Why don't "-use" verb-noun pairs obey initial stress derivation?

Question

It's well known (and several past questions on this SE have covered) that to convert a two-syllable Latin-derived English verb into a noun, you shift the stress to the first syllable. This is apparently called "initial stress derivation". Some examples:

• permít (v.) ~ pérmit (n.)
• addréss (v.) ~ áddress (n.)
• recórd (v.) ~ récord (n.)
• expórt (v.) ~ éxport (n.)

However, I've just realized there's a class of verbs that obey a different rule: verbs ending in -use. When these verbs change into nouns, they keep the final-syllable stress, but devoice the final consonant: so the verb ends in /uz/, and its derivative ends in /us/.

• abuse: abú[z]e (v.) ~ abú[s]e (n.)
• diffuse: diffú[z]e (v.) ~ diffú[s]e (adj.)
• reuse: reú[z]e (v.) ~ reú[s]e (n.)
• excuse: excú[z]e (v.) ~ excú[s]e (n.)

My question: Why do verbs that end in "-use" follow this pattern rather than initial stress derivation?

(Or perhaps not strictly "rather than" - I can also think of one example that both shifts the stress and devoices the /z/:)

• refuse: refú[z]e (v.) ~ réfu[s]e (n.)

Answer 1

This is what Sir Ernest Gowers has to say in "Fowler's Modern English Usage" (Oxford 1965) under the heading noun and verb accent:

When there is both noun and verb work to be done by a word, and the plan of forming a noun from the verb, or a verb from the noun, by adding a formative suffix (as in stealth from steal) is not followed, then one word may be called on to double the parts. In that case there is a strong tendency to differentiate by pronunciation, as in use (n. us, v. uz); such a distinction is sometimes, as in use, unrecorded in spelling, but sometimes recorded, as in calf and calve. Is is not possible to draw up a complete list of the words affected, because the impulse is still active, and the list would need constant additions, especially of words whose pronunciation can be modified without change of spelling.

He goes on to give partial lists of the two main forms that the question identifies; the largest class being words "whose accent is shifted from the first syllable in the noun to the last syllable in the verb"; the other group are words, especially monosyllables, which "are differentiated not by accent but by a modification in the noun or verb of the consonantal sound at the end, which is hard in the noun and soft in the verb."

He adds that very often this change of voice is recorded in spelling (eg cloth, clothe; bath, bathe; etc) and that this confirms "the fact that the distinguishing of the parts of speech by change of sound is very common".

Answer 2

This is just like house (v.) with a voiced final consonant becoming house (n.) with an unvoiced one.

Normally we reflect this change in spelling:

• advise (v.) vs. advice (n.)
• wolve (v.) vs. wolf (n.)

But with the -se verbs, it doesn’t always work out that way.

It’s because the “silent e” for the verb used to be pronounced, and an -s- between two vowels was automatically voiced, just as in clothe, breathe, or mouthe as verbs. We retained the voicing, which is now phonemic, even once the final -e became completely silent.

Answer 3

It seems possible to me that it is more or less a coincidence that abuse, diffuse, reuse, excuse all end in "use", are all pronounced with /uz/ as verbs and with /us/ as nouns, and are all stressed on the last syllable whether or not they are verbs or nouns. "Initial stress derivation" is just one pattern for forming derived nouns from iambic verbs (typically ones where the first syllable is a prefix); there are a fair number of examples of it, but it doesn't apply to all two-syllable verbs from Latin. There are also iambic verbs that have corresponding nouns formed by "zero derivation" with no stress change, or that have iambic nouns formed by changing the end of the verb.

Here are just the examples I've been able to find:

No difference in spelling or pronunciation between noun and verb:

acclaim, account, address*, advance, array, arrest, assault, attack, attempt, command, concern, consent, control, decree, defeat, delay, delight, demand, design, dismay, dispute, distress, distrust, divorce, eclipse, effect, employ, exchange, exhaust, supply, support, surprise, report, return, result, reward, reform, regard, revolt, resort, remark, respect, regret, retort

(*However, some American English speakers may pronounce "address" with the stress on the first syllable.)

Noun and verb that have different endings, but both are iambic:

-nd to -nt: descend, extend, ascend : descent, extent, ascend
-nd to -ns: expand, expend, defend, offend, suspend, respond : expanse, expense, defense**, offense**, suspense, response

(**However, "offense" and "defense" may take initial stress in some contexts, especially in American English—another example of the variability that has been mentioned in other answers and the linked page).

One verb/noun pair that seems particularly similar to the "use" examples that you listed is "devise/device"; likewise "advise/advice". With these, the alternation between /z/ and /s/ is indicated by the (non-etymological) use of "se" vs. "ce".

The comparison to "produce" brought up in the comments doesn't seem on the right track to me. The verb "produce" ends in voiceless /s/; the verbs abuse, diffuse, reuse, excuse end in voiced /z/. The spelling "-use" is ambiguous and can represent /us/ in English, but it doesn't in these four verbs. (The reasons for that are I think partly etymological and partly analogical: I wrote an an answer to a separate question where I try to give a more detailed explanation of why word-final "se" represents /s/ in some words from French but /z/ in others: Why are “using” and “user” pronounced with “s” as “z” while “use” just uses “s”?)

The Oxford English Dictionary indicates that at one point, the noun "produce" was actually pronounced with stress on the last syllable, quoting a line of verse from Dryden:

1700 Dryden To J. Driden in Fables 98 You hoard not Health, for your own private Use; But on the Publick spend the rich Produce.