Teleodynamics

The ‘language-readiness’ of human brains most probably resulted from modification of structures present in non-human primate brains, but identifying such homologues and the nature of their modifications has been highly problematic. In a recent article, Arbiband Bota suggest that these problems can be overcome using a neuroinformatics approach. But its assumptions ignore many non-local, activity-dependent, regressive, and allometric effects of neurodevelopment that violate assumptions of classic homology. What if these effects are what matter most?

Download PDF of full article

Explaining the neurological basis for language ultimately hinges on determining how human brains differ from nonhuman primate brains. In their article reviewing the debates over the location, function and primate homologues of Broca’s language area, Michael Arbiband Mihail Bota argue that analyzing homological correspondences between human and non-human primate brains, using a neuroinformatics database, can resolve many longstanding questions about the ‘language-readiness’ of human brains. The concept of anatomical correspondence via common inherited information is not problematic in itself, but establishing homologies a posteriori in living species, on the basis of anatomical similarities, poses several difficulties, especially where nervous tissues are involved. Many confounding factors can contribute to superficial similarities or obfuscate structural resemblances between brains, irrespective of phylogeny, limiting the precision of homological assessment. This highlights the need for a computational approach able to compensate for diversity of criteria, methodology and nomenclature by exploiting intrinsic redundancies implicit in the data.

The homology puzzle
Although homology is often discussed as a comparison of adult-to-adult body structures, the anatomical similarities and topologies that constitute homologous relationships emerge from patterns of gene expression and cell interactions during development. In most regions of the body, structural differentiation is largely a reflection of spatially proximate cell-cell interactions, but in the brain, cellular migration, axonal growth, and activity-dependent culling of neurons and connections superimpose many nonlocal influences on this basic logic. So many standard criteria for homology assessment cannot be relied upon to be concordant with one another in brain structure comparisons.

The problem of non-local influences on tissue differentiation is most vividly exemplified with respect to axonal connectivity. The growth of axons from one brain region to another non-contiguous region creates a context for discrete cell-cell interactions at a distance, including self organizing, regressive (apoptosis and axonal elimination) and activity-dependent modifications of cytoarchitecture and function. Even phylogenetically unprecedented connection topographies can emerge de novo as a result of afferent changes that alter competitive interactions between axonal projections. These are not mere laboratory curiosities. Corresponding patterns of naturally evolved functional-connectional shifts have been identified in species with specializations that are analogous to some of these manipulations and differential expression of regressive modifications also might correlate with size. Indeed, many non-local interaction effects are sensitive to size (i.e. allometry; see below). Despite decades of intensive investigation of the importance of these secondary developmental modifications, however, only a few researchers have seriously considered their role in brain evolution and their consequences for homology. Considering that the major motivation for adopting this informatics approach is to aid analysis of a quantitatively deviant brain exhibiting a non sequitur functional specialization (language), departures from the standard primate brain ‘bauplan’ is the kind of nonhomology we might expect to discover. Although not explicitly encoded in the NeuroHomology DataBase (NHDB), non-local developmental effects should appear as non-concordant relationships between hodology (connection patterns) and location.

Avoiding the ‘fallacy of simple location’

The classic descriptions of language functions attributed to Broca’s and Wernicke’s areas, respectively, have long been subjects of debate, fueled by uncertainties of physiology and localization. Homological comparisons to monkey cortex offer a way to deconstruct function and structure of these areas, taking advantage of greater laboratory access to physiological information in monkeys. Arbiband Bota focus on two theories of language function for which the homological mapping of Broca’s area to monkey homologues plays a central role. Their favored theory - based on the discovery of ‘mirror’ neurons in the macaque ventral premotor cortex (area F5) - proposes that this specific region and the functions ascribed to these neurons constitutes the homologue to Broca’s area (which they locate to Brodmann’s area 44).  A competing theory, proposed by Aboitiz and Garcia, argues instead that the functions attributed to Broca’s area are more distributed between multiple frontal motor and prefrontal areas. Contrasting these theories on anatomical homology criteria, to the exclusion of other forms of data (e.g. aphasiology and imaging), inadvertently shows how important these other kinds of data are to assessing homology.

The identification of a region of monkey brain that is homologous to Broca’s area is limited by any ambiguities of its structural and functional delineation in humans. The determination of the substrate for Broca’s aphasia and the corresponding language function has been controversial throughout the 20th century. Despite common belief, considerable neuropsychological evidence suggests that damage confined to BA44 is not the sole and probably not the primary cause of the aphasic syndrome that bears Broca’s name.  Many adjacent regions are currently thought to be involved in the symptomatology of Broca’s aphasia, including cortical areas 45, 47, the motor face strip, and the anterior insula, as well as deep white-matter tracts and parts of the basal ganglia. Even the preserved brain of Broca’s first celebrated patient was recently reanalyzed by CATscan andMRI, and was found to include damage to all these structures and more. This heterogeneity of frontal language processing is consistent with tracer studies in non-human primate brains that show auditoryprojections terminating anterior to premotor cortex in prefrontal areas, and parietal and multimodal projections terminating in posterior premotor areas. It is also consistent with electrophysiological studies demonstrating that ventral prefrontal areas contain auditory responsive cells whereas ventral premotor cortex contains the visuo-manually responsive ‘mirror’ cells. Electrical stimulation, aphasia studies, and in vivo imagery of language processing tasks in humans have also consistently shown that left ventral prefrontal areas are linked with mnemonic and auditory components of language tasks and that ventral premotor, motor, and insular areas of cortex are linked with motoric components of language tasks. These considerations suggest that a NHDB that included these multiple sources of data would be as likely to deconstruct the very conception of Broca’sarea, as to delineate a simple structure-by-structure homology in the macaque brain.

The whole point of these homology assessments is to distinguish between language recruitments (antecedent systems recruited for this unprecedented function) and language adaptations (deviations from antecedent forms reflecting adaptation). The importance of maintaining this distinction is well exemplified by the discussion of the ‘mirror system’ hypothesis for language. Evidence from macaque brains suggested a homology linkage to frontal language functions via the following chain of inferences: (1) mirror neurons exhibit what might be called multi-perspectival responsivity to multi-modal (primarily visuo-manual) stimuli; (2) both imitation of speech and reciprocally understood referential communication involve perspective shifting, which is superficially analogous to a ‘mirror’ response; (3) these cells are located within the macaque frontal cortex in a premotor region (area F5) that is considered to be the cytoarchitectonic homologue to BA44 in the human brain; and (4) many researchers have suggested that BA44 is the anatomical substrate of the functions associated with Broca’s aphasia. On the basis of these inferential steps Rizzolatti and Arbib hypothesized that mirror neurons probably provided an important ‘pre-adaptive’ substrate that was eventually recruited for language function in human evolution. But mirror neurons do not provide macaques and other non-human primates with language abilities, and (as noted above) it is not at all clear that lesions destroying BA44, the putative mirror-neuron-containing region, in human patients results in persistent language impairment. Mirror-neuron functions might have been recruited to aid language-readiness, but we should not necessarily expect them to be the locus of a distinctive language adaptation. To explain that we must look for something different.

Scaled-up homologies
One final issue of homology should have been more cautiously explored and more explicitly implemented in the NHDB. Because of the disproportionate size of the human brain, any investigation of its evolution must address questions of allometry (i.e. differential scaling of structures or functions correlated with differences in size). To ignore this misses the most widely investigated feature of human brains, and one that is crucial to issues of establishing homology. Allometric analyses identify the fraction of quantitative change in a given structure that is ‘expected’ on the basis of overall size increase and what fraction deviates from this expectation. This can involve cytoarchitectonic features as well as gross morphology. Affected cytoarchitectonic features include relative neuronal size, cell morphology, neural/glial ratio, lamination, myelination, ratio of gray matter to white matter, and ratio of koniocortex (‘primary’) to eulaminate cortex (‘association’). All these vary systematically with respect to absolute brain size, and largely irrespective of common ancestry, and can also be highly divergent in closely related species that differ significantly in size. The comparatively large gap in human versus monkey brain size inevitably involves architectonic deviations that are crucial to establishing homologies, particularly increased differentiation and subdivision of cortical areas. Considering the importance of this factor it is curious that Arbiband Bota reiterate a simple but misleading comparison between a large heterogeneous region and one of its subregions in non-human and human brains.

To support their focus on human BA44 and macaque area F5, they cite an MRI-based study showing that ‘frontal’ cortex in humans is not allometrically deviant in size from that in other apes.  They agree with its authors that these data contradict previous studies showing that human prefrontal cortex is disproportionately large .  However, this interpretation involves a clear violation of one of the criteria that the NHDB is intended to control for: comparisons that equate hierarchically nested homologues. Differences in methodology contributed to this error. MRI data have insufficient resolution for cytoarchitectonic comparison, but cytoarchitecture is necessary for determining the boundaries of prefrontal cortex. The MRI analysis relied on a gross morphological feature - the central sulcus - to distinguish ‘frontal’ from non-frontal cortex, lumping together prefrontal, orbital, premotor, motor, cingulate and anterior insular cortex (plus some other structures). But ‘frontal’ cortex delimited this way is hierarchically much more inclusive than prefrontal cortex alone, and the allometric predictability of the whole sector cannot be assumed to be inherited by this one part. Although Arbib and Bota missed catching this dishomology, or the misunderstanding it suggests, the NHDB should be specifically designed to avoid such errors, even if glossed over inthe original research reports.

Regimenting the comparative use of the many types of homology criteria is an essential step in explaining the basis for our language-readiness. Arbiband Bota could do the field a great service by providing a more sophisticated informatics approach to homology criteria, but it will not merely be a matter of data collection and tabulation. Much work remains to be done before neural homologies are truly unambiguous.

Response to Deacon: Evolving mirror systems: homologies and the nature of neuroinformatics

Michael A. Arbib1,2 and Mihail Bota2

1Department of Computer Science, University of Southern California, Los Angeles, CA 90089-2520, USA
2Department of Neuroscience, University of Southern California, Los Angeles, CA 90089-2520, USA

We are delighted that our article [1] has received a thoughtful response from Terry Deacon [2]. However, his discussion of homologies ignores key elements of our views on the evolution of brain mechanisms supporting language and does not fully address the details of our approach to neuroinformatics, the NeuroHomology Database (NHDB).

Evolving the language-ready brain
Deacon [2] asserts that ‘Mirror neuron functions may have been recruited to aid language readiness, but we should not necessarily expect them to be the locus of a distinctive language adaptation. To explain that we must look for something else that is different’, as if this were a critique of our approach. However, we said explicitly that Arbib [3] has amplified the approach of Rizzolatti and Arbib [4] to hypothesize seven evolutionary stages.