There are multiple means by which languages acquire new words, a process called "word formation." We borrow words from other languages (for example, "amok" from Indonesian or "manga" from Japanese). We combine words and sometimes then streamline the resulting combination (for example, "Internet" and ultimately "the Net"). We turn acronyms into words (for example, "FUBAR", from "Fucked Up Beyond All Repair."). We attach new meanings to old words (for example, "gay" to mean homosexual, "hacking" to mean computer mischief, or "frag" to mean killing a superior office (derived from fragmentation grenade, a tool used at the time the term was coined to carry out the deed of killing a superior officer to which it refers)).
Sometimes we create new words entirely from whole cloth, as subtype of neologism. Heinlein introduced "grok" meaning "to intuitively and fully understand" in one of his books. The Battlestar Galactic screenwriters invented "frak", a swear word comparable in meaning to "fuck" (although arguably derive of it in much the way that "dang" is derivative of "damn"). The term "robots" was "introduced to the public by Czech writer Karel Čapek in his play R.U.R. (Rossum's Universal Robots), which premiered in 1921." Lewis Carroll also created many natural sounding words of this kind, although few have entered general usage. J.K. Rawling brought "muggle" into the lexicon and the term is now used in a more general sense than it was in her books in a sense comparable to "layman" or "unenlightened" or "ordinary people". But, the simple, created out of whole cloth new root words are a pretty small part of the total additions to the lexicon in any given decade. Under what conditions does this happen?
The examples I've provided were creations of fiction authors whose works had loyal followings and who were intentionally trying to create an exotic feel of mild foreigness. They also followed linguistic rules of English related to joining phonemes into words (i.e. they are initially logatomes). They sound like natural simple Anglo-Saxon words.
Most new words also seem to arise in subcultures where they can be used in conversation among a smaller group of people more likely to know the novel word and with a social incentive to identify themselves with the group of people who know and use the word, before being disseminated into the larger culture, rather than arising diffusely.
Of course, there are other ways that words arive. "Googol" was a number named "in 1938 by 9-year-old Milton Sirotta (1929–1981), nephew of American mathematician Edward Kasner. Kasner popularized the concept in his book Mathematics and the Imagination (1940)." Quark was a nonce word used by James Joyce in Finnegan's Wake before it was adopted by Murray Gell-Mann is a physics term. The term meme was coined by Richard Dawkins (in analogy to "gene" from the ancient Greek mimeme) in his book, The Selfish Gene. Indeed, the sciences may be a particularly fruitful area for new word creation.
How often do these kinds of words enter the general lexicon in absolute terms and relative to other kinds of words? Do they arise in situations other than intentional speculative fiction, and if so when? Do they have any more or less staying power in the language than other new words? How long does it take for the people who use these words to cease to know their origins? Do these words generally refer to concepts that could previously be described only with multiple words or not at all, or do they generally replicate the meaning of single words that already exist? Is there anything inherent in the word itself that distinguishes those that enter general use after an attempt is made to introduce them from those that don't?
A related issue is how often new concepts that aren't expressed by any single word in our language (also known as lexical gaps) arise, and how common are they?
There are certainly cases where this is possible, even a mature language. New things are invented that must have nouns to describe them. French has the word "chez" which means "the home of" and has parallels in other romance languages (e.g. Spanish "casa", that is absent in simple one word form from English. English has stumbles over a gender neutral term third person singular pronoun referring to a human being equivalent to "him" or "her" although there are competing proposals for ways of dealing with it (using the word "one" for that purpose has been proposed at least since 1770 by Robert Baker, but has never gained universal acceptance in day to day language since it comes across as too formal, while using the word "they" in the singular has fairly wide acceptance in casual speech but is usually considered a grammatical error in formal speech), and also lacks a consensus way across dialects to distinguish between the second person singular ("you") and the second person plural ("you guys" or "y'all"). English lacks a single word that captures in a gender neutral way the phrase "significant other" although "partner" is increasingly being appropriated for that purpose. Why don't new words swiftly get invented and widely adopted even when the need for them is widely known? How many concepts are there out there that are longing for simple root words to describe them that instead linger like high school girls hoping to be asked out to the prom? While the "no word to describe it" in this language myth is greatly overused by journalists, certainly there are cases when it is true.
And, sometimes, elegance of expression does matter. People who know short, logical Chinese number words are better at memorizing and manipulating long numbers than people who know the English equivalents. It is easier to keep accounts with Arabic numerals than Roman numerals. If physicists had been forced to use Newton's notation to do calculus instead of the notation of Leibniz, man would probably never have made it to the moon. People with larger vocabularies are more successful in life, in part, perhaps because they can better match words to the meaning that they intend.
Also interesting is the question of why some languages, for example, Yiddish, have produced many words that have been comfortably borrowed into ideomatic English, while other languages that have many people familiar with them do not, or do not sound as natural. For example, Yiddish words (perhaps because they are fit to a Germanic language) tend to sound more natural in English than words borrowed from Greek.