After my post about using an when an immediately following parenthetical begins with a consonant (but the first word after the parenthetical begins with a vowel), I felt the need to get some Google counts. What follows are the stats for strings of the form “a/an to me (at least/anyway) X”.
The finding is that Julian Barnes is not alone: there are a lot of people out there writing things like an–to me–unknown singer.
I wonder about the converse: a–obviously–preposterous idea. I easily found an example, “a (obviously refurbished) replacement unit”, but didn’t investigate systematically (and it seems there are a lot of people typing things like “displayed with a obviously wrong pixel ratio”, so we need to control for baseline use of a before vowels).
Tables and graphs come after the jump…
First, the raw numbers, for all the cases where more than zero hits were returned for “a/an to me X”. I used Bruce Hayes and Tim Ma’s Query Google. As a control, I also searched for words that begin with a consonant–you can see that they pretty much never have an. Just for fun, I tried a bunch of words that begin with [h] and then an unstressed vowel, but got hits for just one, hilarious.
|word||initial segment type||hits for “a to me X”||hits for “an to me X”||% “an”|
Now in histogram form (losing the information about how well attested each word type is). You can see that the C-initial words pretty much never take an, but for the V-initial words, it’s actually not that unusual:
I tried the same thing with “a/an to me at least X”. I was curious to see whether a longer parenthetical would weaken the tendency to use an with vowel-initial X, but the data are probably too sparse here to tell:
|word||initial segment type||hits for “a to me at least X”||hits for “an to me at least X”||% “an”|
In histogram form:
And finally, “a/an to me anyway X”, with hardly any data, and none on V-initial X:
|word||initial segment type||hits for “a to me anyway X”||hits for “an to me anyway X”||% “an”|