M. M., Middleton, E. L., and Buxbaum, L. J. (2021). Treating limb apraxia via action semantics: a preliminary study. Neuropsychol. Rehabil. 31, 1145–1162. doi: 10.1080/09602011.2020.1762672
Tang, Y., Zhang, W., Chen, K., Feng, S., Ji, Y., Shen, J., et al. (2006). Arithmetic processing in the brain shaped by cultures. Proc. Natl. Acad. Sci. U.S.A. 103, 10775–10780. doi: 10.1073/pnas.0604416103
Uddin, L. Q., Kinnison, J., Pessoa, L., and Anderson, M. L. (2014). Beyond the tripartite cognition—emotion—interoception model of the human insular cortex. J. Cogn. Neurosci. 26, 16–27. doi: 10.1162/jocn_a_00462
Uddin, L. Q., Yeo, B. T. T., and Spreng, R. N. (2019). Towards a universal taxonomy of macro-scale functional human brain networks. Brain Topogr. 32, 926–942. doi: 10.1007/s10548-019-00744-6
Uttal, W. R. (2001). The New Phrenology: The Limits Of Localizing Cognitive Processes In The Brain. Cambridge, MA: MIT Press.
van der Weel, F. R., Agyei, S. B., and van der Meer, A. L. H. (2019). Infants’ brain responses to looming danger: degeneracy of neural connectivity patterns. Ecol. Psychol. 31, 182–197. doi: 10.1080/10407413.2019.1615210
van Orden, G., Hollis, G., and Wallot, S. (2012). The blue-collar brain. Front. Physiol. 3:207. doi: 10.3389/fphys.2012.00207
Vingerhoets, G., Alderweireldt, A. S., Vandemaele, P., Cai, Q., Van der Haegen, L., Brysbaert, M., et al. (2013). Praxis and language are linked: evidence from co-lateralization in individuals with atypical language dominance. Cortex 49, 172–183. doi: 10.1016/j.cortex.2011.11.003
Viola, M. (2020). Beyond the platonic brain: facing the challenge of individual differences in function-structure mapping. Synthese 199, 2129–2155. doi: 10.1007/s11229-020-02875-x
Viola, M., and Zanin, E. (2017). The standard ontological framework of cognitive neuroscience: some lessons from broca’s area. Philos. Psychol. 5089, 1–25. doi: 10.1080/09515089.2017.1322193
Wernicke, C. (1874/1969). “The symptom complex of aphasia,” in Proceedings of the Boston Colloquium for the Philosophy of Science 1966/1968, eds R. S. Cohen and M. W. Wartofsky (Dordrecht: Springer Netherlands), 34–97. doi: 10.1007/978-94-010-3378-7_2
Yassa, M. A., and Stark, C. E. (2011). Pattern separation in the hippocampus. Trends Neurosci. 34, 515–525.
Zednik, C. (2018). “Mechanisms in cognitive science,” in The Routledge Handbook of Mechanisms and Mechanical Philosophy, eds P. M. Illari and S. Glennan (London: Routledge), 389–400.
Zednik, C. (2019). “Computational cognitive neuroscience,” in The Routledge Handbook of the Computational Mind, eds M. Sprevak and M. Colombo (Abingdon: Routledge).
Zerilli, J. (2019). Neural reuse and the modularity of mind: where to next for modularity? Biol. Theory 14, 1–20. doi: 10.1007/s13752-018-0309-7
Keywords: cognitive neuroscience, degeneracy, functional brain mapping, levels of analysis, many-to-many mappings, mechanistic explanation, neural reuse, structure-function relationships
Citation: de Wit MM and Matheson HE (2022) Context-sensitive computational mechanistic explanation in cognitive neuroscience. Front. Psychol. 13:903960. doi: 10.3389/fpsyg.2022.903960
Received: 24 March 2022; Accepted: 27 June 2022;
Published: 22 July 2022.
Edited by:
Tomer Fekete, Ben-Gurion University of the Negev, IsraelReviewed by:
Gualtiero Piccinini, University of Missouri–St. Louis, United StatesVanessa Lux, Ruhr University Bochum, Germany
Copyright © 2022 de Wit and Matheson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Matthieu M. de Wit, matthieudewit@muhlenberg.edu; Heath E. Matheson, Heath.matheson@unbc.ca
†These authors have contributed equally to this work