No one feels guilty for skipping the methods section of a scientific paper. The dry prose describes procedures that the authors used to obtain their data, dutifully mentioning every detail such that another researcher could conceivably replicate the experiment from start to finish. Most readers focus instead on the discussion section, which summarizes the results, offers a meaningful interpretation of them, and makes predictions for future experiments. By comparison, the methods section holds little allure.
Portraits of the Mind, a new book by Carl Schoonover, aims to change that. With a collection of stunning images that illustrate a wide range of research techniques in neuroscience, it invites lay readers to revisit the methods section. Consider the image of a monkey’s visual cortex which, eerily, has a smiley face superimposed on it. The smiley face is no accident. To produce the image, researchers aimed a high-speed video camera at the monkey’s visual cortex, an area in the back of the brain that processes visual information. The camera captured small increases in blood volume, which correlate with increases in brain activity. Crucially, the monkey was looking at a picture of a smiley face when researchers snapped the photo. The reproduction of the smiley face in the cortex demonstrates a key feature of the visual system: adjacent objects in the real world occur as adjacent representations in the brain.
Published by Abrams, Portraits of the Mind has the look and feel of a coffee-table art book. It offers short texts to accompany each image and concise essays, contributed by eight specialists, to introduce each chapter. Schoonover knows that some readers will want to enjoy the images on their own terms: “If you feel your eyes glaze over where the text gets complicated and qualified, fear not, just move on — you won’t be missing much.” This statement, however, does a disservice not only to Schoonover’s careful texts but also to the complexity of the images, most of which require more work to interpret than the smiley face. In one image, for example, a set of neurons in the retina literally glow in the dark; curiously, the dendrites (branches) of these neurons do not reach out at random, but all align in the same vertical direction. To produce the image, researchers used a genetic copy-and-paste method. They inserted a jellyfish gene, which produces a florescent green color, into a mouse’s DNA, and switched on the gene only in a subset of neurons. The result is aesthetically pleasing, to be sure, but also reveals a striking interplay of form and function that we can glean only from the text: the illuminated and vertically aligned neurons happen to be precisely those which, in the retina, detect objects that move in an upward direction.
Portraits offers concise descriptions of many more methods, including revolutionary techniques from the late 1800s as well as a slew of more recent developments from the 2000s. The list is long and diverse, but also, as the book points out, troubling: these methods, impressive as they are, actually produce so much data that no model can yet make sense of it all. The “Brainbow” method offers a case in point. Brainbow also works with fluorescence, but it uses DNA sequences for multiple different colors, such that a given neuron can glow not just green, but red, blue, chartreuse, yellow, or orange. This method allows researchers to distinguish individual neurons from each other despite the fact that they are tightly packed together (thus, a green neuron can be distinguished from its neighbor, which might be red, and so on) and produces some of the most striking images in the book. It does not, however, provide any clue as to why one specific neuron forms a connection with another neuron; furthermore, the actual color assumed by a given neuron is not meaningful, but random, making the images difficult to decipher.
Brainbow and other methods are thus “limited, paradoxically, by the dizzying volumes of data they generate, which must then somehow be interpreted” (emphasis mine). Two possible approaches to the interpretation problem recur in the book; the tension between them is apparent in Schoonover’s texts as well as Terrence Sejnowski’s introduction to Chapter 6, “The Brain as Circuit”. The first approach is to simply wait for a genuis: eventually, someone with a special vision will create a new model that radically alters our interpretation of neuroscientific data. It’s happened before. In the late 1800s and early 1900s, Spanish scientist Santiago Ramón y Cajal combined newly developed methods for visualizing individual neurons under a microscope with his own peculiar artistic talent. His efforts led to the formulation of the Neuron Doctrine — the now widely-accepted idea that individual cells form the basic computational unit of the nervous system — and earned him the Nobel Prize for Medicine in 1906 along with Camillo Golgi. (Golgi, incidentally, vehemently disagreed with the Doctrine).
The second approach to the interpretation problem involves machines, not people. Computers can detect patterns in large data sets that no human could possibly see on his or her own. If we let computers crunch the numbers, we might eventually build connectomes, or wiring diagrams that indicate how each neuron connects to every other neuron. Researchers have already created a nearly complete connectome for the worm C. elegans, whose nervous system contains just 302 neurons. With increased computational capacity, connectomes for other animals and humans might eventually become possible as well. This approach has the obvious advantage that non-geniuses can still produce valuable work, by programming computers to extend findings from smaller data sets to increasingly larger ones.
Sejnowski, a professor at the Salk Instiute for Biological Studies in San Diego, pioneered the computational approach to neuroscience. His essay in Portraits describes how, in 1980, he showed that an artificial neural network could learn some surprisingly difficult tasks, such as pronouncing English words. Today’s computers can create far more sophisticated neural networks that simulate not just individual neurons, but also fine dendritic branching and multiple synapses. Yet even Sejnowski admits that the computational approach, for all of its advantages, falls somewhat short: “…even if we could reproduce all the anatomical details and signals in a brain, this wealth of knowledge would not in itself explain how a brain functions, or goes awry. What we need is a twenty-first-century Cajal who can understand the function of these circuit diagrams by simulating the signals as they are processed by the circuits themselves.”
In a book focused on methods, the desire for genius strikes a funny note. Dry as it is, the methods section of scientific papers permits any researcher to replicate the experiment — no genius required. Of course, the researcher still needs certain technical skills, as well as background knowledge, but she need not be the next Cajal. After replication, the researcher can propose novel interpretations for the data and variations on the methods used to obtain it. In other words, she can advance the state of our knowledge, one step at a time.
Ultimately, though, the egalatarian spirit prevails in Portraits of the Mind. For readers seeking eye candy, the book delivers nothing short of beauty. For readers seeking to learn something new, the book offers a great introduction to neuroscience — no genius, and no technical skills, required.
Schoonover, Carl. 2010. Portraits of the Mind: Visualizing the Brain from Antiquity to the 21st Century. New York: Abrams.