A few days ago, Larry Horn sent this note to the American Dialect Society's discussion list:
On an article lauding the Texas Rangers’ defense in today’s NYT sports section, I did a double-take on reading that
The defense—anchored by shortstop Elvis Andrus and the impregnable glove of Adrian Beltre at third base—has saved more runs above average than any other team but the Rays.
Once I got past the metaphor in which baseball gloves may or may not become pregnant, my first thought was that the writer (Neil Payne) had meant “unimpregnable”, i.e. incapable of being impregnated, just as “uninflammable” means 'incapable of becoming inflamed'. I checked the OED and found to my surprise that, as they say, “there is no such word” as unimpregnable, and that the im- (i.e. iN-) of impregnable can only be a negative prefix, so that impregnable already (officially) means what I had thought unimpregnable would mean, rendering the doubly-prefixed form otiose. Evidently, unimpregnable does not and never did exist.
Except that it did and does. Google Books includes 300 or so hits for unimpregnable (while asking me if I meant impregnable), including one from William Harvey’s classic 17th century treatise on the circulation of the blood, several from assorted 19th century journals, monographs, and novels, one from a 1917 novel by Jack London ("Had he been more than a normal thoroughbred dog, he would have continued to assail his unimpregnable enemy until…"), one from a 2010 novel by Francine Prose ("Having been incarcerated in the Castel Sant'Angelo, he had somehow escaped Valletta's unimpregnable fortress-prison and had left the district without permission…”), and so on. Of course, unimpregnable has the virtue (absent, I would argue—with some support from regular Google entries—from its monoprefixal counterpart) of being unambiguous. But officially, it seems, unimpregnable is a hypernegation, and the NYT sportswriter was not employing a hyponegation. Go know. (I still think unimpregnable *is* a word and belongs in the OED, though.)
But there are a couple of etymological glitches here. Specifically, there are two kinds of pregn- words as well as two kinds of -in words, plus the -atable ~ -able problem.
First, impregnable is actually a Middle English malapropism for imprenable, as the OED explains:
Etymology: Corrupted < impreignable, imprenable, < French imprenable, < im- (im- prefix2) + prenable able to be taken, < pren-, stem of prendre to take. The g was evidently in imitation of the g mute in reign, deign, and the like, though it appears to have sometimes led in 16th cent. to the pronunciation /nj/
Thus the glosses "Of a fortress or stronghold: That cannot be taken by arms; incapable of being reduced by force; capable of holding out against all attacks", and "fig. That cannot be overcome or vanquished; invincible, unconquerable, proof against attack".
In contrast, impregnate "To make (a female) pregnant; to cause to conceive; to get with young" is from "late or medieval Latin imprægnāt-us, past participle of imprægnāre", which comes from the same Latin verb praegno "to be pregnant" that pregnant does.
And the OED's im- prefix2 is the "assimilated form in Latin of the negative prefix in- prefix3 before b, m, p", whereas the im- of impregnate is the assimilated form of a prefix that the OED glosses as
used in combination with verbs or their derivatives, less commonly with other parts of speech, with the senses ‘into, in, within; on, upon; towards, against’, sometimes expressing onward motion or continuance, sometimes intensive, sometimes transitive, and in other cases with little appreciable force.
Additional confusion is provided by the fact that some Latin-derived verbs in -ate have -ble forms without the -ate, or both with and without it.
Thus calculate ~ calculable ~ calculatable, separate ~ separable ~ separatable.
Actually, calculatable and separatable are not in the OED, nor in Merriam-Webster's — but both of these words do have (small numbers of) respectable-looking hits on Google books. For that matter, so do impregnatable and unimpregnatable, though these have different meanings from their -ate-less counterparts.