Why do we say an historical but a history? This question was originally posed by @shanselman on Twitter.
I have a theory about this. In my speech, and I believe that of many other Americans, an "h" in an unstressed syllable is either not pronounced or barely pronounced, except when it follows a vowel sound or a pause. This is an adjustment that is made unconsciously; people often don't notice that they're not pronouncing the "h". I say "a historical", but "some istorical".
The rule for a/an is that you use "an" before words which start with a vowel sound, and "a" before words that start with a consonant sound.
Both "a historical" and "an (h)istorical" are consistent with these rules; here by (h), I mean the "h" is pronounced very lightly, if at all. Most people use the first, but some people use the second. I think nearly all Americans pronounce the "h" in "historical" when the word stands alone, but after an indefinite article, some drop the "h" and use "an".
In the word "history", the first syllable is stressed, so the "h" is always pronounced. So "an history" isn't allowed by these rules.
Here are the final words of the relevant article in ‘The Cambridge Guide to English Usage’:
Nowadays the silent h persists only in a handful of French loanwords (heir, honest, honour, hour and their derivatives), and these need to be preceded by an. The h of other loanwords like heroic, hysterical and hypothesis may have been silent or varied in earlier times, leaving uncertainty as to whether an was required or not. But their pronunciation is no longer variable and provides no phonetic justification for an. Its use with them is a stylistic nicety, lending historical nuances to discourse in which tradition dies hard.
I broadly go along with that, noting at the same time that h-dropping more widely is a feature of some regional and social accents, and that in such cases the indefinite pronoun may indeed be realised as /an/.
I believe the main reason so many people say an historical is simply that they were taught that way. It's a special exception to the "use 'a' before a consonant sound" rule, and it seems older generations of English speakers were often specifically taught this special case.
The fact is that a historical is perfectly correct, more logical, and preferred by a large number of English experts. And it is overwhelmingly overtaking an historical in modern usage (see Ngram).
While I agree that the practice most likely originated as Peter Shor describes; namely, that many speakers do not (or only barely) pronounce the 'h' in historical, due to it being unstressed; I believe this phenomenon is not particularly strong in American English. I would guess it began in Britain (and probably continues more strongly there), with its h-dropping dialects.
Some speakers drop the "h" sound in the word "historical", which means that the consonant "n" needs to be added to the article "a". However, it's been quite a while since I last heard this, everyone I know says "historic" with the sound "h" pronounced.
Background information: this word has been taken from Greek, but apparently it entered the English language via French. In French the letter "h" isn't pronounced, hence the English imitation which has now become outdated