English pronunciation / spelling guides appear to state that the letter/grapheme "a" is pronounced either as the "short a" with IPA symbol /æ/, as in "mat" or the "long a" with IPA symbol /eɪ/, as in "baby."
However, apparently there exist multiple words where all references I checked provide the spelling of "short e" with IPA symbol /e/ or /ɛ/ which I take as a variation in the IPA transcription not in the underlying pronunciation.
What I'm unable to come up with or find a reference for is when will the grapheme "a" be pronounced as /e/ (or /ɛ/) instead of the short "a" /æ/ that the typical rules would predict?
Examples of words that have "a" transcribed as /e/:
– temporary
– nefarious
– compare
– share
– dare
– subsidiary
(many more have two alternative pronunciations listed, one with /æ/ and one with /e/)
I recognize that all these examples include the sequence "ar" but I do not know if this is only because I was searching for this sequence in a corpus (based on the original example of "nefarious") or because there exist no other cases.
Actually, I find that my most trusted reference "The ABC's and all their tricks" by Margaret M. Bishop is including some of these words in the group of "-arr-" as in "carry", which my dictionaries transcribe as /ˈkæri/. This mixes two different pronunciations under the same rule. Is there no "rule" to help predicting the /æ/ or /e/ realization?
Best Answer
Except for words imported from non-Latin/Germanic languages, words with ortography matching the following regular expression will have the matching part realized as /er/:
*a[ei]r[iey]*provided that the matching syllable has the word-stress.In the "Shorter Oxford English Dictionary" (BrE pronunciations) the expression matches over 450 headwords (including false positives where "a" is not in a stressed syllable) with only a couple exceptions, like "Bohairic" (via Arabian), "dairi" (via Japanese), "etaerio" (From Greek via French; exception to the non-Latin precondition). In the cases that I investigated Merriam-Webster's Collegiate was also providing the phoneme /e/ for the syllable that includes the letter "a".
It appears that if the consonant was not "r" then the realization would be most often /ei/ (as in "tale", "calix", "agave" (BrE), "baby","navy", "zamia") but the "r" reduces that to the less heavy /e/. The rise is likely related to the stressed syllable being open.
There exist other examples, with all requiring that the 'a' be in a stressed syllable but there appear to be more exceptions than matches in the cases that I investigated ("scarus", "marum", "larum", "garum", --> /e/ but most other matches are a false positive).
This observation mostly matches the first comment of Janus Bahs Jacquet but strengthens the conditions for the raise to happen. Will be glad to revise if someone is able to provide a condition that provides a higher accuracy.