The brains of autistic individuals

Manuel F. Casanova

This was originally provided as a handout to accompany my presentation “Abnormalities of cortical circuitry in the brains of autistic individuals” at the Autism One conference in May, 2006.

The hardest thing to do is to live immersed in the present. Even while tending to our chores we always try to tip toe into the future. People with autism tend to live unceremoniously in the present. Although fixed in the moment, their language does not partake in the industriousness of give-and-take conversation. The end result is a language that is all “muscular,” lacking the “twang” of the vernacular. People misconstrue their communication attempts as bordering on rashness; as being too rough. Moreover, superficiality for autistic persons is a pejorative term. They find it difficult to take cues from people talking to them. In a world made routine by verbal constructs, their way of thinking is judged to be unfamiliar. For they experience the world with heightened sensations and depict the world at an angle. This is the pathos of the condition, because a different way of thinking does not copy, it rivals.


The view that specific mental processes are correlated with discrete regions of the brain has been the object of countless empirical studies. These studies suggest that many of the faculties that define human behavior, providing variations among individuals and species, are located within the outer covering of the brain. This structure, also called neocortex, is formed early on during development by the supernumerary aggregation of modules. The smallest module capable of processing information is called a minicolumn. These modules or minicolumns are composed of both cells (neurons) and their projections. Together, minicolumns provide for standardized circuits repeated some 600 million times throughout the whole extent of the neocortex. The analogy to these minicolumns is that of a microprocessor (CPU) in a computer. Small changes in the basic template of minicolumns, repeated throughout the extent of the neocortex, provide for major behavioral manifestations. In effect the aggregation of supernumerary minicolumns is suggested to have given rise to our species (Homo sapiens) and to some of its unique attributes, e.g., language.

Recent studies suggest that minicolumns are smaller and more numerous in autism. Furthermore, cells (neurons) within each minicolumn of autistic individuals are reduced in size. Since the metabolic efficiency of neuronal connectivity is a function of cell size, the presence of smaller neurons in the brains of autistic patients has a dramatic effect on the way that different parts of the brain interact with each other. Functions that require longer projections (e.g., language) may be impaired while shorter ones (e.g., mathematical manipulations) may be preserved or reinforced. In terms of size minicolumns in autistic individuals fit into the tail end of a continuous distribution with the general population. Computer modeling has shown that smaller minicolumns may confer a given individuals with a different way of thinking (processing information) one that emphasizes particulars. By themselves these findings do not necessarily represent an abnormality. However, minicolumns are insulated from adjacent units by an inhibitory surround. In some autistic patients this inhibitory surround is defective and thus may provide for a leakage of stimuli to adjacent modules. The end result is an amplification of signals which has a deleterious effect on sensory integration. Knowing what is wrong with the brains of some autistic patients provides for putative interventions for those individuals who are handicapped by problems of sensory integration.


The first few paragraphs of this introduction relate our findings in autism. If you feel lost regarding some of the terminology used, do not despair for you are in good company. The next few sections are aimed at dispelling any doubts that you may have. For those that already have some knowledge in terms of neuroscience, you may want to skip to the section on autism and minicolumns after reading this introduction.

The brains of autistic patients have “minicolumns” that are smaller and more numerous than normal. Furthermore, brain cells (neurons) within each minicolumn are reduced in size. Since the efficiency of connections among neurons is a function of cell size, the presence of smaller neurons in the brains of autistic patients has a dramatic effect on the way that different parts of the brain work together. Brain activities that require longer projections (e.g., face recognition) may be impaired while those that depend on shorter connections (e.g., visual discrimination) may be preserved or reinforced.

The outer covering of the brain or neocortex is arranged as a hierarchy of interdependent modules. Studies suggest that many faculties that define cognition and behavior are located within the neocortex. The smallest module capable of processing information is called a minicolumn. These small modules or minicolumns are composed of both cells (neurons) and their projections. They were given the descriptive name minicolumns because of their microscopic proportions (mini) and rectilinear arrangement (columns).

The total number of minicolumns is defined during the first forty days of fetal development. This window of vulnerability coincides with reports of autistic behaviors in fetal/developmental conditions such as rubella babies, infants exposed to thalidomide, and tuberous sclerosis. In all of these conditions a defect is presumed to have arisen at an early stage of development, one that coincides with the formation of minicolumns.

Information is transmitted through the core of the minicolumn and is prevented from activating neighboring units by surrounding inhibitory fibers. In autism minicolumnar size reduction involves primarily the peripheral compartment that provides the inhibitory surround. This means that stimuli are no longer contained within specific minicolumns but rather overflow to adjacent minicolumns thus providing an amplifier effect. This may explain the hypersensitivity of some autistic patients as well as their seizures.

The following paragraphs provide the background information needed in order to understand how minicolumns behave in both the normal and abnormal state. I will start with a short section on brain development and then describe how different areas of the brain acquire particular functions. A subsequent section will describe how minicolumns are altered in autism. I will finalize with a short section on how these findings may be of use in providing for putative therapeutic interventions.

Brain development

The outer covering of the brain or neocortex is arranged as a hierarchy of interdependent modules [Figure 1]. These modules originate at a very early stage of development when newly formed cells near the core of the brain migrate towards its outer margin. Since the corresponding cells migrate along a radial guiding process, their ultimate arrangement is columnar [Figure 2], [Figure 3]. Because these cells develop together and are in close apposition most of their connections are bound together into what has been called a canonical or representative circuit. Connectivity of elements within these vertical arrays exceeds connectivity between them by several orders of magnitude. Unsurprisingly, voltage recording microelectrodes indicate that cells within these rectilinear arrangements share similar stimulus response properties. This means that all of the neurons within a particular minicolumn receive information from the same source. Also, stimulating one element of a minicolumn will provide for a rapid cascade of connections and the transmission of information among its component cells. These functional and anatomical characteristics define minicolumns as the smallest units or systems capable of information processing within the brain.

Minicolumns are analogous to microprocessors in modern day computers. In their function they represent the holistic properties of the brain as a whole: receiving stimuli from elsewhere in the body, processing information, and providing for some type of response. Whenever any of these functions is altered the whole minicolumn becomes dysfunctional. It is believed that Alzheimer’s disease is the result of a minicolumnar abnormality. Degenerative changes know as neurofibrillary tangles occur primarily in those cells that provide for the response or output of the minicolumn. By targeting and damaging these few cells, the whole minicolumn ceases to function. This helps explain both the rapid deterioration of mental faculties observed in this condition and why cognitive impairment seems out of proportion to the amount of cell loss.

A main advantage to the modular organization of the neocortex entails a reduced metabolic expenditure by limiting connectivity. Instead of connecting every single cell within the cortex to different brain regions in a haphazard fashion, projections are organized into modules. Single cells project to target sites and information gained or transmitted is transferred to other neurons within the same modules. Another advantage for dividing the cortex into modular arrangements is that of plasticity or the ability to recover from injury. The fact that minicolumns exhibit the same circuitry allows them to replace each other in case of injury. This implies a spectrum of injury severity where mild lesions cause no major loss of function, i.e., the prevailing minicolumns take over the function from missing minicolumns. As in logic circuitry, where a single type of logic gate (either NAND or NOR) suffices to construct networks for any Boolean function, so it has been argued that the circuitry of a minicolumn acts as a component having the same function regardless of its location in the cortex.This is a common theme in evolution in which complex variations come into being from tinkering with a limited reservoir of components. They take inputs and guide them through the same logic-level registration. The end result is an output that appears modulated by the source of the input, output target, and inhibitory influences. Our mental activities are therefore primarily dependent on connectivity, that is, on how we putt these logic gates together in different combinations. The view of canonical circuits having the same function regardless of cortical location was first espoused in the literature by O. D. Creutzfeldt [1].

V. Ramachandran [2] tells the story of a patient whose arm had been amputated above his elbow. When touching the patient’s cheek as part of the neurological examination, the patient felt a sensation on the thumb of his amputated, and otherwise “non-existent”, arm. Experimental evidence sustains that the representation of the face on the surface of the brain (neocortex) is near the hand. Ramachandran believes that, in this particular patient, the vacated area in the neocortex corresponding to the hand is invaded by sensory input from the adjacent facial skin. Still the information coming to the neocortex and its minicolumns has to be processed exactly the same way for this phantom limb phenomenon to work.

The above mentioned explanations should not be not be construed as saying that minicolumns are clone-like entities. Different brain regions exhibit a certain degree of minicolumnar variability, e.g. in their width and cell density. Comparison across species indicates that this variability may, in part, reflect the complexity of the information being processed. At any given age minicolumar variability is thus more prominent in humans than in other primates. However, regardless of any minicolumnar variability, all mammalian species have cortices that are arranged in terms of the same rectilinear arrangements. This has been the case for all brain regions thus far examined. Non-mammalian species, i.e. reptiles, lack this minicolumnar arrangement.

Variability among the multiple components of the minicolumn (e.g., number of neurons and amount of synapses) may contribute to the fault tolerance of larger networks such as macrocolumns. Redundant systems can be reliable even when the underlying components are error-prone. McCulloch [3] characterized unreliable networks of threshold elements as “logically stable” when elements’ thresholds could vary in tandem—not changing the function computed by the network as a whole—and “logically unstable” when thresholds could vary independently. He showed that, provided that individual components are more than 50 % reliable, redundant systems of unstable nets could be designed for greater reliability than redundant systems of stable nets of the same size. Paradoxically though it first appears, the functional plasticity of minicolumns, their ability to compute different functions than other minicolumns within the same module, can be used to the module’s advantage with respect to sensitivity to error.

Thus far one study has indicated a trend toward significant differences when comparing the variability of minicolumns among brains of autistic patients and controls. One possibility is that the limited number of patients within that series decreased the power thus necessitating a larger outcome measure (i.e., variability) than available in the sample. If lack of variability was proven in larger series it could help explain certain behavioral traits observed in autistic patients, e.g. the insistence on sameness, repetitive movements, and restricted range of interests.

Lateralization of brain function

It has been suggested that the reason the human brain has grown in relation to that of other species is the addition of supernumerary minicolumns. As compared to the mouse brain the human brain exhibits a thousand-fold increase in surface area but only a two- or threefold increase in cortical thickness. If bigger brains were the result of adding neurons serendipitously to the neocortex there would have been a corresponding increase in both surface area and thickness. Minicolumns therefore account for the large increase in brain size (as compared to overall body size) when comparing humans to other species.

With increasing brain size, areas of the cortex have acquired different functions. The small brain of the mouse can be parcellated into a dozen or so different brain regions. By comparison the human brain has anywhere from fifty to a couple of hundred different brain regions depending on the technique employed. Some of these areas seem to act independently from similar regions in the opposing hemisphere. This is known as lateralization of function or cerebral dominance. An example of cerebral dominance is that of language. In a majority of people language lateralizes towards the left hemisphere. Both autopsy studies and structural neuroimaging studies have shown that the area devoted to language processing is significantly bigger in the left hemisphere as compared to the right hemisphere.

A few years ago, I was interested in pursuing the reason(s) why different areas of the brain acquired cerebral dominance. More specifically, I was interested as to whether a particular area that exhibits cerebral dominance for language manifested changes at the microscopic level that could explain features seen at higher levels of resolution. Using computerized algorithms developed in my laboratory my research team found differences in the size and number of minicolumns that accounted for the lateralization of language. Furthermore, these differences were present in humans but not in closely related species, e.g., non-human primates. These findings attracted a lot of attention in the layman press. Since language is a uniquely human endeavor and the minicolumnar changes were specific to the language region, the finding was touted as a putative speciation event, i.e., a change that gave rise to the unique capacities of Homo sapiens. From this initial research, the question soon arose as to whether abnormalities in the lateralization of minicolumns could underlie human conditions characterized by language abnormalities, e.g., autism.

Autism and minicolumns

Autism is a medical condition whose core features include both language and brain size abnormalities. Autopsy, head circumference, and structural imaging studies all suggest that the brains of autistic individuals are, on average, larger than normal. Furthermore, a significant percentage of autistic patients have delayed and deviant language development. Some patients remain mute throughout their lifetime. Furthermore approximately one third of autistic individuals suffer from seizures, a phenomenon that usually localizes to the neocortex. The available facts from both clinical and structural imaging emphasized the need of studies focusing on the neocortex and its modular organization.

Our first study of autism involved nine patients and an equal number of controls. Three brain regions were examined, including a specific portion of the language regions of the brain. Striking disturbances in minicolumnar morphometry were found in autistic individuals. More specifically, minicolumns in the brains of autistic individuals were smaller in size and more numerous in all three areas examined [Figure 4]. These results were soon corroborated using a different technique called the gray level index. Furthermore the results were not a result of other superimposed conditions like mental retardation. Down syndrome patients have normal sized minicolumns despite having smaller brains. If anything, they acquire their mature size sooner than normal individuals. This has been seen, in the context of Down syndrome, as suggestive of accelerated aging.

Our findings, smaller minicolumns in the brains of autistic individuals, have been replicated as part of an international study in collaboration with Dr. Christoph Schmitz of the University of Maastricht, the Netherlands. This new and independent sample examined the brains of six autistic patients and an equal number of controls. All facets of the study, including the identification of brain regions, analysis of minicolumns, were done blind to diagnosis. In autism, the presence of smaller minicolumns in brains that are larger than normal suggests their total increase in numbers. In essence, because minicolumns are smaller, you need a larger number of them to occupy a given area of the cortical surface.

The total number of minicolumns is defined during the first forty days of fetal development. This window of vulnerability coincides with reports of autistic behaviors in fetal/developmental conditions such as rubella babies, infants exposed to thalidomide, and tuberous sclerosis. In all of these conditions a defect is presumed to have arisen at an early stage of development, one that coincides with the formation of minicolumns. Although genetic influences play a primordial role in defining the total number of minicolumns, environmental influences modulate many other aspects of minicolumnar morphology [Figure 5]. In a recent study following the maturation of minicolumns from early gestation to the ninth decade of life a continuous curve was plotted with a sharp inflection during the first two years of life. This is the time frame when minicolumns are most susceptible to environmental influences and coincides with the time window of opportunity for humans to learn language.

What is the meaning of smaller minicolumns? First, this question has been approached from the standpoint of computer modeling by Dr. Lennart Gustafsson’s group in Sweden. Results suggest that smaller minicolumns tweak information processing (noise/signal ratio) in favor of signal. By comparison other conditions characterized by larger minicolumns (e.g., dyslexia) tweak information processing in favor of noise. This means that autistic individuals usually do well in processing stimuli that requires discrimination while dyslexics are better at generalizing the salience of a particular stimulus. Second, minicolumns are compartmentalized. Information is transmitted through the core of the minicolumn and is prevented from suffusing into neighboring units by surrounding inhibitory fibers. The inhibitory fibers act in analogous fashion to a shower curtain. When working properly and fully draping the bathtub the shower curtain prevents water from spilling to the floor. In autism, minicolumnar size reduction involves primarily the peripheral compartment that provides the inhibitory surround. This means that stimuli are no longer contained within specific minicolumns. Stimuli overflow to adjacent minicolumns thus providing an amplifier effect. This may explain the hypersensitivity of some autistic patients as well as their seizures. Third, most environmental information is transmitted to the brain via a structure called the thalamus, a salient exception being olfactory information. The thalamus sends projections which span a finite distance within the cortex. It is believed, but not proven, that thalamic innervation is a way of binding minicolumns into larger modules called macrocolumns. If the terminal field of thalamic innervation remains the same size while minicolumns are smaller, the end result would be many more minicolumns per macrocolumns. Again, this change is similar to the previously described loss of peripheral inhibition “helping” make the brains of autistic individuals behave as an amplifier system. Finally, comparisons of minicolumnar parameters across many species suggests that smaller minicolumns (and more of them per given area) provide complexity in terms of information processing. A visual cortex constructed of smaller minicolumns may provide for added aspects of functionality, e.g., depth or color perception. Researchers believe that this complexity is due, in part, to the overlap of neuronal projections between different minicolumns. That is, dendrites and axons of neurons that remain the same size in autism may have more of an opportunity to overlap when their constituent minicolumns become smaller.

Minicolumnar size is not the only abnormality observed in the neocortex of autistic patients. It appears that cells (neurons) within individual minicolumns are also reduced in size. This has important consequences in terms of connectivity. Long connections require the metabolic sustenance of large cell bodies. Thus, a neuron in the brain that connects all the way to the lower spinal chord requires a fairly large cell body. By way of contrast, a neuron whose projection remains within the cortex, contacting a closely adjacent cell, can manage its metabolic demands with a small cell body.

The small cell bodies in the brains of autistic patients favor information processing through short intra regional pathways, e.g., mathematical calculations, visual processing. Similarly, cognitive functions that require long inter regional connections would prove metabolically inefficient, e.g., language, face recognition, joint attention. In this regard, the less affected region of the brain should be the visual cortex which is composed of a majority of small granule cells having short connections between closely adjacent areas. It is no wonder that in terms of information processing autistic patients use coping strategies in day to day activities that play to their strength. This is one of the main themes addressed by Temple Grandin [4].

Potential therapy

The following paragraphs relate a potential therapeutic intervention in autism. This intervention is not meant to change the number of minicolumns or the size of the same. Rather, our use of transcranial magnetic stimulation is directed at strengthening the inhibitory surround of the cell minicolumn in those patients that are handicapped by sensory integration defects. In this regard therapy would not change a person’s way of thinking. The therapeutic intervention is aimed at specific symptoms which have proven disabling in some patients.

Repetitive transcranial magnetic stimulation (rTMS) offers a noninvasive method for altering excitability of the brain. It potentially induces a short-term functional reorganization in the human cortex. The magnitude and the direction of rTMS-induced plasticity depend on the variables of stimulation (intensity, frequency, number of stimuli) and the functional state of the cortex targeted by rTMS. Since effects of rTMS are not limited to the stimulated target cortex but give rise to functional changes in anatomically and functionally interconnected cortical areas, rTMS is a suitable tool to investigate neural plasticity within a distributed functional network. The lasting effects of rTMS offer new possibilities to study dynamic aspects of the pathophysiology of a variety of diseases and may have therapeutic potential in some psychiatric disorders.

Autistic disorder is a developmental disorder; the symptoms of autism develop at an early age and often do not seem to improve overtime. Therefore, an early intervention is needed to achieve maximum therapeutic effect. Autistic disorder is associated with cortical minicolumnar abnormalities. In brief, the reduced neuropil space (periphery of the minicolumn) reported in autism is the compartment where lateral inhibition sharpens the borders of minicolumns and increases their definition. The primary source of for this inhibitory effect may be derived from axon bundles of double-bouquet cells. The axons of double bouquet cells arrange themselves in essentially repeatable patterns varying between 15 μm and 30 μm wide, depending on the cortical area examined [Figure 6]. Increases in numbers and types of inhibitory interneurons, as seen in the smaller minicolumns of autistic patients, result in greater diversity and more nuanced modulation of minicolumns. This complexity also creates potential for neuropathology. Double-bouquet cells in the peripheral neuropil space of minicolumns provide a “vertical stream of negative inhibition” surrounding the minicolumnar core. Other inhibitory cells in the minicolumn, having collateral projections extending hundreds of microns tangentially, provide lateral inhibition of surrounding minicolumns on a larger scale. The profile of excitatory activity arising from a minicolumn, or cluster of minicolumns, may be represented graphically by a cubic spline, or 3D bell-shaped volume, with the surrounding zone of lateral inhibition generated by collateral projections represented by another superimposed spline with a lower maximum and greater range. Subtracting the latter from the former yields a volume with a ‘Mexican hat’ topology [Figure 7]. This volume represents the ‘net’ excitatory or inhibitory activity resulting from the interaction of excitatory and inhibitory fields. The resulting center-surround configuration may function in the manner of an edge detector, filtering extraneous information and noise to generate an enhanced-contrast boundary between competing macrocolumnar fields. Further, this may provide the basis for overlapping fields of lateral inhibition which interact in a combinatorial manner to influence the excitatory output of each minicolumn in the network.

The value of each minicolumn’s output is insulated to a greater or lesser degree from the activity of its neighbors by inhibition in its peripheral neuropil space (see double bouquet cells above). This allows for gradations in amplitude of excitatory activity across a minicolumnar field. It has been posited that reductions in GABAergic inhibitory activity may explain some symptomatology of autism, including increased incidence of seizures and auditory-tactile hypersensitivity. This hypothesis is consistent with findings of reduced minicolumnar peripheral neuropil space in the neocortex of autistics relative to controls. In this model, reduced peripheral neuropil would result in smaller minicolumns which would coalesce into discrete, isolated islands of coordinated excitatory activity. These islands could serve as potential ictal foci. Moreover, their autonomous activity would hinder the binding of associated cortical areas, arguably promoting focus on particulars as opposed to general features. If cortical systems are in noisy, hyper-excitable conditions in autism, then, according one can postulate which genes could contribute to that physiological state. GABA is the most prevalent inhibitory neurotransmitter. As such, mutations or environmental factors that decrease GABA signaling would increase the brain's excitatory tone. Decreases in GABA production and signaling are known to contribute to hyper-excitable states (as in epilepsy) and cognitive dysfunction. GABA receptors are implicated in autism based on human genetic studies. The GABA imbalance hypothesis is similarly supported by findings of altered plasma and platelet GABA levels, and mutations in the region of a GABA-receptor.

Significantly, by puberty, one third of autistic patients will have exhibited at least two unprovoked seizures. Anecdotal case reports have shown that anticonvulsants (medications that increase GABA levels) have ameliorated autistic traits in epileptic patients. More recently an open trial of divalproex sodium (an anticonvulsant) in autism spectrum disorders showed that patients sustained improvement in core symptoms of autism and associated features: affective instability, impulsivity, and aggression. Anticonvulsants may be of some benefit on autism but at larger doses suffer from serious side effects including stupor and coma. These side effects are due to the non-selective nature of anticonvulsants whose mechanism of action (increasing GABAergic tone) is independent of cell type (e.g., double-bouquet, small and large basket, chandelier). The effects of anticonvulsants stand in contrast to the specificity of slow rTMS. This technique induces electricity in conductors at right angles to an expanding or collapsing magnetic field (law of electromagnetic induction). This effect may be of benefit when selectively attempting to activate the inhibitory cells and fibers surrounding the minicolumn (peripheral neuropil space). These anatomical elements have as a geometric preference being perpendicular to the cortical surface. In addition, a recent study, examined the changes in high frequency oscillations (HFOs) of somatosensory evoked potentials (SEPs) before and after slow rTMS over the right primary somatosensory cortex (0.5 Hz, 50 pulses, 80 % motor threshold intensity). The HFOs, which represent a localized activity of intracortical inhibitory interneurons, were significantly increased after slow rTMS, while the SEPs were not changed. Their results suggest that slow rTMS affects cortical excitability by modulating the activity of the intracortical inhibitory interneurons beyond the time of the stimulation and that rTMS may have therapeutic effects on such disorders. Our proposal/hypothesis [Figure 8] is therefore one where slow rTMS will increase activity of inhibitory cells in minicolumn which will then enhance spatial contrast needed to enhance functional discrimination in patients with autism.


Otto D. Creutzfeldt. Generality of the functional structure of the neocortex. Naturwissenschaften 64 (October, 1977) 507–517.

V. S. Ramachandran. A brief tour of human consciousness: From impostor poodles to purple numbers. New York: Pi Press, 2004.

W. S. McCulloch. Agatha Tyche: of nervous nets— the lucky reckoners. In: National Physical Laboratory. Mechanisation of thought processes. London: H. M. Stationery Office, 1959. p 611–625.

Temple Grandin. Thinking in pictures: And other reports from my life with autism. New York: Vintage, 1996.

Valid XHTML 1.1

Contact persons for this Web site are Manuel F. Casanova (principal investigator) and Andrew E. Switala (administrator).