The quest to find the physiological substrate of a thought has been elusive. What should we be looking for?
American psychologist Karl Lashley spent much of his life in a quest to unravel the relationship between brain function and intelligent behavior, searching for the ‘engram’ or locus of thought. In the 1920s and 30s he conducted a set of experiments where he meticulously destroyed different parts of the cerebral cortex of rats, allowed them to recover from the surgery, and then tested their ability to learn and remember certain tasks such as running a maze or distinguishing between two patterns. The degree of impairment depended only on how much of the cortex was destroyed—and not which part. It was not until he had removed at least half the cortex that learning and memory became severely impaired. In light of these findings Lashley concluded that learning and memory were distributed across the cortex and not localized in any one place—that every part of the cortex was equal in its capability, or equipotential.
Inspired by these findings, one of his students, Donald Hebb, went on to study the impairment resulting from brain damage and surgery in humans. Again, the results were astonishing. Clearcut removal of parts of the cortex outside the speech area often had little, if any, detectable effect. A man who had a prefrontal lobe removed continued to score an extraordinary 160 or higher score on IQ tests. A woman who had lost the entire right half of the cortex continued to have an IQ of 115, better than two-thirds of the normal population (See related article The Curious Outcomes of Neurosurgery). Hebb conceded that these were the striking cases, but it begged the question: how was it possible that while large brain injury often had severe impact on intelligence, sometimes it did not. Once a concept had been learned, Hebb noted, it was not easily lost by brain damage.
Yet, during these same decades, a rather different set of studies was taking shape, carving up the cortex into localized ‘functional’ areas. As early as 1907, German neurologist Korbinian Brodmann had divided up the cortex into ‘functional areas’ based on anatomi- cal connectivity between the sensory apparatus and the cortex. Brodmann went as far as to propose that these areas of the cortex represented different functional ‘organs’. By the 1940s the localization theory had gained further support. Largely consistent with the anatomical maps of Brodmann, researchers had constructed physiological maps. Application of pressure to different parts of the body elicited electrical potentials on distinct areas of the cortex. Every part of the body mapped to a distinct region of the cortex. Neurosurgeon Wilder Penfield drew the most well known of these maps in 1950 called the homunculus – mapping the movement and sensations of the human body to different regions of the brain.
Over the ensuing decades, the theory of localization found its way in the greater psyche of the scientific community. Neural pathways from the retina arrived in a particular region of the cortex, which responded robustly with electrical potentials to visual cues. Neural pathways from the ear arrived at a particular region of the cortex, which responded to sound. Conversely, electrical stimulation of those regions of the cortex resulted in corresponding movement or sensation. Such functional maps showing a visual cortex, auditory cortex, motor cortex and so on, are now printed in every basic neuroscience textbook.
Reconciling distributed and localized function
How is it possible that sensorimotor function maps so clearly on to particular regions of the cortex while higher function such as learning and memory and intelligence that make use of these functions are distributed? How do the various modes of sensory input come together to create the integrated perceptual stream we experience? Almost a century after Lashley, these quandaries remain unresolved.
One of the most fundamental assumptions of neuroscience research has been that the physical substrate of all thought and behavior is the electrical activity transmitted among neurons or neural cells. In these terms, the question is whether any set of neurons can create the behavioral or perceptual response to a task at hand, or whether it must it be a specific set? In this framework the task is to understand how neurons connect with one another to form a complex circuitry of electrical communication. What does each neuron ‘know’? Which neurons talks to which other ones? How do they share information? For many decades neuroscientists have been painstakingly studying this.
The behavior of neurons
Individual neurons have now been probed in numerous ways. In a pioneering study in 1959 Torsten Wiesel and David Hubel found that neurons can be remarkably specific in what they care about. Measuring electrical activity from individual brain cells in the visual cortex of an anesthetized cat, they found that some fired vigorously when a bar was moved in front of the retina in one direction but not any of the others. In the ensuing decades a host of similar studies followed that explored neurons in various parts of the cortex. Some responded to specific colors but not others. Some responded to certain sound frequencies but not others. But not all neurons were so narrow in their worldview. In 2005 Rodrigo Quiroga and colleagues found neurons that responded vociferously to pictures of Jennifer Aniston, but not to pictures of Brad Pitt or pictures of toilet brushes, suggesting that neurons are capable of larger concepts, recognizing complete images and not just select visual features. Still, the collective results seem to suggest that individual neurons hold highly specialized or selective knowledge, although of varying complexity. If knowledge is specialized among neurons and not generalized, how can that be reconciled with Lashley’s findings on cortical damage? Surely damage of these locally specialized cells should completely destroy any abilities that draw from their knowledge. But extrapolating from the work of Hebb and Lashley, damage to those Jennifer Aniston cells in the visual cortex is not likely to get her off your mind.
Cell assemblies as constructs of memories and thoughts
This brings us back to the question of how neurons might share specialized knowledge to create an integrated perceptual whole. Hebb himself put forward one of the most influential hypotheses in this regard. Drawing from the findings of Lorente de No, a French physiologist who had identified closed loops of connected neurons that sometimes spanned multiple ‘functional’ areas of the cortex, he postulated that an individual experience, thought or memory manifests as electrical activity transiently reverberating among a subset of neurons in a closed loop, a ‘cell assembly’. The important aspect of this theory is that it suggests that what counts is not just what neurons hear and get excited about in the external world but how they share what they hear with one another, which neuron shares with which other, and what memory this sharing creates. With this framework in mind, many sorts of simulated models have been proposed, some that are even significant departures from Hebb’s own reverberating loops. However, empirical evidence of such ‘cell assemblies’ or persistent electrical activity among neurons has been elusive.
Towards more integrated and measurable views
A change in method or technique accounts for many things in human history.
― Wilder Penfield,
Empirically measuring and identifying activity of many neurons simultaneously is a technically difficult task on many dimensions. There is also the challenge that presently it is only possible to perform such measurement in rodents and monkeys and not humans. Given that the human brain may function fundamentally differently, perhaps with profoundly different contributions to overall network activity from glial cells (see From Mouse Brain to Human Brain), any findings in animal models may very likely fail to translate to humans. The aggregate activity of the human brain is likely an orchestration of activity among neurons and glial cells that is significantly distinct from other species.
That said, in a broader sense, Hebb’s theory of cell assemblies can be thought of as driving towards identifying structures of electrical activity among neurons that can persist in the network on the timescale of perception, typically a tenth of a second, without dying out. If he is right, the big question is, can we find evidence of such structures in the more aggregate views of brain activity such as EEG, which can be easily related to behavior and cognition in humans? Moving from simplistic spectral analysis and models (see The Blue Frog in the EEG) to identify such features could vastly expand the potential of EEG as a tool of insight into the human brain.