From Jason Cox (with additions and modifications by VHM):
In Taiwan, one often comes across efforts at using zhùyīn 注音 ("phonetic annotation") to hint to readers that a Hoklo Taiwanese reading of the sentence is preferred, rather than a Mandarin reading. Sometimes the characters are "correct" Hoklo Taiwanese (they convey the meaning of the characters directly); sometimes they will simply sound like Hoklo Taiwanese when read in Mandarin. Two examples that come to mind:
si chin e 係金ㄟ — the internet has loads of examples (that would equal shì zhēn de 是真的 ["is real / true"] in Mandarin)
As in the above examples, the Hoklo Taiwanese possessive particle ê [e] is more often represented with ㄟ [ei] instead of the zhùyīn ㄝ [ɛ], neither of which is exactly the same as ê [e] — but ㄟ [ei] being a dipthong, I would have imagined ㄝ [ɛ] would be the preferred substitute for ê [e].
My theory is that ㄟ was chosen not necessarily for the phonetic value, but because of Mandarin phonetic rules — where ㄟ [ei] can be a stand alone syllable, but ㄝ [ɛ] cannot. This rule seems to have been internalized deeply even by people who don't remember it's a rule.
Now, this ê [e], just as is de 的 in Modern Standard Mandarin, is by far the most frequent morpheme in Taiwanese. Michael "Taffy" Cannings, of Tailingua.com, writes:
Iûⁿ Ún-giân conducted a corpus analysis in 2005 on POJ (romanized; Church Romanization) texts and Han-lo (mixed romanization and Chinese character text). The study differentiated syllables and words, and conducted a frequency count for both. In both counts, ê was the most frequent result. I presume the word count is what Prof. Mair is after; in that case ê accounted for 8.13% of the entire POJ corpus, with the second most frequent being "i" (he/she/it) with 2.46% of the total.
The study itself is called Táiyǔwén yǔliàokù sōují jí yǔliàokù wéi běn Táiyǔ shūmiànyǔ yīnjié cípín tǒngjì 台語文語料庫蒐集及語料庫為本台語書面語音節詞頻統計 ("Taiwan Language Corpus collection and corpus-based Taiwanese written syllable word frequency statistics")
I should point out the usual caveats with regards to speech/written language differences, and also note that a large percentage of the POJ corpus is Christian material, so there will be some disparity between everyday street Taiwanese and the results obtained by Iûⁿ. However, I don't doubt that ê would still top the list if the corpus were more balanced.
Taiwanese ê [e] may be written with English "e" (I've often seen it written this way, even in the midst of Chinese characters), Mandarin de 的, Classical Chinese zhī 之, and even as Japanese no の (the latter three all possessive particles in their respective languages). And, as we have see above, it may be phonetically transcribed by the zhùyīn 注音 ("phonetic annotation") symbols ㄝ or ㄟ. One thing is certain: there is no established Chinese character for this most frequent morpheme in Taiwanese. This reminds me of a classic article by Robert L. Cheng (Zheng Liangwei) entitled "Taiwanese Morphemes in Search of Chinese Characters" in the Journal of Chinese Linguistics, 6.2 (June, 1978), 306-314. The same is true of many common (and plenty of not-so-common) morphemes in all of the Sinitic topolects.
[Thanks to Mark Swofford and Grace Wu]