Why do we say 頼もしい instead of 頼ましい?

Question

In a previous question, I posted a list of adjectives produced from verbs using the しい suffix. In each example, it seems that しい attaches directly to the 未然形:

勇む　　→　　勇ま　＋　しい　　　 isam-a-sii
悩む　　→　　悩ま　＋　しい　　　 nayam-a-sii
喜ぶ　　→　　喜ば　＋　しい　　　 yorokob-a-sii
妬む　　→　　妬ま　＋　しい　　　 netam-a-sii
呪う　　→　　呪わ　＋　しい　　　 norow-a-sii
慕う　　→　　慕わ　＋　しい 　　　shitaw-a-sii
好む　　→　　好ま　＋　しい 　　　konom-a-sii

In all of the above words, it appears that the 未然形 has the -a- surface form.

However, in these three words, it seems that it has the -o- surface form:

好む　　→　　好も　＋　しい　　　 konom-o-sii
狂う　　→ 　　狂お　＋　しい　　　 kuruw-o-sii　
頼む　　→　　頼も　＋　しい 　　　 tanom-o-sii

What I'd read previously is that -a- and -o- are both considered the same underlying form because of the sound change /au/ → /o:/. In other words, the -o- surface form is underlyingly -a-, but changes to -o- as part of /o:/.

However, these three adjectives do not contain the long vowel /o:/, so I don't think that explains why they have an -o-. And if this is the result of a sound change, it doesn't appear to be a regular sound change, because most adjectives have -a-. In fact, both konom-o-sii and kuruw-o-sii have -a- versions:

kuruw-o-sii　　　kuruw-a-sii 　　
konom-o-sii　　　konom-a-sii 　　
tanom-o-sii　　　*tanom-a-sii 　　(I can't find evidence that ×頼ましい exists)

So how can these -o- forms be explained?

Answer 1

I answered this question a while back, but unfortunately I only half-understood it, and there was some wrong information in it. After some time researching this question, and now an entire semester of a course about Japanese adjectives, I am ready to answer it again. As such, I have deleted my previous answer.

So, since しい appears to attach to the 未然形 of a verb, why with some small subset, does it attach to the -o- form of the 未然形 rather than the -a- form?

That is because しい does not attach to the 未然形 of a verb.

Rather, it attaches to the 被覆形【ひふくけい】¹ of a verb. Most often, this form resembles the 未然形 -a-², but there are a number of other examples。Some examples which still exist in modern Japanese include: 乏【とぼ】しい 恐【おそ】ろしい 恨【うら】めしい 宜【よろ】しい

As far as I can tell (and indeed, as far as I can find in the literature), there is not a solid rule for determining the 被服形 of a verb. -a- is the most common, e.g. 喜【よろこ】ぶ→喜【よろこ】ば + しい. However, as seen in the examples above, for words in which the previous syllable ends in -o (ぼ, そ, ろ, の, etc.), there is a trend toward the 被覆形 also ending in -o.³

---

Notes:
¹ The 被覆形 is essentially a linking form. It is most often seen in reference to compound nouns, e.g. 雨【あめ】 + 雲【くも】 = 雨雲【あまぐも】 (め→ま) and 白【しろ】 + 橋【はし】 = 白橋【しらはし】 (ろ→ら）. The opposite of 被覆形 (i.e. the form when used as a standalone word) is 露出形【ろしゅつけい】.

² This is perhaps easily understood because (at the very least after 上代), Japanese only has 5 vowels and it will thus almost certainly look like something.

³ In fact, modern 喜ばしい was also originally *喜ぼしい. e.g.

ヨロコボシ［悦］伊豫国_与利白祥鹿_平献奉_天在_札方有札志与呂許保志_止奈毛見_流 （四十六詔・続紀神護景雲三［769］年）

伊豫国【いよのくに】より白【しろき】祥【しるし】鹿【しか】を献奉【たてまつり】て在【あ】ればうれし よろこぼし となも見る

^{Note: The example above was originally printed vertically, with the bold characters having a vertical line to the left.}

---

References:
蜂矢真郷(2012)「上代の形容詞」『萬葉』(萬葉学会) 第212号, pp. 1-35