SENSE5 8/18/90; rev. 1-19-98.  (original 1983)
copyright David Cole
University of Minnesota, Duluth

Sense and Sentience

    Surely one of the most interesting problems in the study of mind concerns the nature of sentience. How is it that there are sensations, rather than merely sensings? What is it like to be a bat -- or why is it like anything at all? Why aren't we automata or responding but unfeeling Zombies? How does neural activity give rise to subjective experience? As Leibniz put the problem (Monadology section 17):

It must be confessed, however, that Perception [consciousness?], and that which depends upon it, are inexplicable by mechanical means, that is to say, by figures and motions. Supposing that there were a machine whose structure produced thought, sensation, and perceptions, we could conceive of it as increased in its interior size with the same proportions until one was able to enter into its interior, as he would into a mill. Now, on going into it he would find only pieces working upon one another, but never would he find anything to explain Perception. [Montgomery trans.]

    Attempting to answer the question of how a physical system can be sentient is a foray into speculative psychology. The problem, it seems to many, is to overcome exactly the obstacle posed by Leibniz, of providing an explanatory link between the physical and the subjective. Leibniz himself thought the gap was unbridgeable, that no account of physical processes can possibly explain subjective experience. Modern functionalism suggests that we should look to an intermediate level of explanation for understanding mental phenomena. The level I will appeal to views mentality in terms of information processing. This divides the problem of understanding the link between physical and mental into two sub-problems, one relating central nervous system processes to information processing, and the other relating information processing to mental processes, including that which is my topic here, sentient experience. The hope is that the principles bridging these are more perspicuous than the initially unfathomable gap to which Leibniz pointed.

Section I: Background Historical and Biological

    Despite the fact the information processing approach to mentality rejects dualism, many of the considerations I will raise owe to Descartes. And functionalism concedes to Descartes that there is a "Real Difference between Mind and Body" (the subtitle of Descartes' Meditations)check subtitle, although it is a difference of level of description rather than substance. There is a contemporary tendency to focus on aspects of Descartes' thought which have subsequently been largely abandoned, often with great difficulty. But in several respects he is truly modern. Notable among these in this context is his break with the Aristotelian account of biological phenomena in terms of types of soul. Descartes attempts to account for animal behavior by purely physical processes; this is a decisive advance. Descartes is then at pains to distinguish beings that merely sense from those that are sentient. Descartes supposes that this distinction marks an important underlying metaphysical difference, and he draws the line between the sentient and the merely sensing such that only human beings are above the line. Although he appears to be mistaken on both counts, the evidential and functional considerations he raises shed light on the difference between mere sensing and sentience that will be the focus of my discussion, along with a point about self-awareness raised by Kant, as well as some contemporary observations.

    At least three basic mistakes have been made in thinking about sensations. All three are evident in the Way of Ideas, traceable from Descartes, through the British Empiricists, to the positivism of the first half of this century. First, sensations have been treated as things. They have been regarded as ontologically similar to physical objects in being discrete entities with determinate character, properties of their own (a mark of Substance), and as things which are now present to the mind, now not, and now present again. Yet, of course, they have always been seen to be a special sort of object, mental objects, and generally (the exception being in neutral monisms, which make them basic) regarded as somehow dependent for their existence upon minds, as physical objects are somehow dependent upon space for their existence.

    The second mistake follows on the first, and that mistake is to treat sensations as atomic, as not resolvable into components of any kind, let alone into components which are not present to consciousness. For, it seems to have been reasoned, the whole being of sensations consists in their being present to consciousness; hence they cannot have another form of being which consists in something not present to consciousness.

    The third mistake, related to the first two and again coming largely from empiricist tendencies in philosophical psychology, is to hold that sensations are temporally prior to, and can exist independently of, intellectual operations of understanding and categorization of the information from the senses. The Mind, on this view, consists of a bundle of sensations; mental operations, such as abstraction, comparison, combination, etc., are performed upon some of these sensations; others are neglected but are full-bodied sensations nevertheless.  The cognitive is built upon an infrastructure of qualia.  These qualia are similar in all organisms with similar sense organs, regardless of their differing cognitive abilities.  Organisms with different sense organs, e.g. bats, have different qualia.

    To avoid these mistakes, it seems to me that one first must realize that sensations are not objects at all (which is not to say that we do not have sensations or that we do not sense). Rather, the having of sensations, a sort of event, does occur, but sensations are not things had. Secondly, one must realize that these mental events may well be resolvable into subevents which may not be themselves discernible to the one having the sensation, that is, not themselves present to consciousness. This is the thrust of much work in contemporary cognitive science.

    Thirdly, and the point I shall dwell on because it is both the most enlightening and controversial, one must not suppose that having a sensation can occur unless the data from the sense organs is subjected to complex processes of categorization, understanding and hypothesis testing. Intelligence is a precondition for sentience, not vice versa. While I shall press this point hard, it is in fact just a particular development of venerable criticism of tabula rasa theories of mind.

    This is not the place for a full-scale defense of the information processing model of mentality, but a few words in passing are appropriate. On the one hand, it seems clear that the central nervous system is well suited to information processing. Sense organs are well-suited to obtaining information about the environment. Indeed, that is surely the reason the sense organs exist; provision of information is their proper function (cf. Ruth Millikan's work). Some processes in the central nervous system are known to indicate the presence of certain character in the information presented to the senses. Thus, there is physiological evidence that analysis of information occurs in the nervous system. The same result is supported by more global, ecological considerations. It is clear that survival might be enhanced by acquisition of information about the environment, and by use of this information in controlling behavior. Finally, at least in the case of our own species, there is the phenomenological evidence: we are aware that our senses provide information about our surroundings, and that we use this information in a variety of ways. This confluence of different types of evidence -- physiological, evolutionary, and phenomenological -- suggests that we look to an information processing model for the explanatory bridge between the physical and the mental.
I shall have little to say beyond this about the way in which neuronal processes relate to information. Information from the sense organs is initially encoded either by firing rates of individual neurons, or by which of a group of neurons are active. Both schemes are used in artificial systems. For example, in automotive engine control systems information about angular velocity may be represented by pulse frequency, and the location of the shaft turning (distributor, crankshaft or axle) corresponds to distinct lines. Ultimately, in both organism and artifact, the information from the sensors is used to control behavior.

    Sense organs generate information about the environment, but they vary considerably in the generality of that information. Internal sensors (e.g. of blood salinity) are the most specific; taste is also narrow, whereas vision provides enormous amounts of information that is very general, in that it can potentially indicate the presence of most anything. And, most importantly, it is information about conditions at a distance from the organism. This has numerous advantages, as is apparent by comparing looking about a room one first enters and attempting to gather comparable information blindfolded.

Section II: Sensing and Sentience

    It should be clear at the outset that we must adequately distinguish two phenomena.2 First, sensing is something which animals, some plants, and some machines can do. Sensing involves a sensing organ or device which enables the system of which it is a part, organic or inorganic, to actively respond to environmental circumstances. A system may sense light, sound, temperature, or the presence of water, prey or predator. Sensing involves responses, perhaps entirely internal, on the part of the system, but sensing is usually manifest in behavior. Although everything "responds" passively to outside forces and conditions, the response of a system which can be said to sense is active. The discriminations of which a system is capable tell one what it senses. Nevertheless, determining just what is sensed by a system may require careful experimental design, as is the case with determining what migratory birds and fish sense in fixing their long courses. Sensing is more highly developed in animals than in plants, reflecting the information requirements imposed by motility.

    Second, sentience, having a sensation or a feeling, or "qualia", is a phenomenon which goes beyond more than mere sensing, for it involves an internal state in which information (typically) about the environment is treated by the system so that it comes to have a subjective character. We know what this is like from our own case. Each normal person has had sensations of cold, bright light, sound, and pain. It is from such occurrences that we understand the reference of "having a sensation".

    Once we distinguish sensing from sentience, we may note that sensing is neither a necessary nor a sufficient condition for sentience. Existing artificial systems may sense without being sentient; that is, they may make sensory discriminations but not have sensations. An example would be a modern pollution minimizing and efficiency maximizing electronic automobile engine control system, which senses oxygen levels, various temperatures, manifold pressure, and so forth. And a human may be sentient without sensing, for it may internally generate the sensations, in which sense organs play no role -- it may hallucinate or dream.

    It is not just artificial systems which may sense but fail to have sensations. Presumably some primitive animals sense without sensations, and under certain circumstances so do human beings. The operation of a human body, maintaining homeostatic conditions, is a complex affair, in some respects similar to chemical refinery operation. Biological sensors monitor blood acidity, light level, and so forth, and control the size of the iris in the eye, heart rate, and a host of other biological processes. Although the control is neuronal, and involves sensing, it is effected without mediating sensations.

    Even intelligent response to sensory stimuli is not by itself a sufficient condition for the having of sensations. But it is the most important step in the right direction. Whether or not a sensation is had must depend upon the exact character and complexity of the information processing which leads to the behavior. And in the case of animals there is always a tendency to exaggerate intelligence, in that more information processing is attributed than is necessary to account for the behavior.
Unlike many lower animals, primates and especially humans must have very many and very generalized feature detection abilities. This is because so much of human response to stimuli must be learned. All of this information from the senses must be available, in humans, at the highest level of information processing, where detailed comparisons of color, shape, textures, sizes, and so forth, must be made. Almost no human behavior could be unqualifiedly a universally appropriate response to some sensory stimulus; the appropriate response is almost always contextually determined in complex ways. And because humans are creative, are makers, they have additional need to reflect on sensory features, in ways that perforce could not have been anticipated genetically, so as to adjust a shade of a color, or the shape of something -- and not just for artistic purposes, of course: subtle differences in shape make all the difference for a constructed device of whether it will work or not, hum nicely or blow up.

    The main idea I wish to emphasize is that having a sensation, however phenomenologically simple, is actually a very complex business. It is a complex doing with data. The data come from the sense organs, and the doing is a very special kind of information processing, a kind that human brains can perform, but don't always perform. And it is processing that is highly dynamic, with constant updating and ever shifting foci of attention.

    The sorts of activity involved in having a sensation include comparing phenomenal features with one another, say contrasts in color, or with features remembered from past sensory experience; categorizing features; remarking their intensity; distinguishing outline from internal features; remarking characteristics of timbre; using features in hypothesis testing about the identity and unseen character of the thing seen, heard, felt, etc. And the comparison and categorizations must be done over time: one must be in a position to notice if the color, shape, attitude of the observed remain unchanged during the period of sensing or not, and if changing, whether one or another pattern of change fits. If none of these things go on (and none of them need be fully conscious), no sensation is had.

    When the repertoire of possible response by a system to an environmental condition is small, that response will be automatic with no mediating sensation. But as the complexity of the possible responses of which the system is capable increases, then the information from the sensors assumes an increasingly different status from mere triggerer of effectors. This is clearly seen where both automatic, reflexive, and sentient responses to stimuli occur. In the case of human beings, for example, a burn on the fingertips causes the arm to be withdrawn -- prior to any sensation of burning pain. Similar withdrawal responses to normally painful stimuli can be elicited in humans where the spine no longer communicates with the brain -- the response is mediated at the spinal level and there is no sensation. But not all response can be that way.

    The stereotyped, reflexive withdrawal of a burned extremity solves a short-term goal of damage containment. But to serve a long term goal, the human must then be induced to consider ways of avoiding such damage in the future. The lesson cannot be that one must not go near stove, it must be that one can touch some surfaces and not others under certain conditions, indicated by a variety of sensory information. And, of course, the human must be induced not to use his or her considerable skill to defeat a reflex. In the absence of sentient inducement to the contrary, humans might well seek to overcome the reflex, say by using the other hand to force the reflexively withdrawing hand to endure the heat. Pain takes up where the reflex leaves off, as a consideration to be taken into account in doing high-level strategy planning as we seek to achieve our ends.


Section III: Hypothesis testing and sentience

    Most of our responses, to be appropriate, require complex information processing, calling up other information about the environment, not all of it available to the senses at the present, and possibly the application of learned general principles and logical inference. And there may well need to be hypothesis testing. Humans must be able to match their current non-propositional sensory input with propositional knowledge in order to respond intelligently to a novel situation. As an example, I wonder if the unfamiliar type of cheese which a friend gave me a week ago has gone bad. I remove it from the refrigerator and take a whiff. It definitely has a strong pungent odor. Milk should never smell this way -- but perhaps it is just right for this cheese. I attend to various nuances of the aroma and at the same time attempt to recall how it had smelled when I smilingly accepted it last week. I look at the coloration and recall cheeses-I-have-known with similar shades of green and yellow. Even after moving it and examining it in better light, I am unsure and rewrap and return the cheese to the cold darkness. Even in this mundane situation, I must have a variety of non-propositional information available for categorization and hypothesis testing (Does that green come from mold? Recent mold?). I must be able to direct my attention, my current highest level information processing, consuming most of my information processing resources, at various features of the environment. The information from the senses becomes the material of sentience, when and only when it is being used as data by a complex analysis and hypothesis testing system.
Consideration of deciding what to do about information from the senses affords us a slightly different way of looking a sensation having. The utility of a sensation which forces itself upon us, such as a sharp pain (as opposed to a subtle nuance of bouquet, say, which we must make an effort to discriminate) is that it interrupts our cognitive activity and demands attention. It drives thinking about what to do about it. Our choices are typically many, and we can weight proximate cause removal, distal cause removal, ingestion of any of a number of analgesics, consultation of an expert, or endurance of the pain. We can't help thinking about the pain, but our choice about what to do about it is voluntary. Our reactions to stimuli which are not voluntary, e.g. reflexive limb withdrawal when burned, are not mediated by sensation. The sensation arises only when sensory information is present to our decision making processes.

    For most of us, paradigms of sensation having include savoring the taste of pineapple sherbet, or the blat of a trumpet, or a backrub or a sunset. Less savory, but just as paradigmatic, are experiences of strong stenches, electric shocks, and headaches. It is important to notice just how rare, for good or bad, this sort of mental event is in our lives. Life is not a series of sensations. Savoring information from the senses is a highly reflective activity, and expensive in the mental resources it requires. Comparisons are made, thoughts are suggested, associations abound, attention meanders about playfully, and an awareness that we are having the sensation moves in and out of the foreground. In savoring, we milk the data from the senses for all they (and we) are worth.
So on the analysis proposed here, having a sensation is something that goes on in some information processing systems when non-propositional data is being analyzed. The having of the sensation is not a distinct event from the analysis. The phenomenological character of the experience of the sensation depends upon the information processing, not every feature of which is conscious. Much of the input is pre-processed by feature discriminators, and various types of image-enhancers, and this pre-processing is not conscious. Nor are we conscious of much of the processing by which associations and memories are called to center stage, consciousness. The character of the experience itself depends upon the complexity of the information processing required to determine appropriate reaction. This reaction may be completely internal, involving no overt behavior. The input may be just analyzed and categorized (which involves hypothesis testing); perhaps mulled over, perhaps not, and not regarded as requiring action. But an experience takes place, and it just is this information processing.

    The analysis and hypothesis testing are, I argue, crucial to the existence of subjective experience. To see this, one has to ask "what functional properties make a process of neuron firings a having of a sensation?" Non-propositional information proceeds in considerable quantity through many systems without the occurrence of sensations. For example, at this moment millions of television receivers are processing information, information about various environments. But the response of the television receiving system to this information is completely stereotyped -- it doesn't treat it as information at all. It merely converts it to a pattern of lighted dots on a screen, which requires minimal processing. Sentience can appear only when the data are used for categorization involving hypothesis testing and comparison, where the data are referred to over and over, dynamically, and the range of possible categories, expectancies, and hypotheses is large enough so that the response is far removed from automatic.

    Normal humans have these capabilities, and so at least sometimes have sensations and can savor qualia. But subhuman animals to not have these capabilities or do not have them to the same degree as humans. Most notable of the behavioral deficiencies of sub-human animals, as Descartes noted, is the absence of linguistic behavior. The concern about language is not a concern about the incapacity by a dog to communicate the thought (example owes to Bertrand Russell) that the dogs parents were honest but poor. It is a concern that the thoughts are not there to be communicated, and so sensory inputs are not being put to the uses that create qualia. Animals do not have to figure out what is the right thing to say about anything. Natural language, even at the level of development of a human 6-year-old, is an extremely complex categorization system, the use of which requires enormous amounts of information processing. The task of describing, of determining the aptness of color and shape and other descriptors is evidence of a reflective sensitivity to perceptual features that gives clear evidence of the information processing abilities of humans, and so was rightly emphasized by Descartes as a sign of sentience. The ability to use natural language to describe what one senses is however a sufficient but not a necessary condition for sentience. Comparable categorization might take place in a propositional representation system, a "language of thought", without any expression in natural language. While this is conceivable, it seems to be unlikely that such non-communicable complexity would evolve. The utility of complex general discrimination abilities depends very much on information and general rules that must be learned, and there is serious reason to doubt that any single individual, deprived of communication of information garnered by others, could hope to begin to acquire sufficient experience to justify the sophisticated information processing abilities that humans have. The character of the human mind, including the ability to analyze the information from the senses, is made possible by the social character of human life. Thus while it is logically possible for a non-linguistic organism to have the information processing complexity that is sentience, it is doubtful that a species of such organisms could evolve. It is in the course of this information processing, and only in the context of such processing, that sensation, rather than mere sensory discrimination, occurs. Being a bat is much more like being a rock than it is like being Batman.5


Section IV: Sentience and Metaconsciousness

    The preceding considerations support the thesis that sentience in lower animals, to the extent that it exists at all, cannot be much like the sentience we humans know. A different analysis of the conditions necessary for sensation having can begin by noting the oddity of

1. S (now) has a sensation of cold but S does not know that S has a sensation of cold.

This is odd, but it might be the case if S did not know what cold was, having never experienced it before and so unable to categorize the sensation as one of cold. A more probable example would involve having a sensation of puce, without being aware that it was that particular color which one sensed.

But that line cannot explain the oddity of

2. S has a sensation but S does not aware that S has a sensation.

The sense of "aware" at issue here is a general one, synonymous with noticing, and does not require any facility with a natural language. In the sense at issue, cats are aware of mice, and mice of cats. Such awareness merely requires recognition capacity, but the ability to soliloquize. Can a cat, e.g., be aware that it is having a sensation Can there be a sensation in the absence of such awareness?
Generally, to be aware of something x does not require that one know that x is x -- one can be aware of something without knowing just what it is. For example, seeing a large object on the living room carpet, you are aware of the object, without necessarily knowing that it is the transmission from your teenage son's motorcycle. But in being aware of the object, you must distinguish it from other things, in this case, the carpet which forms the background of the sorry scene which you perceive.

    What then does recognizing sensory information as a sensation require? Having a sensation requires being aware or knowing that one is having a sensation in at least the attenuated sense that one distinguishes the sensation from other things. To meet this minimal requirement, one need not have a concept of this or that or the other sensation, one need not have a concept of "red" or "acrid" or "shrill". But one does need to be able to distinguish the sensory state of oneself from other things, namely states of the world outside oneself.

    Ordinarily, we perceive objects in the world, seeing through the internal representations, much as one sees through a window without noticing it. In seeing a tree through a window without noticing the window, one does not perceive the window. To do so requires a shift of attention. And it seems that sensation-having is like perception in this regard. Although sensory information exists and is used in perception or sensing, if it itself is unperceived, it does not constitute a sensation. In perceiving a tree without noticing that in doing so certain information is present as an internal modification in oneself, one does not perceive the internal modification or state of oneself. In such cases one is lost in the world and does not perceive one's own states; one does not have a sensation. Sensations are had only when there is a shift of attention to one's own states. Then the sensory information assumes quite a new status. From being merely the means of sensing, it becomes an object for one in its own right. Then one can say "this is happening to me, I am having a sensation."

    One may later be able to recall the sensory data one had while one was "lost in the world", and it may seem as though one had sensations throughout. But this is an illusion. Sensations exist only when they are noticed; their esse is percipi. And noticing a sensation is different from noticing its cause: noticing a sensation is noticing one's own state, a state which typically, but not always, has a proximate external cause. As I write this, I sometimes have sensations of the paper and sometimes do not. When I am concentrating completely on what I am writing, I do not. But I can make very rapid shifts and note that I am seeing the red ink appear on the yellow paper as I move the pen. I make this shift very often just now because of the content of what I have just written; it makes me self-conscious. (The availability to me of eidetic information from short term visual and auditory memory may mislead us here; it may make it seem as though we had sensations when we did not at the time the information arrived from the senses.) But this is not typical of my dealing with the world; sensations are rare, and not, as the British Empiricists thought, the building blocks of all mental life.
Moreover, apparently there is individual variation in the extent of sensation having. As K. A. Mathews, et al., write:

People seem to be cognizant of their internal bodily sensations only to the extent that their attention is focussed inward onto those states. Moreover, this conclusion appears to be valid across a wide variety of different contexts. That is, similar effects were obtained whether self-attention was varied experimentally by the presence or absence of a mirror, or was varied dispositionally by persons' propensities to be high or low in self-consciousness. (pp. 177-178)6

Interesting, the same results were suggested with regard to the experience of emotion:

...following the lead of Schachter and Singer..., a number of personality and social psychologists have come to view the experience of emotion as being strongly influenced by three elements: The existence of some perceptible internal state that differs from one's baseline level; the focussing of sufficient attention on that internal state to result in awareness of its existence; and the use of some knowledge structure to interpret the state. (p. 179)

This suggests that lower animals will not experience emotion as we do.

    An antecedent of the argument that sentience requires self-awareness can be found in Kant's first Critique. Kant holds that at least the possibility of self-awareness is a necessary condition of my having sensations:

It must be possible for the "I think" to accompany all my representations; for otherwise something would be represented in me which could not be thought at all, and that is equivalent to saying that the representation would be impossible, or at least would be nothing to me. (B131-132)

    Although it remains unclear to me just exactly what Kant's claim is here, it seems to be at least related to the thesis I have argued for above. But Kant goes on in the very next sentence to say "[t]hat representation which can be given prior to all thought is entitled intuition", and I am arguing that this thesis, shared in one form or another with the British Empiricists, is mistaken, in that thought, although not necessarily conscious, is required for the very existence of a sensation, which I take to be what Kant calls, in humans, an "intuition". In any case, I believe that Kant's purposes are not my own, as he is explicitly concerned with personal identity, the unity of consciousness, and not with the nature of sensation.

    And, unfortunately, Kant does not seem to have an argument for his thesis. Jonathan Bennett remarks that Kant suggests that he armed with an account of awareness, another of self-awareness, and an argument which links the two. This would be formidable equipment indeed, but Kant does not have it; and that is not surprising, for there seem to be no good reasons for saying that a dog's visual field, say, is "nothing to" the dog, or in general for saying that where there is consciousness there must be self-consciousness. (Kant's Analytic, Cambridge U.P., 1966, p. 105).

Here I am trying to provide reasons for saying, not that a dog's visual field is "nothing to" the dog, for it determines the dogs behavior, but that it probably cannot be appreciated by the dog as sensation. Perhaps that is what Kant meant.

    Unfortunately, most all animals appear to lack the requisite self-awareness. Griffin (op.cit) cites work by G.C.Gallup which suggests, on the basis of differential response to their images in mirrors, that Chimpanzees are self-aware. Griffin remarks:

So far, Gallup's type of experiment has yielded positive results only with Great Apes. Despite intensive efforts, gibbons, monkeys, and other laboratory animals have failed to react to mirror images as replicas of their own bodies. Instead, they seem to treat the mirror image as though it were another animal. Gallup therefore concludes that no other animal has the capacity for self-awareness.

Therefore my negative conclusions regarding "lower animals" should not be taken to apply to these higher anthropoid apes.

    Now it seems to me that it is theoretically possible to build a machine which is capable of treating sensory information as states of itself, and so which, on this account, have sensations (contra Descartes). Since I identify sensation having with a certain kind of information processing, I am bound to hold that where that processing occurs there will be sentience, no matter what the physical stuff of which the information processor is made. But it does not seem that subhuman animals are such self-aware systems. The form of reflexive awareness which I am arguing is requisite for sensation-having does have biological advantages, as is demonstrated by our success, but it is not necessary for the activities of lower animals. It comes with the awareness of self needed for complex social relations, making the appearance/reality distinction, and, related to the latter, doing science, constructing a model of the world, a model which includes the knower-perceiver. For lower animals, as for existing machines, environmental conditions can be sensed and responses made without any awareness or representation of the intervening sensory information as information possessed by the system.

    So here the analysis of sensation which reveals a necessary capability of distinguishing states of oneself from states of the world, and the functionalist considerations of the complexity of response possibility which is requisite for sensation-having, come together. For distinguishing states of oneself from those of the world is a particular complex response needed for many complex overt and covert responses or discriminations, such as recognizing an illusion as an illusion, or understanding how it is that saccharin can taste sweet. Qualia have a function, but one that has a place only in the most cognitively capable organisms and artificial information processing systems.