That would be a hard thing to prove. By our practice the burden of proof is on those who believe -ability is a suffix. It is highly likely that an adjective ending in -able was formed (or borrowed) first. But, if anyone has evidence that first attestation of a given word ending in -ability preceded first attestation of a word ending in -able or that a sense of a word ending in -ability is not derivative of the word ending in -able, that would be evidence for keeping it.
If we accept a purely diachronic approach to such matters, the absence of such evidence would be dispositive. However, intuitively, I believe that it is a suffix and a productive one: that speakers think of it as a unit to be grafted on to a verb base. Furthermore, other dictionaries have it: to wit, WNW, RHU, Collins, Online Etymological Dictionary. Quinlon's Affixes doesn't have it. I think the disagreement between our practice and these lemmings suggests this is an RfD or, possibly, BP matter. Move to RfD. DCDuringTALK 16:50, 29 September 2010 (UTC)
Keep unless you can explain why -able + -ity = -ability. Actually, even then, keep, since [[-ability]] seems the best place to put that explanation. —RuakhTALK 22:05, 29 September 2010 (UTC)
Dictionaries that have it don't draw a sharp line, usually saying both that it is from Latin-abilitas < -abilis + -itas and from -able + -ity. Online Ety Dict. denies that is descends from ability, which might have been the only remaining plausibility. I would be a little surprised if we can find many cases in which it is a true diachronically derivational suffix. If we don't keep it we will certainly need explanations at [[-able]] and [[-ity]] and redirects to one of those two, probably [[-able]]. DCDuringTALK 22:49, 29 September 2010 (UTC)
You're both smart guys. I'm prepared to drop it if necessary. RFV doesn't really work for affixes, as affixes by their nature are parts of words, so you can't attest them as discrete units. Mglovesfun (talk) 23:06, 29 September 2010 (UTC)
AFAICT we had finessed the question of attestation by saying we needed three examples of attested terms having been formed from an affix, apparently using the magic number three to maintain some connection with WT:CFI. Arguably this is too stringent. Productive affixes most clearly demonstrate their productivity by forming nonce words, according to morphologist Ingo Plag, in Word Formation in English. But such word are not attestable unless they are serial nonces. I have no alternative to CFI to suggest, but would at least follow lemmings from the better class, such as those above and Partridge, who has "-bility" in Origins. DCDuringTALK 23:36, 29 September 2010 (UTC)
No counterargument. But I believe that language speakers and writers can invent and language hearers and readers will understand any neologism ending in -ability that follows the rules, whether or not there is a corresponding word ending in -able. As this is outside of CFI, the matter seems one for RfD where the lemming rationale applies AFAICT or for BP. DCDuringTALK 11:51, 30 September 2010 (UTC)
Yes we could move it to rfd and I'd be happy with that. I think the attestation route should be left open, though. Mglovesfun (talk) 14:56, 30 September 2010 (UTC)
Originally nominated at RFV, consensus was that it should be here. Though, I'm still not sure why. Mglovesfun (talk) 09:07, 5 October 2010 (UTC)
I think RFV was a better place for this, because the issue is whether any terms have been coined using -ability as a suffix (rather than taking an -able and sticking -ity on it). What is the rationale for deletion? Equinox◑ 23:57, 5 October 2010 (UTC)
I dunno, I moved it here per other people's comments. I was happier with rfv. Mglovesfun (talk) 11:15, 6 October 2010 (UTC)
Some OneLook coverage -ability at OneLook Dictionary Search. DCDuringTALK 14:57, 6 October 2010 (UTC)
Yes, I think the consensus was that this wasn't going to be cited, but it might be worth keeping in spite of CFI. Which I oppose. Well, in this case I oppose it. Mglovesfun (talk) 22:46, 8 October 2010 (UTC)
Keep. Equinox says "the issue is whether any terms have been coined using -ability as a suffix (rather than taking an -able and sticking -ity on it)", but doesn't bother to justify that statement. (Perhaps it's intended simply as a statement of personal opinion — in which case I suppose it's a valid one — but somehow it doesn't come off that way to me.) Mglovesfun implies that the CFI require this to be "cited" in some specific way, but it obviously falls under the "clearly widespread use" clause. If you want me to cite it, I can easily do so, using one cite from google books:"probability", one from google books:"readability", and one from google books:"commensurability".
Perhaps a consensus will form here along the lines of Equinox's assertion; if so, this may then become suitable for RFV. In the meantime, it doesn't make sense to list this at RFV. It's not fair to our citers to ask them to cite something when there's no agreement about what kinds of citations are acceptable.
The nominator of this entry is Mglovesfun, with his introducing statement "Aren't all the derived terms suffixed with -able then with -ity? Mglovesfun (talk) 13:53, 29 September 2010 (UTC)" (see RFV). So what is the entry charged with? Per what principle, even an implied one, should the entry be deleted? I do not know. I can only guess a principle of the sort "If a suffix appears to be a result of concatenation of two suffixes, it should be excluded". I know of no precedent of application of this principle, nor is this principle expressly stated in CFI. In order to vote delete, I would have to tentatively accept this principle without having investigated its implications and consequences. Hence I err on the side of keep. --Dan Polansky 09:15, 11 October 2010 (UTC)
Semi-relevant comment: I'm glad I nominated this as it's gonna be good stuff to have on the entry's talk page when this debate is closed. Mglovesfun (talk) 08:33, 13 October 2010 (UTC)
I did once consider adding the Italian equivalent -abilità, but decided that it was just -abile + -ità. I am still of the same mind. SemperBlotto 08:41, 13 October 2010 (UTC)
What makes sense for the Italian may not make sense for the English. It may be a general characteristic of -ità that a preceding -e is dropped, but it's certainly not a general characteristic of -ity that a preceding -i- is inserted. We don't have taciturn → *taciturinity, nor mediocre → *mediocirity. —RuakhTALK 12:15, 14 October 2010 (UTC)
The etymology is certainly wrong, since there is not a suffix -abilitas in Latin. Latin adds -itās to an adjective ending in -ābilis. --EncycloPetey 04:58, 19 October 2010 (UTC)