It's been a while since we talked about snowclones; "every schoolboy knows...", on 2/27/05, seems to have been the most recent formula to get attention here.
Now, over on the American Dialect Society mailing list, the snowclone "once a X, always a X" has come up, in Barry Popik's pursuit of the instance "Once a Dodger, always a Dodger", which he's (so far) traced back to 1934. You can follow the discussion on the archives stored at the ADS site.
Today Larry Horn took to wondering about the general formula:
I wonder if there are earlier cites for "Once a(n) X, always a(n) X". When I was a young'un, I remember being taught on my professor's knee "Once a phoneme, always a phoneme". That was after 1934, though.
I too was taught on my professor's (figurative) knee about the "once a phoneme, always a phoneme" principle, but that was about 25 years after 1934. And I too wonder about other instantiations of the formula.
Update from Larry Horn, a few hours later:
There are 886K google hits for "once a" "always a", featuring on the first couple of pages no Dodgers or phonemes, but a motley collection of marines, cheaters, friends, Caesarean[s] (which come to think of it I've read about), deserters, tuba players, and such. Plus another 32.4K for "once an" "always an", featuring Arabs, Indians, English majors, orphans, and addicts. Some of these (marines, Caesareans, addicts) are no doubt proverbial, others are formed productively as needed. Anyone want to tackle the issue of first cite for the construction type?
zwicky at-sign csli period stanford period edu
Posted by Arnold Zwicky at May 17, 2005 12:11 PM