Since no one takes this, from what I could find:
Against Marking Accent Locations in Japanese Textbooks:
The fact that native listeners do hear an accent on a devoiced syllable indicates that associating an accent invariably with a high pitch cannot be an accurate description of the language.
The Production and Perception of Japanese Pitch Accent, section 1.3.1 "Devoicing and Accent":
Yoshida (2002, p. 41) discusses three possible realizations of accented high vowels that occur between voiceless consonants. Among them, (1)
is thought to be most common (Sakurai, 1998).
- Devoicing occurs, moving accent to the following mora.
- Devoicing is avoided to realize accent on that mora.
- Devoicing occurs with accent staying on that mora.
Handbook of Japanese Phonetics and Phonology, section 8.3 "Vowel Devoicing":
In short, there is a tendency to shift accent due to vowel devoicing. However, young speakers place accent on devoiced vowels and show no such accent shifts.
As for your questions:
- Why doesn't that rule apply to 助ける? Are there some other examples on such exceptions?
You are observing a most common scenario: moving devoiced accent to the following mora. Other examples would be any word where accent would otherwise fall on devoiced syllable. OJAD most of the time shows both variants (for instance 危機, 記者, 施設).
- I can also remember hearing it pronounced as ① somewhere. Does that pronunciation exist?
Comments to you post argue that this is what you get when you forcibly try to accent devoiced syllable.
Update: Why it makes sense to talk about accent on devoiced mora
Only the first link above tries to explain this (in a somewhat obscure way), other take it for granted, yet it doesn't seem to be obvious even to all native speakers, let alone us mere mortals :).
In some other languages accent (also called stress) is produced by one, any combination of two, or all three of:
- Pitch raise
- Sound volume raise
- Vowel length increase
Particularly because of (3) long vowels in Japanese are often perceived as "stressed" by non-natives, and (2)+(3) make it difficult to grasp how devoiced mora could possibly be accented (though actually it's possible to vary pitch of a consonant by rising corners of the mouth (i.e. "smiling")).
The first thing to note is that (2) and (3) do not apply to Japanese pitch accent (as the name suggests). The second thing is that in (1) it's not the pitch level that matters, but the pitch change. Normal speech is not a mere alteration between high and low pitch, but actually a cascade of pitch downsteps (audio: 三人寄れば文殊の知恵).
Contrary to the notation, pitch change occurs not between the morae (there are no gaps in speech so there's no "between" to begin with), but within the mora that comes after the accented one. Here are two audio samples from OJAD: 助ける and 助けて. Note how differently け is pronounced: in the first sample the pitch actually rises during け and falls on る. In the second sample the pitch falls on け. Hense OJAD shows たすける{LHHL} and たすけて{LHLL} respectively (the low pitch on た is arguably the result of pronouncing the word in isolation and in real speech may "inherit" the pitch of the preceding syllable).
Another example is 記者: while second mora is both louder (2) and longer (3) then the first, only pitch (1) matters, and pitch downstep is quite pronounced (at least to my ear), hence きしゃ{HLL}.
Finally it's worth remembering that pitch accent of a phrase isn't always a concatenation of individual accents, the OP's case could even be たすけてください{LHHHHHHL} (and the accent pattern may stay even when ください is dropped).