"Initial" is pronounced "inishal," so why isn't the verb "initiate" pronounced "inishate"?

Question

Trying to answer a recent question about the pronunciation of the consonant "c" in the word word appreciate made me realize something I'm ignorant about: although I've read in a fair amount of places about the process of palatalization that caused words like issue to be pronounced with a "sh" sound /ʃ/, most of what I've read is about "yod coalescence" (see this Wikipedia article) which from what I understand involves a change like [sj] > [ʃ]. The two consonants "coalesce" or merge into one, and neither one is left over. But in some words, such as the verbs appreciate or initiate, the /ʃ/ doesn't seem to result from coalescence because the "i" is still pronounced as a separate vowel. In addition, the Wikipedia article says "Yod-coalescence has traditionally been resisted in RP" but I think even the most traditional British speakers would use /ʃ/ in words like initial and initiate (correct me if I'm wrong about this). So it seems to me that there must have been several processes of palatalization that applied within English at different times.

My question here is about words with the spelling pattern CiV where C stands for any palatalized consonant or group of consonant letters, such as "t," " c," "ss," "s," "g," "st" (corresponding to pronunciations like /ʃ/ /ʒ/ /dʒ/ /stʃ/) and V stands for any vowel letter. Why and when was the "i" completely lost in words like initial /ɪˈnɪʃəl/ (presumably after going through a semivowel stage), and on the other hand, why was it retained (or restored) as a syllabic unstressed vowel in words like initiate (v.)/ɪˈnɪʃiˌeɪt/?

The only pattern I can find is that words spelled with CiV that have the "i" pronounced as a separate vowel tend to have it followed by a primary or secondary stressed vowel. Here are some example suffixes to show what I mean:

With /i/:

• -tiate /ʃiˌeɪt/ (verb ending)
• -tiality /ʃiˈalɪti/ (noun ending)
• -tiary = /ʃiˌɛri/ (in American English) (noun ending)

Without /i/:

• -tial /ʃəl/
• -tion /ʃən/
• -tious /ʃəs/

The second set is composed of suffixes that are all pronounced with the completely unstressed "schwa" vowel. But I have no clue why this would make a difference.

Compare these related words:

• differential /ˌdɪfəˈrenʃl/, differentiation /ˌdɪfəˌrenʃiˈeɪʃn/
• religious /rɪˈlɪdʒəs/, religiosity /rɪˌlɪdʒiˈɑːsəti/
• beneficial /ˌbenɪˈfɪʃl/, beneficiary NAmE /ˌbenɪˈfɪʃieri/

I'd be interested in a description of either the historical sound changes or of the current phonological rules and patterns that result in this alternation.

(Note: I know this is sort of a technical "linguistic-y" sort of question, but I believe it still falls clearly within the scope of this site. I've decided to post it here rather than on the Linguistics Stack Exchange because I think I will have a better chance of getting an answer, or at least some useful comments, by asking here.)

Answer 1

You may say the word “initial” as “inishal” but that is not universal. It is peculiar to you and likely the region where you learned English. I say “initial” and “initiate” the same way, pronouncing the second “i” in both cases.

There are no rules for this except “accents differ” and “laziness is catching.” Letters get dropped out of laziness and then kids learn to talk that way and pretty soon everybody in a region is sharing that laziness and we call it an accent. People who live in Toronto all say “Toronno.” It is easier and it caught on and now it is part of the regional accent.

Answer 2

I think I may have found part of the answer, so I'll put in in a post here.

After posting my question, it occured to me that a possibly analogous case is preference, /ˈprɛf(ə)rəns/, where the middle "e" can be elided, and preferential /ˌprɛfəˈrɛnʃəl/, where it cannot. So I decided to look for more information about vowel elision and English stress and metrical structure.

So I looked for information about this in general and found this: Optimality Theory and Prosody, by Michael Hammond, at the University of Arizona. Here's the portion that I found helpful (some of the bolding was added by me):

Syllables are organized into patterns of alternating prominence (METRICAL FEET). In English, each foot contains a stressed syllable on the left and at most one stressless syllable on the right ([σ (σ̆)]).

[...]

Under certain conditions, a stressless vowel can go away in fast speech (SYNCOPE)

1 a. at the beginning of words:

• paráde -> práde
• Torónto -> Trónto
• Canádian -> Cnádian

2 b. before a stressless syllable

• ópera -> ópra
• géneral -> génral
• chócolate -> chóclate

3 c. before a stressless syllable before a stressed syllable

• réspiratòry -> réspratòry
• glòrificátion -> glòrficátion

Compare:

• óp(e)ra vs. òperátic
• gén(e)ral vs. gènerálity
• glórif'y vs. glòr(i)ficátion
• réspiràte vs. résp(i)ratòry

A rule-based account: Remove a stressless vowel if it precedes a stressless syllable or if it is word-initial.

Problem: the syncope rule misses the generalization that vowels syncopate only when an optimal (=disyllabic) foot would result. The two environments above can only be reduced to a single environment when the output is considered.

A constraint-based account:

                a.       STRESS: pronounce stressed vowels.
                b.       FOOTLESS: avoid unfooted syllables.
                c.       STRESSLESS: pronounce unstressed vowels.

STRESS >> FOOTLESS >> STRESSLESS

I'm not sure how exactly to apply this to the case of CiV words, but the metrical explanation seems to apply in both cases.