THE INDIVIDUATION OF THE SENSES
The Individuation of the Senses[1]
How many senses do humans possess? Five external senses, as most cultures have it—sight, hearing, touch, smell, and taste? Should proprioception, kinaesthesia, thirst, and pain be included, under the rubric bodily sense? What about the perception of time (Le Poidevin, this volume) and the sense of number?
Such questions reduce to two.
1. How do we distinguish a sense from other sorts of information-receiving faculties?
2. By what principle do we distinguish the senses?
Aristotle discussed these questions in the De Anima. H. P. Grice revived them in 1967. More recently, they have taken on fresh interest as a result of a collection of essays edited by Fiona Macpherson (2011a). This entry reviews some approaches to these questions and advances some new ideas for the reader’s consideration.
I.
In order to count the senses, one must know not just how they are distinguished from one another, but also how they are differentiated as a group from other capacities. Obviously, faculties such as breathing and digestion are not senses, because minimally, a sense is a faculty by which organisms gather information[2] about the current state of their environment. But this is not a sufficient characterization, for it misses an important characteristic associated with the traditional notion of a sense. (In what follows I do not distinguish the external and the internal environment. As we shall see the question whether thirst is a sense modality does not turn, for me, on whether its milieu is internal or external.)
The body gathers environmental information for homeostatic regulation. For example, blood CO2-level is monitored by dedicated chemoreceptors and used to regulate breathing. But this information feeds directly into homeostatic control and is not available to the subject. At least, it is not directly available to the subject—you may come to believe that your blood CO2-level is high, but only if you know that this is what is causing you to breathe faster. Certain kinds of blindsight are similar. There are patients who, because of damage to a part of the visual system, lack conscious vision, but are able to execute visually guided action. These patients can become adept at knowing certain aspects of their environment—for example, the shape of things they handle—simply by monitoring their own bodily responses—in this case, grip shape. (Goodale and Milner 2004 describe the everyday life of one of these patients, DF.) Consider also the various agnosias, acquired inabilities to recognize higher-level objects such as faces or people. As Oliver Sacks (1970) observes, visual agnosics (who possess acute visual discrimination for lower level detail such as colour, oriented lines, and shapes) lose their ability directly to recognize higher-level objects. They can, nevertheless, recognize these objects by recognizing distinctive parts—for example, faces by protuberant teeth and other such distinguishing marks. What most people recognize holistically, directly, and sub-personally, these patients identify indirectly.
Another point to consider: when subjects voluntarily over-ride homeostatic breathing-control—for instance, when they deliberately breathe slowly and deeply after a run—they are not contesting or “withholding assent” from the output of the CO2 detecting system. Having no access to this, they exercise direct control of breathing independently of this output.
By contrast, subjects have direct access to the output of the visual system. When my visual system determines that something is blue, the thing looks blue. My response to this sensory appearance depends on the cognitive context. Normally, I would infer that what I am looking at is blue, but if I have some reason to believe that the lighting is unusual or that my senses are deceiving me, I might withhold assent to the appearance, and refuse the inference. Vision does not control belief; rather, it sends a “message” about how things are with regard to certain visual characteristics.[3] It is up to me whether to believe the message in light of all of the evidence available to me. In short, vision provides me perceptual “content”, which I use in a context-dependent way.
One theory is this: the senses (as well as the emotions, etc.) post their output to a temporary store called “working memory” where it is held for a short period. Various cognitive systems, including belief-formation systems, have access to this content in working memory, and use it in conjunction with other information. (Baars 1997 calls this temporary store the “global workspace”.) J. J. Gibson’s (1977) notion of “affordance” is relevant here: sensory states can be regarded as delivering to working memory a message that objects or events afford the subject the possibility of action. I would, however, qualify Gibson’s claim by stipulating that “action” should includes epistemic actions such as inferring or questioning or revising belief; in other words, “affordance” should include epistemic affordance.
The “traditional senses” characteristically process current environmental information and make it epistemically available for the non-autonomic, context-relative, initiation of action, including epistemic action. In humans, such epistemic availability is provided through consciousness and rationality—by consciously experiencing something as blue, I gain rational support for the belief that it is blue. In lower animals, these conditions may not be met. A honeybee sips from flowers that bear one coloured pattern but not from those that bear another; its action is visually guided. But we have no idea whether it performs this act of discrimination by anything involving conscious awareness. Nonetheless, its action is context-dependent: it may be time to get back to the hive, and so the bee may abstain. Thus, the bee’s visual system cannot be said to control action directly: it too posts output to a buffer to interact with other informational and motivational states. Here too, vision provides informational content. In humans, the interaction with other states is (at least sometimes) rational; in higher animals, the content is often conscious. But these conditions are not necessary for something to count as a sense.
Now, the eyes also provide information for autonomic control. For example, information regarding the brightness of incident light is used to control pupillary aperture and circadian rhythms. (These happen through pathways independent of conscious visual perception.) So one cannot define the sense of vision simply as information-extraction from light (as does John Heil 1983, chapter 1), or by the eyes. The best way to accommodate this kind of regulatory function is to introduce the notion of an inferential process in the brain that computes a function from states of the sense organs to states of the surrounding environment. In other words, these inferential processes extract information about the environment from states of the sense organs. A subset produces perceptual content as defined above. Call the latter perceptual processes. A certain group of perceptual processes constitutes the sense of vision, another group the sense of flavour. The vision group will include the processes that deliver content regarding colour, shape, and so on. The flavour group includes the processes that deliver content on sourness, creaminess, fruitiness, and so on. The question we are addressing is this: how are we grouping perceptual processes when we differentiate the senses?
A second mark that distinguishes the senses from other information-gathering faculties is that they modify response patterns through learning, in particular conditioning. A response to one sensed quality or object A can be transferred to another quality or object B because A and B have been sensed together and have thus become associated. Suppose that Suzy makes coffee while she prepares breakfast. Even on days when she skips breakfast, her stomach growls at the smell of coffee. Her food-anticipating reaction spreads from food itself to the smell of coffee.[4] Information-capture for regulatory purposes does not feed into conditioning in this way. Suppose that most mornings Suzy goes for a run while sucking on a mint lozenge. Each day, she raises her blood CO2-levels while experiencing the minty taste, and her body responds to the former by breathing hard. Now suppose that one morning she sucks on the mint as usual, but does not go for a run, she does not breathe hard. The (unconditioned) response to running is not transferred to the flavour of the mint; associative learning does not occur. It is a mark of a sense that it supports learning. The information that regulates breathing is not sensory, on this way of looking at things, but the smell of the coffee is.
Both the above marks of the senses are forms of environmental flexibility—the first is for accommodating events that are relevant to action in a context-sensitive way, and the second is for taking advantage of environmental regularities that vary from place to place or time to time (and should not therefore be hardwired). Quite often, the same systems afford flexibility of both kinds (or neither). That is, the kind of information-capture that provides input to context-relative initiation of action most often also serves to provide input to processes of conditioning. Are the two marks coextensive? Hume took it for granted that they are; in his empiricist framework, only the senses provide materials for associative learning. Later associationists do not question this, though they are not explicit about how they are using the term ‘sense’. (See Pavlov 1927, Lecture 1, and Rescorla and Wagner 1972 for representative examples.) Still, the two marks are conceptually distinct, and one might want to distinguish event-response sensory systems from association-learning sensory systems (allowing, of course, a substantial overlap).
The conditions proposed thus far are not yet sufficient. The “number sense” by which primates estimate the numerosity of (sufficiently small) collections (Gelman and Gallistel 1978, Carey 2009) seems to satisfy the above criteria—in particular, it seems to provide direct awareness of number. But it is not a sense modality. The reason is that input to perceptual systems comes from transducers—cells that convert incident energy into a neural pulse that carries information about this energy. (For example, the rod cells of the eye convert light into neural pulses that carry information about this light.) Sensory transducers respond to the environment in physically determined ways: for instance, the basilar membrane in the ear is so constructed that different parts resonate, as a matter of physical law, to different auditory frequencies. The number sense lacks transducers. It operates equally well in different modalities—we can quickly estimate the number both of small collections of successive light flashes and of successive sounds (though possibly not of tactile stimuli—see Gallace, Tan, and Spence 2008). Collections of objects and events do not act on cells that emit a neural pulse that carries information about number. Rather, the number sense operates on the outputs of (some) other senses. It is thus a post-perceptual module, rather than a sense modality.
A further complication is that in a genuine sense modality, the neural signal emitted by sensory transducers is, as Brian Keeley (2002) points out, processed in ways that are “historically” (i.e., evolutionarily) dedicated to the recovery of information about external stimuli. Keeley notes that the weak electric current from a charged 9V battery creates a definite sensation on the human tongue. This is because touch and taste receptors respond to the electric current that the batteries emit. However, humans lack information-processing data-streams designed by natural selection to extract informational content about ambient electricity from this stimulation of touch receptors—the sensation produced by a weak current feels like a tactile stimulus on the tongue, or sometimes like a flavour, not like an event of a distinct type. By contrast, electric fish have systems dedicated to processing information about electrical fields, and they perceive quite specific features of these fields. Sharks, for example, detect prey by the disturbances in the electric field caused by their movement (Hughes 1999). So though both humans and sharks sense electric current, sharks (but not humans) have a sense modality for electric fields or currents.
These ideas are important in deciding whether pain and sexual arousal are sensory states; both were posited as such by some historical authors (Dallenbach 1939) on the grounds that each is associated with a special kind of experience or quale, which is, moreover, informative about current circumstances. In the historical debate about pain, subjective experience proved inconclusive. Is the pain of being burned a particularly intense, and hence unpleasant, sensation of heat, or is it a distinct sensation that accompanies intense sensations of heat? Introspection does not decide the question; yet, whether pain is a sense depends on it. The transducer/processor condition is helpful. It appears that pain relies on receptors activated by high threshold values of mechanical, thermal, and chemical stimuli, and processed by a dedicated system in the brain (Craig 2002, 2009).[5] If this is correct, then there is reason to hold that pain is a sense, and that sensations of pain do not belong in the same modality as the intense stimuli that are associated with it—sensations of warmth etc. Thirst belongs in this category too: it has dedicated transducers known as osmoreceptors and a dedicated computational system (McKinley and Johnson 2004).
Sexual arousal is different from pain and thirst. First, there are no transducers for sexual arousal; it arises in a context-sensitive way from other perceptions, some visual, some tactile, etc. Secondly, arousal is more like something perceived than a perception. It is a state of the body that enables sexual performance. Now, there may well be processes that detect some of the bodily changes that constitute sexual arousal, and this may account for its characteristic feel. But this is to say that the feel of sexual arousal is a perception of arousal. This is different from saying that arousal is itself a perceptual state.