Localization

Brain and Mind 3: 1-18, 2002.

Concepts of Localization: Balkanization in the Brain

JENNIFER MUNDALE

Dept.of Philosophy, University of Central Florida, CNH 411, Orlando, FL 32816-1352
USA (E-mail: jmundale@pegasus.cc.ucf.edu)

Abstract. A spate of recent anti-localizationist publications have re-ignited the old debate about the localization of function. Many of the recent attacks on localization, however, are directed at what I will argue to be a narrow and outmoded view of localization, and thus have little conceptual or empirical impact. What I hope to present here is an analysis of functional localization that more adequately reflects the sophistication and complexity of its use in neuroscientific research, both historically and recently. Proceeding first by way of contrast, I examine the anti-localizationist positions of holism and equipotentiationism. Then, I present a four-fold analysis of localization according to physical scope, physical kind, functional scope, and functional kind. Next, I turn to a discussion of the heuristic value of localization in deciphering structure-function relationships. Finally, I hope to show that the overall view of functional localization that emerges from these considerations constitutes a much more elusive target than its critics assume. It serves to mitigate, and in some instances even defeat, some forms of anti-localizationist criticisms.

Key words: anti-localization, complex systems, equipotentiationism, function, functional localization, heuristic, holism, localization, phrenology

1. Introduction

It is a stock scene: Holmes opens the morning paper, and then remarks to Watson, "There's trouble in the Balkans." Perhaps we should adopt a similar sentiment in acknowledging the recent resurgence of the debate between the holists and the localizationists. Every now and then this centuries-old debate erupts, usually in the wake of some major innovation in localization technique. For example; in the nineteenth century, Flourens (1794-1867) set out to destroy the phrenological system devised by Gall (1757-1828), and in the twentieth century, Lashley (1890-1958) tried to discredit the cellular-based or cytoarchitectonic methods so expertly employed by Brodmann (1868-1918) in developing his map of the cerebral cortex. Now, another group of anti-localizationist critics is emerging to challenge the methods and assumptions associated with neuroimaging and other recent technologies used in localization research. Chief among these are Uttal (2001), Lloyd (2000), and vanOrden and Paap (1997). Each of these authors offer conceptual, methodological, and technological criticisms of localization, with the technological criticisms aimed primarily at PET scanning and functional MRI (fMRI).
In what follows, I do not address the technological limitations and difficulties involved in localizing brain functions. Instead, I focus primarily on the conceptual difficulties surrounding localization. I begin the analysis by contrasting localization with other forms of anti-localizationist views (e.g. holism, equipotentiationism). Then, I present a four-fold analysis of functional localization according to physical kind and scope, and functional kind and scope. I also raise some considerations about the temporal aspects of recent localization research. I argue that some criticisms of functional localization are substantially mitigated when viewed in light of the foregoing analysis of localization. I contend that preoccupation with the exclusively spatial characteristics of functional areas constitutes a red herring in the localization debate. To understand how functions are implemented in the brain, I urge that a methodologically suitable view is one that emphasizes the heuristic role of localization. Bechtel and Richardson (1993) offer such a view, and it is also the view of localization that has been operative in neuroscientific localization research, both past and present.

2. Localization Contra Holism and Equipotentiationism

In the present context, of course, 'localization' refers to functional localization, or the determination of functionally distinct, physically discrete areas within the brain. Nearly all brain research presupposes some commitment to functional localization; for, if the brain could be understood only as a primitive simple, a functionally non-decomposable and singular entity, then the very project of analyzing the brain would fail to be relevant. Over a century ago, Meynert (1833-1892) expressed this idea clearly when he said:

If we look upon the cortex as an organ functioning as a whole then the information that it subserves the processes of the mind is all that can be said... To think further about the cortex is impossible and unnecessary... But our hope to understand eventually the function of the hemispheres is raised again by the opposite assumption which leads us straight to an organology of the central surface... Between these two theoretical possibilities the facts have to decide. (in Papez, 1953, p. 64)

More recently, Shallice (1988) expresses a contemporary version of similar ideas:

If Lashley's (1929) idea of mass action were valid, then neuropsychology would be of little relevance for understanding normal function. Any form of neurological damage would deplete by a greater or lesser degree the available amount of some general resource... Knowing which tasks a patient could or could not perform would enable us to partition tasks on a difficulty scale. It would tell us little, if anything, about how the system operated. (p. 18)

Shallice points to Karl Lashley (1890-1958), of course, because he was one of the arch anti-localizationists of the twentieth century. I will say more about him shortly.
      The two major anti-localizationist views are holism and equipotentiationism (henceforth, I will adopt the somewhat less clumsy alternative, equipotentialism). Extreme forms of either, particularly equipotentialism, can lead to the sort of pessimism expressed above about the value of neuroscientific investigation into the brain.
      Often, holism and equipotentialism are treated synonymously, but there are important differences between them. Of the two, equipotentialism is more antithetical to functional localization, because it asserts that the brain is functionally equivalent throughout. While equipotentialists do not deny that the brain carries out multiple functions, they do deny that it does so in virtue of functionally distinct components.
      The classic figure of equipotentialism is Lashley, well known for his criticisms of Brodmann's cytoarchitectonic (pertaining to cellular distribution patterns) method of marking functional divisions in the cortex (Lashley and Clark, 1946). He vigorously opposed any but the mildest functional differentiation of brain regions, arguing instead that:

There is evidence of mutual dependence of parts in which the specialization of structures seems less important than the mere mass of functional tissue. There are indications that within the entire cortex, for certain functions, and within specialized areas, for others, the subordinate parts are all equally capable of performing the functions of the whole. (1933, p. 34)

Uttal explicitly disavows equipotentialism (2001, p. 150, 209), but tacitly expresses support for views that are either consistent with equipotentialism, or that even constitute a mild form of it. He notes with apparent approval, for example, that:

The idea that brain regions are complexly interconnected and that the resulting interactions may be more characteristic of the brain as a whole than of any localized and specialized region has been percolating recently into the literature that compares cognitive processes and fMRI responses. (p.163)

To the extent that he is willing to deny regional, functional specificity, he takes on the trappings of an equipotentialist. However, if he is asserting only that the whole brain is involved in all cognitive processes, his views may just amount to a form of holism.
Holists are of the view that to understand how the brain subserves a given function, it is necessary to understand the whole brain (or least large portions of it) acting as an integrated, functional unit; they deny that any particular function can be understood as the isolated contribution of just one or even a few, select components. While holists typically have been opposed to localization, strictly speaking, holism need not be inconsistent with functional localization, since it doesn't deny outright the possibility of regional, functional variation.
In part, Finger regards the tension between holism and localization to be a matter of pragmatic emphasis:

As for holism, this is a concept that has continuously waxed and waned in popularity over the years. Yet the importance of looking at the workings of the whole brain, whether before or after injury, never had to be viewed as a direct challenge to the localizationist position. To be sure, there were clear differences in emphasis....Localization theory was aimed more at answering questions about the fundamental biological organization of the brain. Holism appealed more to individuals who were trying to understand why brain-damaged patients might show both positive as well as negative symptoms, and why maladaptive behaviors might vary with contextual and situational conditions. (1994, p. 61)

Generally, though, holism does not accord well with functional localization, and Finger also adds that the relationship between them has not always been perceived as a harmonious, non-competitive one.
      In the current literature, also, one finds arguments in support of the view that localizationist approaches need not be inconsistent either with holism, or some milder forms of equipotentialism. For example, the localization issue directly feeds into the reductionist vs. emergentist controversy; i.e., the controversy between those favoring a more mechanistic, componential, and decomposable analysis of the mind-brain, vs. those who view the mind as something that emerges, in some irreducible way, from the complex processes of the whole brain. Though both Bechtel (2001; 2002) and Bickle (2001) favor the reductive-localization approach over the emergent-holistic approach, they also argue, independently, that these aren't jointly inconsistent, but that the appropriateness of each, in part, depends on the level of explanation in question.
      As noted above, there is a tendency in the literature for equipotentialism and holism to be used interchangeably, and one could, of course, hold both positions. Lloyd (2000), van Orden and Paap (1997), and other complex systems proponents, appear to defend a contemporary mix of both equipotentialism and holism. There are several varieties of complex systems (a general term here meant to encompass such designations as connectionist, dynamic systems, parallel distributed processing or PDP, and similar systems), but typically, the components are heavily interconnected and complexly interactive, the processing is in parallel, and the processing burden is distributed. One of the essential features of a complex system is that its behaviors and properties cannot be predicted on the basis of the behaviors and properties of its components (Bechtel and Richardson, 1993, Bechtel 2002, vanOrden and Paap 1996, Bickle, 2001). The complex systems approach to the brain denies that the brain can be understood mechanistically, or by localization and decomposition into functionally transparent components (Bechtel and Richardson, 1993).
      In vanOrden and Paap's example, both holistic and equipotentialist elements of their complex systems model result from the property of reciprocal causality, a feature reflecting the highly interactive, dynamic nature of the system in which every component both affects and is affected by numerous other components. The system is holistic in that, "Reciprocal causality implies that each and every component of a system contributes to every behavior of the whole system "; the system is equipotentialist in that "Each component affects every other component, to the extent that their independent contributions cannot be sorted out in the behavior of the whole" (p. S92).
      Lloyd, nevertheless, connects equipotentialism with the property of distributed processing in his complex systems approach. In attempting to describe a continuum of positions between localized processing and distributed processing, he describes the most extreme form of a distributed system as follows:

...a brain could be a single fully distributed network. Here every anatomically defined brain region has a part to play in every cognitive function, and no region is out of play. 'Equipotentiality,' as Karl Lashley conceived it, is an early example of fully distributed processing. (2000, p. 95)

What Lloyd describes here, though, is not clearly an equipotentialist system, but merely a holistic one. For it to be the former, he would have to add the proviso that every brain region has the same part to play in every cognitive function. Distribution of processing does not, by itself, rule out functional specialization. Moreover, even his own subsequent data about the multiple, though differential activation of Brodmann's areas in tasks assessed by PET scan research is inconsistent with equipotentialism (pp. 99-100). The point is not merely a pedantic one, in that his construal of equipotentialism contributes to a much stronger anti-localizationist conclusion than is actually warranted.
The tendency to conflate holism with equipotentialism is not surprising, as those favoring one position are likely to be sympathetic with the other as well. And, obviously, they are both anti-localizationist, though the holists are less emphatically so. Holism can be seen as a kind of compromise position, or as a more attenuated form of equipotentialism. Yet, equipotentialism doesn't necessitate holism; for, to assert that the brain is everywhere functionally equivalent is not to assert that the entire brain is involved in all functions. For an equipotentialist, the ability perform a given function is directly proportional to the amount of brain area involved, such that lesioning or extirpating a part of the brain may only result in a diminished capacity to perform a given function. Simpler functions may not require the involvement of the whole brain. The essential point for the equipotentialist is that no one area is functionally distinct from any other. Henceforth I will reserve the general term 'anti-localizationist' to refer to the broader thrust of views against functional localization, particularly when addressing views that either fail to disambiguate holism from equipotentialism, or when the distinction is not germane.

3. A Four-Fold Analysis of Localization

At this point in the discussion it is already apparent that localization admits of degrees and varieties. In order to illustrate this more clearly, it is useful to consider four distinct categories or axes of analysis with respect to the localization of function:

1) Physical scope
2) Physical kind
3) Functional scope
4) Functional kind

Clearly, these are not mutually exclusive, and some researchers have expressed their views of localization in ways that combine some of the above categories.
First, by physical scope, I mean the size or scale of the brain piece that is purported to correspond to a distinct function; this a quantitative consideration. Relative localizationists on this axis will insist on the functional specificity of comparatively smaller regions of the brain than the holists will. Historically, localization researchers have increasingly refined their physical scope toward progressively smaller regions of the brain. Finger emphasizes this progression when he writes:

In the long history of the brain sciences, it is possible to conceive of the theory of localization as being applied to the whole and then to increasingly smaller parts. At first the question seemed to be, 'Why is the brain special, and how is it different from other organs such as the heart?'.... After this issue was resolved, attention was drawn to functional differences between gross anatomical divisions of the brain, such as the medulla, the cerebellum, and the cerebrum. Surprisingly, it was not until the nineteenth century that scientists seriously entertained the possibility that the cerebral cortex might be divided into distinct parts, each responsible for a different function. (1994, p. 3)

Similarly, those who were willing to compromise on this aspect of the debate were willing to admit that the grosser divisions of the brain might differ functionally from one another, but that localization to tiny, tightly circumscribed areas was not plausible.
One sees this kind of compromise in Goltz (1834-1902), for example. Although he tended to be an outspoken critic of localization, especially of the cytoarchitectonic work on localization in the cortex, he was nonetheless willing to admit that the lobes of the brain were at least somewhat functionally different. He was particularly vehement in denying precisely circumscribed areas of motor and sensory functions within the cortex. One of his best known experiments involved the purported removal of the motor cortices of a dog who not only survived the surgery, but was paraded in front of the International Medical Congress, in 1881, clearly able to run, jump, see, hear, and smell. These findings were later contested, however, since at post-mortem examination it was discovered that Goltz had not removed as much of the cortex as he had lead his colleagues to believe. This controversy was never fully resolved. Eventually, Goltz did come around to a compromise position, which, in his particularly engaging fashion, he confesses as follows:

Even if I was convinced that the assumption of small circumscribed areas did not correspond with the facts, that did not mean that the idea of Flourens, that the substance of the brain was equivalent everywhere, was correct. I tried to find out what consequences the amputation of single lobes of the brain had, and found that the destruction of the anterior lobes led to entirely different disturbances from the destruction of the occipital lobes.... If I now admit a certain localization, I denounce former sins, so I was told. To this I answered that I had no other calling than to find out and admit the truth. (1888, 1960 transl., pp. 129-130)

This concession was enough for Brodmann to call Goltz a "half-localizer" (1909, 1994 transl., p. 251).
For a more contemporary example, the association areas of the brain have so far proved to be comparatively resistant to fine-grained functional divisions; as association areas, they receive multi-modal input from many different cortical regions so are complexly integrative. The difficulty of localization within association areas is discussed by Andrew Kertesz, a noted neuroimaging expert. His discussion of association areas invokes the present category, physical scope:

Large areas of the brain that are commonly called tertiary association areas seem less clearly related to any particular function, and show a great deal of evidence of plasticity and interchangeability. Some of these tertiary association areas have, nevertheless, at least a hemispheric functional specialization. The unique role of some left hemisphere tertiary association areas in language has been well established since the time of Broca, although some evidence exists for oral and visual comprehension of concrete nouns and automatic speech output in the right hemisphere, which may not be as speechless as once thought. (1994, p. 11)

This example also underscores the point that the issue of localization vs. equipotentialism is not necessarily a dichotomous one; we may ultimately come to discover that some regions of the brain are less functionally specific than others.
      Next, physical kind is a qualitative consideration that refers to the structural characteristics and regularities of a given sample of brain tissue (i.e. kind of neuron, processes, tissues, etc.). I believe this captures a dimension of localization that is frequently overlooked in the literature, and helps to illustrate and support some of the claims I make in the next section. For some functions, their physical counterparts may be spread diffusely through the brain, and thus may not be neatly localizable spatially, but may still be localizable to a specific, qualitatively distinct kind. In cases like this, one might regard the function in question as multilocal. Otherwise put, it is still a move away from equipotentialism to correlate a function with a specific brain kind, even though that kind itself may be diffusely located, because it denies that every piece of the brain is functionally equivalent with another.
      Brodmann employs a sense of localization similar to this in his cytoarchitectonic research; he called it, "localisation according to individual histological elements" where the basic unit of consideration is not an individual cell, but cell groupings (1909, 1994 transl., p. 7).
      There are cases, however, when even a single kind of cell might serve to demarcate a functional region. Purkinje cells, for example, are found only in the second layer of the cerebellum, and their axons carry the sole neuronal output for the cerebellar cortex. They are distinguishable by their immense number of dendritic processes and arborizations; they have a "bushy" appearance. This structure is related to their functional role, which is to receive and integrate multiple cerebellar inputs from multiple somatotopic areas.
      Betz cells, which are extra-large, specialized versions of pyramidal cells, are found only in the fifth layer of the primary motor cortex. Brodmann recognizes this as an exception to the general principle of localization by cell groupings, rather than by a single kind of cell (1909, 1994 transl., p. 6). The fifth layer of the cortex is known to be a heavy output layer and it is prominent in the motor cortex, as one might expect. The fourth layer, by contrast, is a heavy input layer; it is meager to non-existent in the motor cortex, but pronounced in sensory processing areas such as primary visual cortex and primary somatosensory cortex. Thus, the laminar structure of the cortex itself constitutes another physical kind which, though spread throughout the cortex, nonetheless has multilocal functional significance. Among other criteria, Brodmann used regional variation in the comparative thickness and cellular composition of the cortical layers to demarcate functional regions (1909, 1994 transl., pp. 38-45).
      The serotonergic pathways are another example of localization by physical kind. Serotonin is a neurotransmitter produced by neurons that originate in the Raphe nuclei of the brainstem, but whose pathways project to several different regions of the brain. Not all targets of the serotonergic pathway are known, and not all effects of serotonin are understood. Nevertheless, there are some common physiological and structural features shared by all serotonergic neurons, and researchers have begun to associate certain forms of depression with defects involving the serotonergic and other pathways (Kandel, et al., 2000, pp. 1216-1225).
      By functional scope, I mean the relative generality of the function that is purported to be localized. Language, for example, is a very general, highly integrative function; whereas phonological processing is relatively more specific. Integrative functions such as reading, decision making, attention, are thought to involve more than one functional area, so are more complexly instantiated in the brain than functions that are comparatively more basic, such as auditory tone perception, visual fixation, shape discrimination, etc. Differential localizability according to generality of function is readily acknowledged by researchers engaged in charting the functional regions of the brain. For example, Petersen and Fiez, in a famous study involving PET research into language processing, note that, "A functional area of the brain is not a task area; there is no 'tennis forehand area' to be discovered. Likewise, no area of the brain is devoted to a very complex function: 'attention' or 'language' is not localized in a particular Brodmann area or lobe" (1993, p. 513). Similarly, there is no single, isolated area of the brain devoted to vision, but, as in the case of language, there are numerous, interactive subfunctions (color processing, shape processing, motion processing, etc.) that, by comparison, are more conspicuously localizable. This conception of localization also allows for a multilocal instantiation of function.
      Finally, by functional kind I mean a qualitative consideration that refers to the traditional distinction between higher and lower functions. Equipotentialists who compromised along this axis were willing to admit that the relatively primitive functions such as sensory and motor processing were localizable, but that the higher, more cognitively oriented functions were not. Many researchers have held this particular compromise position; Loeb (1859-1924) and von Monakow (1853-1930) are some of the more notable classical figures in this category (see Finger, 1994, pp. 55-8). It is also a position commonly held today. Even Uttal concedes some degree of localization for the lower functions when he writes:

There is no question that specialized sensory and motor regions exist and that other regions are involved in, if not dedicated to, particular cognitive processes in some yet-to-be-discovered ways. On the other hand, however, it seems clear that the more complex the psychological process, the less likely that a narrowly circumscribed region uniquely associated with that process will be found. (2001, p. 13)

Uttal's position, as stated above, represents a compromise position on localization. However, insofar as he conceives of localization as requiring a "narrowly circumscribed region", I believe he misconceives the nature of functional localization.
Uttal reveals his interpretation of localization more explicitly when he claims, "One of the basic guiding assumptions of any attempt to localize a cognitive function in the brain is that a circumscribed locale for that function actually exists" (p. 155). Uttal is not the only one to interpret localization in terms of discrete, neatly bounded, and spatially isolated areas. It has been a common, if informal way to think about localization at least since the time of Gall, and probably, largely because of Gall (Finger, 1994). Some functional delineations, such as those found in Brodmann's map of the cerebral cortex, do consist of such well-defined regions. However, there is no conceptual requirement to restrict localization to this interpretation. A more sophisticated view of functional localization that allows for multilocal possibilities is both conceptually coherent, and factually consistent with localization research. Thus, anti-localizationist criticisms based on enumerating examples of functions that are not restricted to narrow, spatially circumscribed areas, fail to hit their mark.
In a brief review, Marshall makes a similar point in quoting and criticizing Uttal's "restatement" of the problem of localization:

This restatement 'offers, in place of a specific function being precisely localized (that is, instantiated, represented or encoded) in a particular place, the idea of one centre contributing to the operation of a complex system of nodes and loci that are collectively responsible for the behaviour.' Well, yes: this is precisely what every behavioural neurologist and neuropsychologist has argued since (at least) Carl Wernicke's fractionation of the 'aphasic symptom complex' in 1874. (2001, p. 152)

Even Brodmann, one of the earliest and most enduring figures in brain mapping research, assumed that higher functions would result from the combined activity of several different cortical areas. He explicitly distances himself from the phrenologist's assumption of "higher order psychic centres", and with respect to complex functions, he writes,

One cannot think of their taking place in any other way than through an infinitely complex and involved interaction and cooperation of numerous elementary activities, with the simultaneous functioning of just as many cortical zones, and probably of the whole cortex... (1909, 1994 transl., p. 255)

He stops short of equipotentialism, however, adding that "this does not mean that all the individual organs make equal physiological contributions to higher psychic processes" (p. 255).
      The multi-local view that Brodmann expressed nearly a century ago is still endorsed by contemporary localizationists. He is cited approvingly by Raichle, for example, a contemporary leader in PET scan research, who writes, "With this [Brodmann's] prescient admonition in mind, the task of functional brain imaging becomes clear: identify regions and their temporal relationships associated with the performance of a well-designed task." (2001, p. 4).
      Raichle notes the importance of identifying temporal relationships among functional regions. For temporal resolution, EEG (electroencephalography), MEG (magnetoencephalography) and related methods are some of the most important and informative techniques. These methods can be used to track the temporal course of activation and (less reliably) localize the source of the activation.
      Not all localization techniques, however, have the temporal resolution necessary to accurately track duration of cortical activity. PET scan research, for example, has been very influential in our understanding of functional localization, particularly with respect to visual processing and language, though this technique isn't useful for temporal tracking. Brodmann's cytoarchitectonic map, arguably the most important foundation for modern localization research, was never intended to provide temporal information. Similarly, many of the landmark clinical cases, such as Broca's aphasic patient, "Tan", established areas of functional specialization without reference to any significant temporal information. So, historically, at least, the temporal domain has not been an essential or indispensable part of localization technique. As noted above, however, it certainly has been a factor in some methods and appears to be an increasingly important component of contemporary localization research. For this reason, I will briefly consider a temporalized version of functional localization as yet another dimension in which a function can be said to be localized.
      Conceptually, instead of understanding the isolated, "snapshot" contribution of a single, functional area, many recent studies have begun to decipher the functional significance of multiple, temporally extended and highly coordinated areas of activation. The added complexity of spatio-temporal localization is not limited to the possibility of simple, multi-local activation over time, but can involve the relational (e.g., synchronizational, oscillatory, inhibitory, sequential etc.) aspects of multiple activations of multiple areas over time. In many cases, adding relational considerations to our understanding of how functions are localized in the brain may require us to consider daunting levels of combinatorial complexity. For other purposes, however, that complexity can help to explain how certain computationally-demanding processes, such as pattern recognition, are implemented in the brain. As Singer explains, "If such temporal coding is added to the principle of population coding the number of different patterns or representations that can be generated by a given set of neurons increases substantially." (1994, p. 202).
      Just as adding a temporal dimension to the concept of localization can require a richer and more complicated understanding of the physical realizer, so it can also invite a richer range of functions, broadened so as to include those of significant temporal duration. Cases of the latter can be found in work that correlates stereotypical movements, complex postures, or behaviors, with sustained, electrical stimulation of motor areas. For example, a study by Graziano, et. al (2002), shows that in monkeys, electrical microstimulation of certain points in the right precentral gyrus for a duration of 500 ms evokes a complex, hand-to-mouth movement. Extending the duration of the stimulus results in extending the pattern of movement, and shortening the duration results in truncating the movement. A series of stimulations at systematically varied points in the precentral gyrus resulted in orderly variations in the movement patterns such that they were able to construct an orderly, topographic map of hand and arm postures in space. As the authors note, this study represents an intriguing elaboration of Penfield's more static, topographic map of the motor cortex (the well-known motor homunculus). Penfield's homunculus reflects the correlation of momentary muscle twitches with brief pulses of electrical stimulation of the cortex, and thus lacks the temporal interest that is present in the orderly mapping of postures. It is tempting to consider what other sorts of maps or localization schemes one might discover were one to focus on functions that intrinsically involve significant, temporal duration.

4. Localization as a Heuristic

In this section I will further support my case for a more replete notion of localization by pointing to methodological considerations. I also have a broader point to argue in this section; namely, that to focus on the merely spatial aspects of functional localization is to fundamentally misunderstand its nature and purpose. Localization is important because it helps to point us to the salient structural or anatomical features associated with a given function. As Bechtel and Richardson argue (1993), the methodological significance of localization is that it serves as a heuristic to help elucidate structure/function relationships in the brain, or in any mechanistic system that lends itself to elucidation via alternately localizing the functional components of system then further decomposing those components into subcomponents, and so forth. Localization is not the ultimate end of this explanatory process. As they explain:

Direct localization does not, however, provide an ultimate explanation, as it does little more than locate an underlying system within a complex system. Even if direct localization is successful, it tells us only what produces the effect, and not how it is produced. (p. 65)

In the scientific literature, also, one finds practical acknowledgment of the indirect way in which localization serves to advance mind-brain understanding. With respect to brain scanning technology, for example, Shepherd writes,

...these studies are at the highest levels of organization; the scans provide overall maps of the locations of brain activity in relation to behavior, but they give little insight into the neural mechanisms at lower levels that are responsible for generating the maps. (1994, p. 9)

Similarly, Kertesz comments on the value of using lesion studies, as follows:

Localizing lesions can and does, however, provide information about function, although this information is often indirect. Sophisticated lesion analysis, including remote metabolic and transsynaptic events integrated with deficit analysis based on knowledge of normal function and functional activation, goes a long way toward clarifying functional-structural relationships. (1994, p. 2)

Now, in order to further explain and support the heuristic role of localization more clearly, I will turn to an illustration.
      Suppose you are a rustic farmer of a few hundred years ago. You own a large tract of land with several different plots. For the moment, assume that your land receives fairly uniform exposure to the elements (sunshine, wind, rain, etc.) Over the years, experience has taught you which crops grow best in which locations. You know, for example, that some patches of land will support nearly any crop, but other patches will support almost nothing. As for the crops, some are hearty and will grow nearly anywhere, while others require more specialized soil and care. When you consider the planting patterns each spring, you 'localize' certain crops to certain areas, and this way of thinking about your property helps you to grow your crops more successfully.
      Then a series of natural disasters hits. Torrential spring rains wash out your first plantings. Next, the river overflows; a record flood inundates your property, carries away your topsoil and deposits silt and other material carried from upstream. The flood is followed by a drought, and the drought is followed by powerful dust storms. Assuming you have managed to endure all this, and you can still find all your old landmarks, what will have happened to your crop localization scheme? To realize that it might not bring the same results as it has in years past is to realize that it wasn't the 'location' as such that accounted for your results (again, assuming otherwise uniform exposure to the elements); it was the soil composition that made the difference. Now that the soil has shifted, you will have to begin working out a new localization scheme. There is no reason to suppose that the boundaries between any remaining good locations and bad will be neatly ordered, spatially isolated, narrowly circumscribed areas, though of course they might be. And if the relatively good growing areas you should happen to find aren't arranged into neat, spatially well-defined parcels of soil, but rather, are scattered in patches, here and there, this is no reason to conclude that the soil is everywhere equivalent, or that everything could grow everywhere.
      Of course, the rustic farmer wasn't quite so rustic as to be completely ignorant about different soil types. Some things could be gleaned from the grosser appearance, such as whether it was clay, sand, limestone, peat, loam, etc., and a farmer would know from experience which crops were suited to which kind of soil. But more direct and sophisticated assessments about Nitrogen content, mineral content, pH, etc., could not be performed. This situation is roughly analogous to the way in which our understanding of structure-function relationships in the brain have evolved; as noted above, localization was first applied to the grosser structural features, then progressively refined to include smaller, more micro-level features (Finger, 1994, p. 3). As with the rustic farmer, if the goal is to advance understanding of structure-function (soil vs. crop fertility) relationships, localization is a means to an end, not an end in itself. It tells us, in effect, "look here"; i.e., look at the physical features of this part of the brain for clues about how functions are actually carried out by this structure. This is not to say that the process is infallible, merely that it is useful to further discovery. Once the structural components of a salient region are understood, the localization heuristic that lead us to consider those structural features in the first place becomes something like a Wittgensteinian ladder. A deciphered localization is still important for other purposes though, such as in neurosurgery, clinical predictions, and other contexts. Also, it remains important to consider how a given region is connected with other functional regions, but again, the spatial location of the connecting pathways as such, is not as important as knowing what regions are connected, and whether or not the connections are bi-directional. As Kertesz notes, "It is obvious that complex cellular systems that interact must be interconnected, and this occurs most readily if cells are close together. Functional contiguity, however, may be even more important than geographic contiguity in the cortex" (1994, p.6).

5. Unfettered Localization: Gall's Blind Alley

I now will approach my claims about the nature of localization from another standpoint, by citing a historical progression from a failed system of localization to a theoretically productive one, and what I take to be the crucial difference between them. The example of a failed localization scheme in this case comes from Gall, whose significance is being revisited in recent philosophical literature (Fodor, 1983; Bechtel and Richardson, 1993; Radden, 1996; Mundale, 1998; Zawidzki and Bechtel, in press). For an example of more productive localization, I point to the work of Brodmann. Though I have previously commented on both Gall and Brodmann in other work (Mundale, 1998), my current conception of their role in localization theory has shifted somewhat, mostly with respect to Gall's contribution.
Gall is often taken to represent an early step in establishing structure-function correlations between the brain and psychological faculties. He regarded the brain as the organ of the mind. The skull, which he supposed to tightly conform to the brain, was merely an indirect reflection of the surface of the brain. This meant that a person's psychological propensities could be read from the brain indirectly by examining the localized bumps and depressions in the skull. Many commentators have emphasized Gall's materialistic approach to the mind, including Clarke and Jacyna, who write:

For him, character and intellect were simply the sum of the combined functions of his brain 'organs', so that character was the brain. Instead of subscribing to the usual explanations for mental processes, which attributed them to the usual extracorporeal forces or delegated them to various viscera of the body, Gall maintained that a 'plurality of organs' on the surface of the brain, each with its own specific moral or mental faculty, functioned together to create the intellect and the character. Mental processes had natural causes which could be identified and determined. (1987, p. 234)

It is tempting to think, therefore, that Gall's localizations were in some way connected to some underlying rationale about the peculiar structural features of the brain. But this was not the case.
Gall claims to have arrived at his localization scheme by observing regular correlations between psychological propensities and skull shape, in humans as well as animals. Though structural or anatomical considerations played no role in how he determined his functional localizations, this would not present a problem if, having arrived at those functional localizations, he had then supposed there to be something structurally or anatomically significant about each of the brain regions he delineated. In fact, the only structural feature that was of general interest was the physical size of a bump. The larger the bump, the larger the underlying brain area, and the larger the brain area, the greater the proclivity to manifest its associated function. Zawidzki and Bechtel (in press) make a generous allowance for Gall on this last point, noting that, although the importance of the size of a brain area has largely been discredited, it was replaced by a related assumption, now central to PET scan research, that increased activity in a region corresponds to an increase in metabolic processes (blood and oxygen flow).
Given Gall's renowned ability as an anatomist and dissectionist, it is particularly strange that such a gross and obvious feature as size should be the only structural consideration of significance. Zola-Morgan comments on this matter as follows:

On the one hand, as a result of his anatomical dissections, Gall made several important anatomical discoveries that stand today. On the other hand, Gall also attempted to develop a functional anatomy that served the purpose of his doctrine of localization. Gall's functional anatomy was not grounded in empirical analyses and bore no relationship to his very careful and thorough descriptive anatomy. For Gall, descriptive anatomy and functional anatomy were entirely separate from each other. He argued that it was rare for (descriptive) anatomy to lead to the elucidation of function... (p. 362)

Apart from other well-known difficulties with Gall's methods, another way to understand the failure of this theory is in terms of this structurally unfettered system of localization. Brain structure, except in the most shallow sense, bore no relation either to the functions Gall attributed to the mind, or to the manner in which he localized them. Though Gall clearly regarded the brain as the organ of the mind, and though he obviously helped to establish the concept of cortical localization as such, he effectively insulated the mind from any meaningful interaction with brain anatomy and physiology. He also effectively shut off any avenues of independent confirmation of his theory from future developments in brain mapping research.
As I have noted in previous work, however, one of the crucial contributions Gall did make was to help inspire Brodmann's notion of an organology of the cortex (Mundale, 1998), a contribution that Brodmann himself acknowledges (1909, 1994 transl., p. 250). The critical link that Brodmann provided was to ground functional differentiation in intelligible, structural features of the cortex, and the resulting functional differentiations then lent themselves to independent confirmation through other methods of localization. In short, he made the connection between function and structure empirically respectable. Also, with Brodmann's work, one is in a position to understand why location, as such, is not the issue, but rather the underlying, physical differences in the brain tissues that vary from one location to the next.
To put this in broader terms, one of the vital insights of Brodmann and other twentieth-century neuroscientists consists of a crucial, physiological application of the Aristotelian principle that the function of an object is constrained by its material composition and the arrangement of its material constituents. In other words, what something does depends, in part, on its physical structure. My toaster does not keep my perishables refrigerated, and my ears do not respond to olfactory stimuli. As Aristotle put it,

When the matter is one, different things may be produced owing to difference in the moving cause; e.g. from wood may be made both a chest and bed. But some different things must have their matter different; e.g. a saw could not be made of wood, nor is this in the power of the moving cause; for it could not make a saw of wool or of wood. (1044a 25-29)

Though wood is multifunctional, it is not omnifunctional. The material nature of the wood, though it does not determine what one is able to do with it, nonetheless constrains what one is able to do with it. Generally, the physical nature of any object constrains what functions it will be able to perform. Conversely, functional demands constrain the kinds of structures that will be able to carry them out. It's not just that form follows function, but function also follows form; they are reciprocally heuristic. This is particularly so in a biological context, in which the actual implementational possibilities are a much smaller subset of the logical possibilities one might imagine. For Brodmann, when cortical regions differed markedly in their cellular composition, density, distribution, etc., it was a good, though not infallible bet that these cytoarchitectonic variations would correspond to functional variations. In other words, regions thus differentiated were a reasonable place to focus further attention, and subsequent localization research methods of the twentieth century largely bore out the reliability as well as the productivity of his heuristic bet (Mundale, 1998).

6. Conclusion

A common theme between equipotentialists and some holists is the denial of a reciprocally heuristic relationship between structure and function. For that matter, the multiple realizability school of functionalism (e.g., Fodor, 1974, Putnam, 1975) also partakes of this view; hence, Putnam's claim that "We could be made of Swiss cheese and it wouldn't matter" (1975, p. 291). These approaches deny that there is anything interesting to learn about function by investigating its neurological implementation; they also deny that neurological implementation can contribute anything useful to our understanding of function (see also Shapiro, 2000, and Zawidzki and Bechtel, in press). Some versions of the complex systems approach also deny that there is anything but the most incidental relationship between structure from function. Taken to the extreme, however, it is hard to see how such a view manages to avoid some form of metaphysical mysticism. Granted, there is a legitimate issue about the degree of transparency between structure and function, but do the extremists here mean to assert that the relationship is one of practical impenetrability, or impenetrability in principle? If the latter, then the universe indeed becomes a place of intractable mystery (see also Bechtel and Richardson, 1993, p. 228).
One of my goals in this paper was to present a clear analysis of the sophisticated, multi-local conception of localization at work in modern and contemporary neuroscientific research. It was also my goal to show how such an analysis effectively shields functional localization from a certain 'straw man' line of attack; that is, an attack that tacitly assumes a narrow and outmoded conception of localization that is roughly akin to the sort associated with phrenology. Furthermore, to recognize the heuristic role of localization is to realize that it is not the simple, spatial delineation of functional areas as such that is at issue, but rather, that the areas so delineated provide further clues as to the structural or physiological implementation of the function of interest.
Though I have not addressed all criticisms of localization, I hope to have cleared away at least some of the conceptual difficulties, and to have provided an analysis of localization that could serve to advance further debate and discussion on this issue. My suspicion is that the most significant challenges for future localization research will not be technological ones, but taxonomic ones. These, however, have a way of plaguing everyone's house.

Acknowledgments

Thanks to Prof. John Bickle, for his helpful comments and editorial patience. Thanks also to my colleague, Prof. Don Jones, for his insightful conversations and kind suggestions. Finally, I am grateful to an anonymous scientific reviewer for urging me to consider new research that raises considerations about the temporal nature of localization, and to an anonymous philosophical reviewer whose comments helped to improve the work overall.

References

Aristotle, 1941: The Basic Works of Aristotle, edited and with an introduction by Richard McKeon. Random House, New York.

Bechtel, W., 2001: The compatibility of complex systems and reduction: a case analysis of memory research, Minds and Machines 11, 483-502.

.-- , 2002: Decomposing the mind-brain: a long-term pursuit, Brain and Mind 3, 229-242.

.-- , and Richardson, R., 1993: Discovering Complexity: Decomposition and Localization as Strategies in Scientific Research, Princeton Univ. Press, Princeton, NJ.

Bickle, J., 2001: Understanding neural complexity: a role for reduction, Minds and Machines 11, 467-481.

Brodmann, K., 1909: Vergleichende Lokalisationslehre der Grosshirnrinde in ihren Prinzipien dargestellt auf Grund des Zellenbaues. Barth, Leipzig. Translated into English and edited by L. J. Garey, 1994: Brodmann's 'Localisation in the Cerebral Cortex', Smith-Gordon, London.

Clarke, E. and Jacyna, L. S., 1987: Nineteenth-Century Origins of Neuroscientific Concepts, University of California Press, Berkeley.

Finger, S., 1994: The Origins of Neuroscience, Oxford University Press, New York.

Fodor, J., 1974: Special sciences, or: disunity of science as a working hypothesis, Synthese, 28, 97-115.

.-- , 1983: The Modularity of Mind: An Essay on Faculty Psychology, MIT Press, Cambridge, MA.

Goltz, F. , 1888: Über die Verrichtungen des Grosshirns, Pflüger's Archiv für die gesamte Physiologie 42, 419-467, In G. von Bonin (ed., transl.), 1960: Some Papers on The Cerebral Cortex, as, On the functions of the hemispheres, Charles C. Thomas, Springfield, IL, pp. 118-158.

Graziano, M. S. A., Taylor, S. R., and Moore, T., 2002: Complex movements evoked by microstimulation of precentral cortex. Neuron 34, 841-851.

Kandel, E., Schwartz, J. and Jessell, T., 2001: Principles of Neural Science, 4th ed., McGraw-Hill, New York.

Kertesz, A. (ed.), 1994: Localization and Neuroimaging in Neuropsychology, Harcourt Brace/Academic Press, San Diego, CA.

Lashley, K., 1933: Integrative functions of the cerebral cortex, Physiological Reviews 13, 1-42.

.-- , and Clark, G., 1946: The cytoarchitecture of the cerebral cortex of Ateles: A critical examination of architectonic studies, Journal of Comparative Neurology 85, 223-305.

Lloyd, D., 2000: Terra cognita: From functional neuroimaging to the map of the mind, Brain and Mind1, 93-116.

Marshall, J. C., 2001: Bumps on the brain, Nature 414, 151-2.

Mundale, J., 1998: Brain mapping, in W. Bechtel and G. Graham (eds.), A Companion to Cognitive Science, Blackwell, Malden, MA, pp. 129-139.

vanOrden, G. and Paap, K., 1997: Functional neuroimages fail to discover pieces of mind in the parts of the brain, Philosophy of Science 64 (Proceedings), S85-S94.

Papez, J. W., 1953: Theodor Meynert, in W. Haymaker (ed.), The Founders of Neurology, Charles C. Thomas, Springfield, IL, 64-67.

Petersen, S. E. and Fiez, J., 1993: The processing of single words studied with positron emission tomography, Annual Review of Neuroscience 16, 509-30.

Putnam, H., 1975: Mind, Language, and Reality: Philosophical Papers, Vol. II, Cambridge University Press, New York.

Radden, J., 1996: Lumps and bumps: Kantian faculty psychology, phrenology, and twentieth-century psychiatric classification, Philosophy, Psychiatry, & Psychology 3.1, 1-14.

Raichle, M. E., 2001: Functional neuroimaging: A historical and physiological perspective, in R. Cabeza and A. Kingstone (eds.), Handbook of Functional Neuroimaging of Cognition, MIT Press/Bradford, Cambridge, MA , pp. 3-26.

Shallice, T., 1988: From Neuropsychology to Mental Structure, Cambridge University Press, New York.

Shapiro, L., 2000: Multiple realizations, The Journal of Philosophy, 97, 635-654.

Shepherd, G. M., 1994: Neurobiology, 3rd ed., Oxford University Press, New York.

Singer, W., 1994: Putative functions of temporal correlations in neocortical processing, in C. Koch and J. L. Davis (eds.), Large-Scale Neuronal Theories of the Brain, MIT Press/Bradford, Cambridge, MA, pp. 201-237.

Uttal, W., 2001: The New Phrenology: The Limits of Localizing Cognitive Processes in the Brain, MIT Press/Bradford, Cambridge, MA.

Zawidzki, T. and Bechtel, W. (in press): Fall's legacy revisited: decomposition and localization in cognitive neuroscience, in C. E. Erneling and D. M. Johnson (eds.), Mind as a Scientific Object: Between Brain and Culture, Oxford University Press, Oxford.

Zola-Morgan, S., 1995: Localization of brain function: The legacy of Franz Joseph Gall (1758-1828), Annual Review of Neuroscience 18, 359-83.