As a student of language, I am a strong proponent of etymological studies, since language obviously evolves over time (as I discussed briefly here and here) and I am also a strong believer in George Santayana’s famous quote that “Those who do not learn from history are condemned to repeat it.” I think a careful knowledge of language involves understanding how words have come to mean what they do for us, just as a careful knowledge of science involves understanding how scientific concepts and theories have changed over time (for instance, like the change in the concept of combustion from phlogiston theory to oxygen theory or from Aristotelian to Ptolemaic to Copernican cosmology).
However, knowledge of etymology should not make one take any liberties with that knowledge, as so often tends to be the case with humans. (As Alexander Pope said, “A little learning is a dangerous thing.”)
What I am referring to is what could be (and has been) referred to as the argument from etymology (or argumentum ad etymologia if you prefer a made-up Latin phrase, made up by yours truly [as far as I know]). The Language Log gives a useful format that is often given for this argument:
In <language X>, the word for <concept Y> is based on the word for <concept Z> (or perhaps, a combination of the words for <concept Z1> and <concept Z2>). Therefore, in order to understand <concept Y>, you should think in terms of <concept Z>, recognizing the deep traditional wisdom inherent in the lexicographic history of <language X>.
This is good, but for the purposes of this entry, let me propose a modified version:
In <speaker’s native language>, the word for <concept Y> is derived from the word for <concept Z> (or perhaps, a combination of the words for <concept Z1> and <concept Z2>) in <language X>. Therefore, in order to understand <concept Y>, you should think in terms of <concept Z>, recognizing the deep traditional wisdom inherent in the lexicographic history of <language X>.
I’ll even self-analyze and propose a formalization of a comment I made about the word confidence:
In English, the word confidence is derived from the Latin root fides, ‘faith.’ Therefore, in order to understand the concept of confidence, you should think in terms of faith, etc.
Although I didn’t (and wouldn’t) make this argument – in fact, I suggested that we should think of faith in a specific context as being similar to the normative usage of “confidence” – it could be made when trying to force different usage of a word into a specific, decontextualized definition. There is one big problem with this already, but as the practice of logic goes, let’s consider a counterargument:
In English, the word nice is derived from the Latin root nescius, “ignorant” (lit. “not-knowing”). Therefore, in order to understand the concept of being nice, you should think of terms of ignorance, etc.
The problem in this specific case is that the English nice no longer has any connection to ignorance, and in fact, we see the evolution of the word steadily moving away from that meaning. When the modern word loses the connective strand of meaning from the original, we no longer have any reason to appeal to the word’s origin to help determine its meaning. (Sidenote: Credit for bringing this counterargument to my attention belongs to William Lane Craig and Paul Copan in their book Creation Out of Nothing: A Biblical, Philosophical, and Scientific Exploration.)
The big problem, however, is that appeals to etymology are used rhetorically in many cases to lend an air of prestige or wisdom (as implied by the end of the Language Log’s format above) – it’s that whole tendency we have to assign a sense of superiority to things that are old. (In the realm of ideas, old ideas can be good in that there has been plenty of time to critique them, and so only the best ideas stand, but that truly has nothing to do with their age, just their longevity.) In reality, this is not linguistically valid. Appeals to etymology are really only good for questions of etymology, which rules them out as legitimate ways of determining the meaning of words. As I have discussed, language is not prescriptive (and the tendency to think this is so tempting that even a Language Log contributor slips up by suggesting that “anti-Semitism ought to mean ‘hatred of Semites'” [emphasis not added] even while affirming that the most common usage refers only to Jews), and so we cannot go back and look at the history of a word to force a previous meaning onto the derived word retroactively. To repeat a previous sentiment, that would be linguistic tyranny, but in this case, it is also poor logic. Because of the way language changes, one is lucky if the word bears any resemblance to its origin, and certainly the latter will never be a firm help in determining usage.