We accept two basic principles about language. One is that words don’t magically correlate with things in nature. The only reason “hydrogen” (the word) means hydrogen (the substance) is because people use “hydrogen” to designate hydrogen. It is not a necessary fact that “hydrogen” means hydrogen, it could mean something completely different to a different group of people. Second is the premise that words really do have meanings. Although people in a community may use a word inconsistently we say that there is some meaning that best fits the usage of the linguistic community, and that it is what the word really means to those people. Of course the exact nature of the second claims turns on exactly what a meaning is.
The first claim doesn’t really need defending. Simply observing that the meanings of words have changed over time, and that different words can have the same meaning in different linguistic communities should be enough evidence. And because we have ruled out some magical unchanging force as the provider of meanings for words the only thing we have left are people. Thus we conclude that how the people of a community use words is what gives them their meanings. This is not to say necessarily that all there is to the meaning is a pattern of use, although some would indeed claim that.
And so we now turn to the second claim, that words do have meanings. Actually to be more specific we should say that words in the context of linguistic community that uses them in a consistent manner have meanings, but I won’t repeat that mouthful again, since it follows directly from the first claim. The only evidence we need to support this claim is the observation that people are able to successfully communicate with each other. If I intend for you to pick up the hose, and I say “pick up the hose”, and you do indeed pick up the hose we can conclude that my intended meaning was successfully conveyed to you. If we rule out ESP the only possible way this could have happened is through the words, and thus we conclude that my words conveyed the meaning to you. Certainly this is enough to conclude that most common words have meanings; can we extend this conclusion to all words? Probably not. Some words, such as those used by postmodernist authors, don’t seem to have meanings, since people exposed to them often can’t agree on what is really being said. But this is not a problem for the second claim, since in its extended version we stated that the words must be in the context of a linguistic community that uses them. Some of the terms that postmodernists use don’t have meaning specifically because the linguistic community that is trying to understand them doesn’t use them and because even among postmodernists they aren’t used consistently. So the second claim seems to hold water.
But what exactly is a meaning? There are many ways to approach what a meaning is, but one of the most general was proposed by Frege, who said that a meaning could be given by the truth conditions (over all possible worlds). Here I will work with this definition of meaning, since most definitions of meaning agree with Frege’s at least in the vast majority of cases (in the sense that they claim that the same words have meaning). Now, if we are to give truth conditions for any word it must be done using only physical properties, since there are no non-physical properties (I assume that we are materialists). For example, we might define some adjective X by stating that something is X only when certain physical conditions hold.* In other words, the meaning of a word is equivalent to some o-physical term, or natural kind.
But people, often being ignorant of the underlying physical reality, don’t always use words in a way that is consistent with any natural kind or o-physical term. This means that it is possible to investigate a term and find out what it “really means” (or that it doesn’t mean anything at all), even though that very meaning was defined in part by your usage of the word. How this is possible is illustrated best by simple terms. Simple terms are easiest to study because they have the additional property of being used in such a way that they are assumed to describe a single kind of thing. For example “jade” describes a single kind of mineral, “brown bear” a single species of animal, ect. We can then investigate a word such as “jade” and find out what kind of physical properties of minerals make a rock “jade” instead of something else. And in the case of jade we jus this kind of investigation discovered that there was not one but two kinds of minerals that we commonly call jade, nephrite and jadeite. But since jade is a simple concept (or at least was a simple concept) we can conclude that there really is no such thing as jade, there are really two kinds of minerals. Our investigations revealed, surprisingly, that jade, defined as a single kind of mineral, is meaningless. Fortunately for language most simple terms really do refer to a single kind of thing, “gold” is gold, ect. Why should we think that such investigations can reveal the real meaning of words and that we should modify our usage in light of them? Well, because if we don’t our words will, in some cases, be meaningless, meaning that there will be no physical reason for X to apply in those cases, which in turn means that it would be impossible to communicate clearly and efficiently about X in those cases, obviously something undesirable.
* Some may object, saying this isn’t true for mathematical concepts. However, I would argue that mathematical concepts, such as numbers, can ultimately be defined in terms of the physical, which in the case of numbers would be the cardinality of groups of physical objects, or in terms of other simpler mathematical constructs, but in either case that they are ultimately defined in terms of the physical world. Similarly other abstract concepts, such as the ones employed in this very discussion, can be defined in terms of simpler and simpler concepts, until we reach those that can be defined in terms of physical properties, or at least so I would claim.