Understanding "Is" and "Ought"

billharms@billharms.com 8/1/2000

For some time, I have been interested in the claim that a scientific understanding of human knowledge leaves something very essential out of the picture, namely, the source of the standards or rules of rational thinking and proper evidential support. The reason a merely scientific understanding of human knowledge fails to capture the source of these rules is that science can describe how people reason and form their beliefs, but not how they ought to do so. These rules and standards are an inseparable part of our traditional philosophical understanding of  knowledge. To qualify as knolwedge a belief needs to be True, and not just accidentally so. It needs to be Justified as well. Such value-laden normative terms as "truth" and "justification" seem to be outside the realm of scientific enquiry. This is, by the way, the same claim according to which morality  is supposed to be fundamentally outside the realm of science.

The reason that I became interested in this claim is that I think it is wrong. The reason I think it is wrong is because as a graduate student I spent some time immersing myself in Ruth Millikan's   Language, Thought, and Other Biological Categories (MIT Press, 1984) and, though it is not a point on her agenda, I began to see how viewing "ought" language in terms of biological functions allows us to understand its relation to "is" language, rather than just recognizing that they are different.   It allows us to understand what would make an "ought"-statement true, and why this truth has different kinds of direct implications than the truth of "is"-statements.

The word is not the basic unit of meaning

As far as I can make out, the reason that the relation between "is" and "ought" is such a mystery is that people think that the word is the basic unit of meaning. The confusion unfolds as follows. Human languages differ enormously on the surface, but underneath are certain very consistent features. Every human language is composed of words which are assembled into sentences according to the syntactic rules for sentence formation of the language. Every human language has a basic subject-predicate strucutre, which is to say, that what (indicative) sentences do is to attribute properties to objects. Subjects and predicates are not true or false. It is not until they are put together into a complete sentence that truth and falsity become possibilities. Along with the indicative statement of fact, the conjuction of subjects and predicates can also be used to perform other kinds of speech acts such as commands and questions. On this view, words are the "atoms" of meaning and truth is something that emerges as a possiblity when the name of a thing is conjoined in an indicative sentence with the name of a property. More or less.

Within this kind of a framework, and this is the basic framework throughout linguistics and mathematical logic, "ought"-statements present one with a puzzle. Take "Murder is wrong" for instance. On the surface, the sentence seems to be attributing to the act of murder the property of wrongness, just as "Murder is profitable" attributes the property of profitableness to the act of murder. In the case of profitableness, we can figure out what kind of property this is - some kind of shorthand for the likelihood of your getting paid for it. Having unpacked what "profitable" means, we can proceed to determine whether or not "Murder is profitable" is true or not, or how often it is true. But in the case of "Murder is wrong", it is not at all clear what sort of property "wrongness" is. Instead, the sentence seems to be a sort of disguised command, like "Don't murder!" But commands are neither true nor false, so if "Murder is wrong" is a disguised command, then it is merely masquerading as a statement of fact in an attempt to get you to obey. There can be no sense in which it is true that murder is wrong, or that you ought not to murder.

In order to reject this uncomfortable conclusion, many have been led to postulate that wrongness is after all a property, and "Murder is wrong" is true just in case murder has the property of wrongness. But since wrongness is not any kind of property that science can identify, it must be a "non-natural" property. Now apart from the fact that "non-natural" properties are very mysterious, they have also become by definition off limits to scientific enquiry. And if this is true of "x is wrong", it is also true of the oughts of knowledge:  "x is justified", "x is rational", and "x is true." So not only can't you get an "ought" from an "is", but apparently science cannot even help us understand what an "ought" is!

But, as I said, the word is not the basic unit of meaning, which is quite evident once you realize that a word by itself has no job to do. (The name of a thing does not necesarily indicate the presence of a thing, since it may be used in a sentence to say that the thing is not present.) For those of us who like the thought that human beings are not unique beings wholly unlike the rest of our animal cousins this is good news. For it has become quite clear that animal "languages" if there be such do not have combinitorial syntax (i.e. combining meaningful elements into truth-bearing wholes) but allow communication  in a somewhat simpler and more direct fashion. But since it turns out that that basic (functional) unit of meaning is a "complete representation" like a sentence, then it also turns out that animals can say true things to each other, and perhaps we can learn to understand them as well.

Focusing on words and their combinations corners one into assuming that truth only works one way, via the correct attribution of properties to objects in indicative sentences or their like. This creates the mystery with "ought"-statements. Focusing on the complete representation from a functional point of view opens new possibilities. The place to start is with monolithic signals in simple signaling systems. Take traffic lights. Red and green have something of the nature of commands. Red says "stop" and green says "go." But they are not simply commands, since they also purport to indicate something about the world. Green also says that it is safe to go, or at least that the opposing lights are red. Yellow says slow down or proceed with caution, but also that  the light is about to turn red. So there is a sense in which the green and yellow light can be false, if the things they purport to indicate do not in fact obtain. Animal signaling systems usually work like the traffic lights. The beaver slaps its tail when it perceives danger. The tail slap means that danger is present, but it also  commands that the other beavers run and hide. Which is to say, it is not like our indicative sentence because it commands. It is not like our imperative because there are conventions regarding when the slap is supposed to occur, according to which it can be false. What it is like is what our "ought"-statements seem to be: commands that purport to be true.

The preoccupation with combinatorial syntax results in a theory of language in which referring words always connect to the world by attributing properties to objects. The different modes, indicative, imperative, inquisitive, then indicate the attitude the individual wishes to express toward the state of affairs indicated. To claim that it obtains, to request that it be brought about, to inquire as to whether it obtains. Surely human beings do use words in all these ways. But what traffic lights and tail slaps point out is that this is not the only way signals work. It is not even the usual way. Instead, the role of a signal in a signaling system determines (1) what state of affairs it is supposed to coincide with, and (2) what behavior is supposed to follow from its issuance. Failure of (1) is falsehood. Failure of (2) is the sort of thing we usually call failure of "comprehension." Truth in general does not require the attribution of properties to objects (not to mention the indicative mode). Truth simply requires conventions regarding the tracking behavior of the signal. If there is a rule that says what external state the signal is supposed to coincide with (like danger) then the signal can be true and false. The insistence on subject-predicate structure leaves one rather limited in how one can understand signals. They can have truth-value or they can command, but not both. Simple signaling systems typically do not have subject-predicate structure, and they can have both truth-value and imperative consequences, just as normative statements seem to. And, they can do this without postulating "non-natural properties."

Teleosemantics

Teleosemantics is a theory of meaning pioneered by Ruth Millikan which focuses on the function of language, or better yet, signaling systems, rather than on combinatorial syntax and semantics. The central idea is that true signals fulfill tracking functions, and that correct "understanding" is a matter of responding correctly. The complications of human langauge do get accounted for eventually, and they may indeed be unique and set us apart from the rest of the animal kindom. But the way in which our language sets us apart is not by making us uniquely able to "understand" the world, but by making us at once capable of thinking about the meaning of signals sent by other species and  yet predisposed to misunderstand the basic nature of meaning, to the extent that we often doubt that other animals have meaningful thoughts at all, just because it seem implausible that they are having the kinds of thoughts we have.

"Function," of course, is a quasi-normative term in its own right, and it needs to be in order to provide the basis for a semantic theory. This is because a theory of meaning requires a source of standards which can fail. Signals are false if their tracking rules are not satisfied. Signals are "misunderstood" if the correct consequences do not follow. The notion of a function provides these sorts of standards. If it is the function of the beaver tail-slap system to keep beavers safe from predators by slapping when predators are near and hiding when slaps are heard, then the slap is false when predators are not near and misunderstood (or ignored) if the other beavers do not hide. In order to avoid "smuggling" in the normativity, one needs an account of function.

Millikan prefers an evolutionary account of function. In simple hard-wired animal systems, the adaptive history of variation and selection on the genetic basis of the signaling system specifies its function. Beaver tail-slapping presumably has a long and successful history, and it is success or failure in performing up to those historical standards that provide the fallible conventions necessary for meaning. The tail slap is true now if these are just the sort of situations in which beaver tail-slaps have made adaptive contributions in the past. What the other beavers are supposed to do on hearing the slap is just what they have done in the past upon hearing them that has had adaptive value. This is not the same as the state that usually obtained, or the behavior that usually followed, but the states and behaviors which when coordinated resulted in increases in reproductive fitness.

It is imporatant to note that, while Millikan prefers the evolutionary account of function (as do I), other accounts may serve just as well. For the theist, for instance, if God intended that the signals occur at certain times and that they be responded to in certain ways, then God's historical intentions might provide fallible standards for truth and consequence. Similarly, if a signaling arrangment stabilizes culturally due to a productive coordination of circumstance and behavior, this provides historically rooted stardards with respect to which failure can be judged. People are often critical of Millikan's work (unfairly, I would say) because it seems too dependent on adaptive histories, so it is important to realize that any specification of fallible function can serve as the basis for teleosemantic meaning.

The focus on function as opposed to combinatorial syntax emphasizes that tracking conventions and response conventions, truth and consequence, are not always paired as they are in our own peculiar subject-predicate way of talking, but can be conjoined in any way that serves a function. The ways in which truth-conditions (when the signal is supposed to occur) and meaning-consequence (what is supposed to happen in response) are combined in a single signal is as varied as the purposes that a signaling system can serve. Truth and consequence-meaning are not eternal things, they are a matter of whether or not a job is done in the way it is supposed to get done.

The function of normative language

The relationship between our scientific/indicative and normative languages is something like the relationship between our scientific/indicative language and the languages of our animal cousins, with the exception that in our case signaling systems with different functions are implemented in the same brains. Strictly speaking, different functions are served by indicative and normative language. (I'm simplifying madly here. My guess is that there are lots of different signaling systems with lots and lots of different functions.) The function of factual descriptions of the world is to simply indicate current states. The function of  normative language is to regulate. Of course, all this would be much clearer if it were not for the fact that evolving systems are opportunistic about what resources they use so that both our indicating system and our regulating system make use of the same linguistic resources in terms of subject-predicate reference to the world.

So if we stop being hung up on words as the basic units of meaning and the subject-predicate conjuction as the unique way of saying something meaningful, and realize that meaining is a matter of function and function is a matter of history and truth and consequence get combined in all kinds of ways depending on the function of signaling systems, then we can understand the relationship between "is"-statements and "ought"-statements. The function of "ought"-statements is obviously to regulate behavior. In the case of morality the objective is something like social harmony. In the case of the norms of rational thought, logic and good evidence, the function is the effective management of a very complex envronmental tracking system. But whatever is doing the tracking,

it is the same world that is being tracked.

So there are no "non-natural properties", there are just whatever states of this world are the states when the signals involved are supposed to be sent. What is supposed to follow from a signal depends not on what state of the world makes the signal true, but what fulfills the historically determined function of the signaling system. This means, and this is the really important bit, that

just because two signals have the same truth conditions does not mean they have the same consequence-meanings.

Why is this such a big deal? Because that's the difference between "is" and "ought." It is not that "is" and "ought" refer to different properties in the same old way, but that they refer to the same (physical/historical) properties in the fulfillment of different functions. "Is" statements track the world in order to get us coordinated with that world in the service of whatever desires we happen to have. Ought statements track the world in order to get us very directly to do certain things at certain times, regardless of whatever desires we have. "Is"-statements are value neutral, without the usual imperative component, because they are designed to serve any desire. "Ought"-statements include the (usual) imperative component because they have much more specific tasks to accomplish. But both sorts of statements are true or false for pretty much the same sorts of reasons - because they do or do not get the timing right with respect to the varying world and their respective functions.

An antidote to relativism

What this means in general is that a scientific understanding of knowledge does not have to leave the normative elements of knowledge, truth and justification, a mystery. Teleosemantics gives us truth for all kinds of statements, including evaluative statements regarding whether or not beliefs are justified, whether people are irrational, or whether theories are poorly supported. An added bonus is that we get an account of the truth of moral statements as well, at least if morality has a function. Statements are true if they get the timing right with respect to their function. Statements are understood (correctly interpreted) if you get the response right with respect to their function.    That's really all there is to it (apart from the gory details). If you really want to know how names end up referring to objects (and what sorts of objects they refer to), where combinatorial syntax comes from and all of that, go read Millikan (1984). I really do think she has most of it right, and not nearly enough people are paying attention.

Does this mean that science will be able to tell us something about right and wrong, rational and irrational, that we don't already know? I doubt it. The reason why I doubt it is that in order to make specific claims about the functional semantics of a language you need to know a lot about the functioning history of a signaling system, and unfortunately, complex social behavior tends not to leave many fossils. So if you want to be reasurred that murder is wrong on the basis of a fact-based teleosemantic analysis you will probably be disappointed. Presumably, there is some social instinct, some behavioral control circuit (apologies to any cringing neurosceintists) which was selected for generating a certain kind of outrage in circumstances of within-group killing, and we grope to express this in the only language we have, in the manner we have observed in others. But whether there is such a "circuit" is certainly beyond current neuroscience, and if there is one, what its adaptive history is will probably never be known. Actually, the very best information we have regarding the existence and function of such a alleged "circuit" is the fact that just about everybody thinks murder is wrong, and that this seems to go beyond mere cultural conditioning.

And who really does want science to start preaching to us? What is useful and interersting here is that we no longer have to put  up with unrestricted cultural constructivism - the idea that standards of morality and rationality are pure inventions of culture and that cultural standards are uncriticizable from outside. Whatever common social circuitry we share based on our common biological ancestry brings with it standards of evolutionary design. And when we express those standards they give meaning to our words beyond the surface content borrowed from the spoken language we are forced to use. Claims of wrongness and irrationality are true for us. They are true for "them" too. There is little room for dogmatism here, in that we can never be completely sure that we have got it right. So much the better, I say.  But there is good reason to believe that we are not just making it all up, that there is after all a fact of the matter.