The neural basis of our ability to perceive a rich palette of smells can be compared with the neural basis of our ability to perceive a rich palette of colors. The best way to illustrate this is by comparison with pointillist art.
There are two ways to put colored paint on a canvas to elicit the perception of color in the mind of the observer. One is to mix the paints to achieve a particular color impression: red and white to achieve pink, blue and yellow to achieve green, and so on.
The other is to place different colors in individual small “points” of color and let the effect of color mixing arise from a distance, where the colors blend to produce the mixed color in the mind of the observer. This was the method of “pointillism.” It was invented and perfected by Georges Seurat, who had drawn his inspiration and methodology from visual theorists, most notably Wilhelm Helmholtz. Paul Signac was another well-known practitioner of this art form.
Seurat’s painting A Sunday on La Grande Jatte (1884), for example, is produced by thousands of tiny dots depicting several dozen figures on a bank of the Seine outside Paris on a sunny Sunday afternoon. The closer one is to it, the more one can see the individual dots of color, but the less one can see the figures they are forming; further away, the dots fuse and the entire picture comes into focus. The painting hangs in the Art Institute of Chicago. (You can appreciate the description that follows by observing the painting in full color on the Web.)
Reproduction by dots in fact became the method of publishing photographs and other graphic art in newspapers and magazines, in which a picture is made of tiny dots, each of a different shade of gray or of color. In art, pointillism reached its reductio ad absurdum in the work of Andy Warhol, who made large dots themselves the object of the artwork in the pictures of cartoon figures and Hollywood celebrities.
Several years ago, Terry Acree at Cornell University and I independently realized that these examples from the modality of vision can serve as useful suggestions of how smells are perceived. The illuminated dots of colored paint reflecting different wavelengths of light are analogous to the modules called glomeruli in the olfactory bulb, each activated preferentially and differentially by different odor. The patterns of colored dots seen in La Grand Jatte correspond to the patterns of activated glomeruli (see figure 8.1). The perception of the color patterns requires their registering from a distance in order to yield the complex effects of color mixing. By analogy, the perception of smell would require a “distance” provided by a readout that can produce “odor mixing” of the effects of neighboring modules. That distance illusion merges the pointillistic representations of the color patterns, as shown by the patterns in figure 8.1.
The neural circuits in the olfactory pathway in the brain are organized to carry out this series of operations. This series was summarized in figure 7.1. We take up each step along the way to gain insight into how pointillist representations of smells create their perceptual worlds in our minds.
The analogy with color has another very revealing similarity. What we call color is a perception that arises from electromagnetic waves of different wavelengths activating our photoreceptors to different extents; it is the brain that creates “color” by the way it processes the signals according to the different wavelengths. Similarly, “smell” is not present in the molecules that stimulate the smell receptors. It is the brain that creates the perception of “smell” from the differences between the features of the different smell molecules. This is a fundamental basis of smell and flavor. Hence, the subtitle of this book. As we move into the brain, we pass from the domain of the stimulus through processing circuits to the domain of the brain that creates our subjective world. Those processing circuits are next.
Comparing Color and Smell
We have pointed out that color is not contained in the wavelengths or photons of light. It is a sensory quality, created by our brains. Every time we view a flower, we are reminded how dull a scene is if the different wavelengths give it only shades of gray instead of, say, a bright red, which is red only because our brains have neural circuits that select that wavelength of light and create that internal perceptual quality. In the same way, we postulate that our smell world would also be only a smear of shades of inchoate sensations set up in our noses if we did not have the circuits in our smell pathway that select specific types of molecule-stimulated activity to give them a quality we can identify as distinct from all the rest. The ability to create this quality—philosophers call it a qualia—of a specific smell starts with the remarkable glomerulus.
Processing the Pointillist Image
In all other sensory systems, the brain employs successive stages of processing to extract what is most important for the immediate needs of detecting and discriminating among its stimuli. In the smell pathway, a sequence of stages similarly processes the initial “odor image.” Comparisons such as that in figure 7.1 with the visual system make this sequence more intuitive.
First is the processing of the initial odor image in the olfactory bulb at the layer of the modules forming the image. This image is then enhanced by a powerful system of lateral inhibitory microcircuits. The enhanced image is sent to the olfactory cortex, where a cortical microcircuit with widespread connections reformats it into a content-addressable memory. The final stage is for this memory representation to be sent to the highest centers in the neocortex, where a complex cortical microcircuit gives rise to conscious perception.
In sum: form a pointillist image, process it locally, format it globally, represent it in memory, enhance it with emotion, and perceive it consciously.
Each of these steps is performed by its own microcircuit, which will be described in subsequent chapters.
The Glomerulus: A Universal Smell-Detecting Device
In our analogy with pointillist painting, the points are the olfactory glomeruli. The glomeruli constitute an amazing little structure, the most distinct multicellular unit in the brain. Little wonder that Edgar Adrian advised me to “look to the glomeruli,” that we have regarded it as the basic functional unit of this sensory system, and that molecular biologists are focused on its organization and function.
A glomerulus is essentially a densely packed meeting place, where signals from the nose are transferred to the brain (see figure 7.1). Here the terminals of fibers (axons) from the receptor cells in the nose connect to the short branches (dendrites) of nerve cells in the olfactory bulb. A single glomerulus receives not just hundreds but thousands of incoming nerve fibers from the receptor cells. In the rabbit, for example, there are some 50 million olfactory receptor cells and some 2,000 glomeruli. This gives a ratio of, on average, 25,000 cells for each glomerulus. We say that this defines a convergence ratio of 25,000:1. This is one of the highest ratios of a given cell type and a given target in the brain.
The molecular biologists Linda Buck, Richard Axel, Peter Mombaerts, and their colleagues have shown that all the fibers coming to one glomerulus express only one type of olfactory receptor in their cilia in the nose. This means that all the fibers are carrying the same information, making the convergence ratio even more impressive.
Imagine that 25,000 people are talking to you at the same time. How would this work? It would not if it involves being in a monster cocktail party where everyone is saying something different or the same thing at different times. We would call this noise, and technically this is what it is. However, if they are all shouting, in unison, “Happy Birthday” or singing a song together, the words will be loud and clear. Technically, this is the “signal.” In signal theory, the converging simultaneous inputs raise the “signal-to-noise ratio,” thus making the signal much more distinct.
We postulate that signal-to-noise enhancement is a key operation of the glomerulus. This is used in orthonasal smell to detect and discriminate among specific signals in the environment that may be critical for survival, such as the scent of prey to attack or predators to avoid. Similarly, it is likely used in retronasal smell to detect and discriminate the volatile components of food released within the mouth. These components may also be critical for survival, in signaling the ripeness of fruit or whether fish or meats are fresh or rotting. They may have been critical for exploring new food sources, as during the great migrations of people that occurred during human evolution. And today, we use them to enjoy our food and discriminate among our wines.
Why should nearly all animals, including humans, need this module for their sense of smell? The odor environment consists of thousands of different odor molecules and odor objects, out of which we must identify only those that are behaviorally significant. This environment constitutes the background of asynchronous noise, out of which the animal must detect the signal. To do this, the olfactory system does not have to spend expensive neural tissue to track its stimuli in space, as does the visual system with its constantly changing visual scenes. Instead, the olfactory system can build immovable detection modules that wait for their appropriate stimuli to come to them. To adapt the analogy made famous by the Oxford philosopher Isaiah Berlin, the visual system is like the fox, in which each location knows many things, and the olfactory system is like the hedgehog, in which each glomerulus knows one thing.
The glomerulus is thus a detection device par excellence. Just as the points of paint at a particular wavelength in a pointillist painting give brighter colors than if the paints were mixed on the palette, so we can imagine that the points of “smell molecule features” represented by a glomerulus shine more brightly (can be perceived more sharply) against a mixed-smell background. Note how these follow the principles of pixelation of a visual image explained in the previous chapter. This special property may be an important reason why the glomerulus is the nearly universal module in the processing of odors by animal species, from humans to insects.
How Fine An Image?
In our analogy with vision, the fineness of resolution of the odor image must depend on several factors. First, of course, is the number of receptor cells; presumably the more the better. This is consistent with the dog having up to 100 million, around 10 times the number for rodents and humans. The next factor is the number of different types of smell receptors. Here the rodent comes out on top, with over 1,000, while the dog has some 800, and the human has around 350. Finally is the number of glomeruli. The dog has several thousand glomeruli, and my colleague Charles Greer and his colleagues have recently shown that the human has even more, up to around 6,000. It appears that more receptor cells, receptor types, and glomeruli give higher resolution in our smell images. However, the greater complexity of the human brain in analyzing the images becomes a critical factor, as we shall see.
Single Modules Interact Specifically
So far in the human, we have several thousand modules, each responding independently. They are acting like separate “labeled lines.” But it is another axiom of signal theory that the receiver—the brain—cannot make any sense of the information in those labeled lines until it can compare them. There must be mechanisms for correlating and comparing the activity of the pathways with one another. Specifically, lateral interactions between the glomerular modules are needed. These interactions begin through interneurons called, appropriately, periglomerular cells, which connect to neighboring glomeruli (see figure 7.1).
Physiological studies have shown that the periglomerular cells respond to odor input with single impulses or with impulse bursts. One of the physiological actions that has been identified is inhibition of mitral and tufted cell dendrites issuing from surrounding glomeruli. It has been hypothesized that this may enable a more active glomerulus to inhibit its less active neighbors—the kind of lateral inhibition we have discussed. This action would contribute to signal-to-noise enhancement of the odor image.
Another kind of interglomerular action may be excitation. This could involve direct excitatory synapses from a subtype of periglomerular cells that has an excitatory neurotransmitter. But it could also involve a special type of inhibition, in which an excitatory periglomerular cell axon excites a distant inhibitory cell, producing a distant but local inhibition of the mitral and tufted cells in the target glomerulus. In a study conducted in 2003, Michael Shipley and his colleagues at the University of Maryland obtained evidence for this kind of action. Some of the periglomerular cells with these actions target glomeruli at considerable distances away, indicating that extraction of the odor pattern involves complex interactions that are coordinated over several glomeruli. Tom Cleland and his colleagues at Cornell University have suggested that the global effect of these actions may be to normalize the excitability at the glomerular layer, to keep it within a working range, regardless of the intensity of odor stimulation.
We are only starting to learn about the subtypes of periglomerular cells and the kinds of interactions that occur between glomeruli. At this point it seems safe to say that one effect of these lateral interactions is to begin the extraction of the spatial pattern so that it can be read more effectively by the next level of microcircuits involving one of the brain’s most enigmatic cells, the olfactory granule cell.