There was a question in class whether or not language could construct knowledge or merely convey it, and while the crux of what simmered by the casual toss of semantics had to deal with "the chicken or the egg" paradox (whether or not to consider the signifieds as empirically "prior" to the application of sound images), there is perhaps a possible "portmanteau" of text that might help answer that question by example.
Locke and Richards and Ogden agree that language has no "standard in nature," rather that they gain meaning through sedimentation (something Campbell said?) and for Richards and Ogden, "It is only when a thinker makes use of them [words] that they should stand for anything, or have 'meaning' (Richards and Ogden 1274). In this way, this new line of thought runs counterpoint to the metaphor of language as vessel of meaning (Lakoff and Johnson). It also seems to lean towards way language as incompetent; unable to promulgate knowledge. Yet, for Locke, this predicament may lay inlaid in the discourse of philosophy, of which he bundles in practice such as science, and law. To use a somewhat flimsy example, take the concept of murder.
Clearly, we know what this is. By law, it could be called homicide, or suicide, or manslaughter, each a conceptual sub/re/classification of a socially engendered concept:murder. We can say that a man murdered his neighbor, but we cannot say that a bear murdered its neighbor. If, in fact, a bear did kill something near him, perhaps a wolf, we would instead call that "the circle of life," ultimately putting it metaphorically out of our reach of conception. Even if a bear is to kill a human, we do not say that the man was murdered by a bear, we instead say that he was "killed." The action however, to kill: death set in motion by outside action, is ultimately the same. It can be seen then that language can indeed construct meaning, in this instance, not only referring to the act of killing, but its social relation, its intended effect, the personal state of the murderer, and its negative qualities that link and separate it from its other modes of numeration, similar to what Richards and Ogden drew up in their triangle of meaning (Richards and Ogden 1275).
So, language can construct a knowledge, and namely, one that does not exist in nature. Granted, this new construction is not independent, and nor does something "new" have to be considered as such, for even if its substratum lay in the root or proliferation of related significations, the manipulation, or uniqueness in application is sufficient warrant to warrant its authenticity as "original." The real question, using Derrida, is how many words, related and extant based in opposition can there be before the root meaning, or origin is entirely lost?