Mora count for 思ふ in Classical Japanese?

Question

During an analysis I was performing of the poems contained in the 古今集 I noticed a curious pattern. Otherwise perfectly-formed poems included in the collection seemed to ignore the morae-count on a consistent basis with respect to the verb 思ふ【おもふ】, allowing it an extra beat, as you can see in the following examples:

6-7-5-7-7:

をしと思【おも】ふ
心【こころ】はいとに
よられなむ
ちる花【はな】ごとに
ぬきてとどめむ

5-8-5-7-7:

五月雨【さみだれ】に
物思ひ【ものおもひ】をれば
郭公【ほととぎす】
夜【よ】ふかくなきて
いづちゆくらむ

5-7-6-7-7:

夕【ゆふ】ぐれは
雲【くも】のはたてに
物【もの】ぞ思ふ【おもふ】
あまつそらなる
人【ひと】をこふとて

5-7-5-8-7:

いつはとは
時【とき】はわかねど
秋【あき】のよぞ
物思ふ【ものおもふ】事【こと】の
かぎりなりける

5-7-5-7-8:

春【はる】きぬと
人【ひと】はいへども
うぐひすの
なかぬかぎりは
あらじとぞ思【おも】ふ

These examples are not exhaustive; they are merely to demonstrate the phenomenon and to show that it doesn't matter which line it appears in.

What I'm wondering is if the rules on morae-counts are more complex than originally thought and actually have a provision allowing for this, or if there were some other explanation for it.

On a similar note, I noticed while researching this question that あらば gets a pass when on a 5-morae line when paired with 3-kana words (e.g. 「こころあらば」、「いのちあらば」), while the related あれば seems to get a free pass in more settings (about 1/3 of total occurrences). I'm almost inclined to think any potential rule involves the ば form, however that's not a consistent element in the 思ふ cases.

Answer 1

I suspect this might be an example of poetic license or even contraction. Note that all of the 思ふ instances above follow on another mora from the お行, leaving open the possibility that をしと思{おも}ふ, for example, was actually read as をしともふ, thus producing the expected mora count. I note too that 思う has a pitch pattern of おもう{LHL}, making the お effectively unstressed and more likely to be elided if preceded by another お sound. This kind of contraction is common in faster informal speech, and might have been utilized for poetic effect.

Answer 2

You can read (or sing) them as:

■■■■■■□□　[を]{お}しとおも[ふ]{う}～－ (like ■■■■■□□□　おしとおもー)*
■■■■■■■□　こころ[は]{わ}　いとに～
■■■■■□□□　よられな[む]{ん}～－－
■■■■■■■□　ちるはな　ごとに～
□■■■■■■■　－ぬきて　とどめ[む]{ん}

■■■■■□□□　さみだれに～－－
■■■■■■■■　ものおも[ひ]{い}　[を]{お}れば
■■■■■□□□　ほととぎす～－－
■■■■■■■□　よふかく　なきて～
□■■■■■■■　－いづち　ゆくら[む]{ん}

■■■■■□□□　ゆ[ふ]{う}ぐれ[は]{わ}～－－
□■■■■■■■　－くもの　はたてに
■■■■■■□□　ものぞおも[ふ]{う}～－ (like ■■■■■□□□ ものぞおもー)*
□■■■■■■■　－あまつ　そらなる
□■■■■■■■　－ひとを　こ[ふ]{う}とて

■■■■■□□□　いつ[は]{わ}と[は]{わ}～－－
□■■■■■■■　－とき[は]{わ}　わかねど
■■■■■□□□　あきのよぞ～－－
■■■■■■■■　ものおも[ふ]{う}ことの
□■■■■■■■　－かぎり　なりける

■■■■■□□□　はるきぬと～－－
□■■■■■■■　－ひと[は]{わ}　い[へ]{え}ども
■■■■■□□□　うぐ[ひ]{い}すの～－－　
□■■■■■■■　－なかぬ　かぎりは
■■■■■■■■　あらじとぞおも[ふ]{う} (like ■■■■■■■□ あらじとぞおもー)*
　

■■■■■■□□　こころあらば～－
■■■■■■□□　いのちあらば～－

... in 8 beats.

*When the [字余]{じあま}り(extra mora) is a 母音(vowel) or [撥音]{はつおん}(「ん」), it's often merged with the previous sound and they're read like a single mora, i.e.:

おしとおもう(6) → おしとおもお → おしとおもー / おしとおも(5)
あらじとぞおもう(8) → あらじとぞおもお → あらじとぞおもー / あらじとぞおも(7)

And as for the 字余り which is not a 母音 or 撥音, such as in 「こころあらば」「いのちあらば」, I just heard that it's added intentionally, so that the line or the song itself would sound different and stand out from the other lines or other people's songs. (I didn't know this, just this morning I asked someone who's studied classical Japanese and 百人一首 at senior high...)

Answer 3

The reason why such a pattern appears is, as currently understood, that Japanese was not a mora-timed language yet at the time 古今和歌集 was compiled.

Until around the 16th century, meters were counted by syllable, or in other words, by the number of vowel clusters. That makes adjacent vowels in a continuation can be freely merged into one unit, and a bare vowel syllable appears in the middle of words can be virtually "discounted" from the meter.

All of your examples involve hiatus, and easily fall under this pattern. This rule is considered regular; tanka (waka) at that time was largely manneristic being an important tool of communication and accomplishment among the nobles, that shows little discretionary deviation in format.

An extreme example contains four out of five lines of a tanka seemingly have an excessive mora (字余り):

いせのうみに i-se-no-u-mi-ni: 6
もしほやくあまの mo-si-fo-ya-ku-a-ma-no: 8
かぜをいたみ ka-ze-wo-i-ta-mi: 6
そらをしぼむる so-ra-wo-si-bo-mu-ru: 7
きみにぞあるべき ki-mi-ni-zo-a-ru-be-ki: 8
（賀茂保憲女集）

but if you count vowel clusters:

i-se-nou-mi-ni: 5
mo-si-fo-ya-kua-ma-no: 7
ka-ze-woi-ta-mi: 5
so-ra-wo-si-bo-mu-ru: 7
ki-mi-ni-zoa-ru-be-ki: 7

which is perfectly normal.
(I haven't found an all-字余り poem yet. Much appreciated if someone let me know :)

Another interesting example is a religious work by 最澄 that involves foreign words, which is rarely seen in the world of tanka:

あのくたら a-no-ku-ta-ra
さんみゃくさんぼだいの sam-mya-ku-sam-bo-dai-no
ほとけたち fo-to-ke-ta-ti
わがたつそまに wa-ga-ta-tu-so-ma-ni
みゃうがあらせたまへ myaũ(?)-gaa-ra-se-ta-ma-fe

At first glance, it looks as if an extravagantly broken 5-10-5-7-9 verse in mora-counting, but still retains regular meter in syllables.