Do words need definitions?

The naive philosophical position is that in order to evaluate the truth of any proposition, one must know the definitions of all of the words in that proposition.  This might be called the "reductive" theory of truth.

But this is quite wrongheaded.

Why is it wrongheaded?  Because it leads to a kind of infinite regress.  Suppose you have a statement p, which is made of words.  What do these words mean?  They have definitions - but those definitions are also made of words.  So in order to evaluate whether p is true, you now have to know the definitions of all the words in p, plus the words in the definitions of the words in p - call that p'.  But all of those words have definitions as well, which means one would have to know the definitions of the words in p, plus the words in p', plus all the words in the definitions of the words in p' - call that p''.  And one can continue expanding this indefinitely, to p''', p'''', and so on.  Eventually one of three things will happen.  Either (1) you will find a loop, in which one of the later definitions involves a word that was defined in one of the earlier definitions, creating a circular meaning. which never comes to a final ending, or (2) you come to words that are not defined by other words, or (3) you will come to words whose definitions you do not know and have no way of finding out.  (In practice, all three will happen, rather quickly.)  But according to this naive, "reductive" theory of meaning, you cannot know the truth-value of any proposition without knowing the definition of all of the words in that proposition.  Does this mean that no one can ever know if anything is true?  Does this doom us to perpetual meaninglessness?  

No.  There is still hope.

Consider a statement of the form "All fneebs are glerbs."

According to the naive, "reductive" theory of truth, one would have to know the definitions of all of the terms in that proposition, including the words "fneebs" and "glerbs" in order to determine that the proposition is either true or false.  That is, only by fully and exhustively understanding the definitions of "fneebs" and "glerbs," can one determine if fneebness entails glerbness, and thus evaluate whether it is true that all fneebs are glerbs.

But this isn't necessarily so.  For instance, one may be able to determine that the statement "All fneebs are glerbs" is false, without knowing precisely what "fneebs" or "glerbs" are - by finding an example of a fneeb that is not a glerb.  Thus there is a possible way forward for meaning: the "via negativa."

One might object, at this point: but how can we know that we have an example of fneeb that is not glerb, if we don't know what "fneeb" and "glerb" even mean?

Consider, for instance, the sentence, "All time is relative."

I may not be able to give you a definition of the word "time" that is fully satisfactory; nor, perhaps, could I give you an exhaustive definition of what it means to be "relative".  But if I could give an example of a time that is not relative, that would disprove the statement nonetheless.

I don't know if I can precisely, exhaustively define "time," but I know enough about time to tell you some things that time isn't.  For instance, time is not a sandwich.  Time is not confined to the interior of King Charles's spleen.  Time is not always 5:00 pm.  And so on.  Similarly, I might not know what "relative" means, exactly, but I can say that if a property is relative, then that property is not necessarily the same in every reference frame.

Another example: if a person says, "All men are healthy," I may not be able to define "man" (or "men"), and similarly I might not be able to define "health" (or "healthy"), and yet if I can produce an example of a sick man, I have disproved the proposition.

This, of course, leaves us open to the problem known colloquially as "No True Scotsman".  According to the old joke, one person (presumably a Scotsman) asserted that "No Scotsman puts sugar on his porridge," to which his friend replied that his Uncle Angus puts sugar on his porridge, leading the original speaker to reply, "Yes, but no true Scotsman puts sugar on his porridge."

Going back to our earlier example: if person A says, "all fneebs are glerbs," and person B, who doesn't know exactly what the words "fneeb" and "glerb" mean, points to a fneeb that is not a glerb, [NOTE: in this example, B is correct: this truly is a fneeb, and it is truly not a glerb], that would prove person A wrong.  Nonetheless, person A might reply: "You don't know what a fneeb is.  You don't know the definition of the word 'fneeb'.  What you are pointing at is not a fneeb."  

How can B respond to A?  It's not easy.  It's annoying and difficult.  

Word definitions are stipulative, that is to say, in a certain precise sense, "arbitrary" - but that does not mean that they are random or infinitely unlimited in their meaning.  There are limits to what we can mean with words.  These limits are generally conventional.  These conventions can change - we can arbitrate new rules in the usage of words - but not without cost.

One way of anchoring meaning is ostension.  That is, for example, B can declare, "When I say 'fneeb' I mean that thing over there," and point at it.  Notice that this is not a definition.  Presumably, there are fneebs other than this particular fneeb that B is pointing at, and perhaps they all share some characteristic or similarity that we might call "fneebness."  But none of that has been determined within this particular conversation.

Now A may still respond, "Well, when I say 'fneeb,' I don't mean that thing over there."  And now the conversation has reached a kind of impasse.  A may acquiesce to B's definitions, or may simply try to change the subject and not speak of fneebs anymore.

But there are other possibilities.

At this point, B can interrogate A, and ask A questions about A's concept of fneeb, in such a way that leads A to realize that A's concept of fneeb is somehow inconsistent or wrongheaded.  From there, A or B my propose alternate definitions of fneeb, ever refining their understanding and getting closer and closer to a language that describes the truth.  In short, B can now engage in philosophy.  And, perhaps, eventually, through their conversation, A and B can work their way to first principles.  Possibly.

But it's crucial to understand that it doesn't start there.  The starting point of language is not clearly defined terms.  Language can "work" before anyone does any philosophizing about it.  Before philosophizing, humans may use language is a culturally contingent, somewhat arbitrary, and even quite inconsistent way.

Better yet, A and B can propose hypothetical models of reality that involve different kinds of definitions of fneeb, and lead them to predictions that can be tested empirically.  And if the models fail these predictive tests, they can propose other, better models, where fneeb is defined differently.  That is, they can engage in science.

But language doesn't start there either.  Science does not precede language, either temporally or logically.

Language begins, and functions quite well, without clearly defined terms.  And though we may gradually refine these meanings, there is no guarantee that we will ever arrive at fully consistent definitions for the terms we use.