As Metehan Hoca said, there are exceptional usages as in our case [dative case] we witness both ways and I ve gone for keeping the vowel. And in this particular case there is another subrule which is ;

These sort of words , if they are in reduplication form [ikileme] the vowel does not drop.Such as ;

Gönül gönüle

Gönülden gönüle

Ağız ağıza

Ağızdan ağıza

It´s understandable. Dropping the stem vowel in reduplicated forms would break the alliteration, the power of repeating.

Thanks for the other view, metehan2001.

This grammar rule is called "orta hece ünlüsünün düşmesi" (= dropping of the vowel in the middle syllable): 

The vowel in the middle syllabel in a grammatical unit consisting of more than two syllabels often drops when this middle syllable has no consonant at the end.

gö-nül

gö-nü-lü (the midde syllable has no consonant at the end) => (gön-lü gönlü

gö-nül-den (the middle syllabel has a consonant at the end) => gönülden

sa-bır

sa-bı-ret-mek => sab-ret-mek) sabretmek

sa-bı-re-den => (sab-re-den)  sabreden

sa-bır-dan => sabırdan

ka-hır

ka-hı-rol-mak => (kah-rol-mak) kahrolmak

ka-hır-lar => kahırlar

hü-zün

hü-zü-nü-ne => (hüz-nü-ne) hüznüne

hü-zün-de => hüzünde

ö-mür

ö-mü-rü => (öm-rü ömrü

ö-mür-den => ömürden

A new approach, scalpel.

It looks like a phonetic law which at a certain time has affected all stems of the type

              (C)VCVC-

What is notable in this group is that the stems are from different origins. The law has come like a wind and blown over all these stems no matter if they are own or borrowed.

There is a tendency for two-syllable words in all languages I guess. This shows in children´s language. Children sometimes even prolong one-syllable words to give them the ideal amount of syllables. I don´t know if this shows in Turkish.

Now the question is if there are stems of the type (C)VCVC- which do not follow this law. My vocabulary is too limited to look for them. I mean if it really was a phonetic law at some time all exceptions should have an explanation (like the word being younger for instance).

The only example that comes to my mind is uzun. Why uzun: uzunu but hüzün: hüznü? The etymological dictionary says the old stem is *uzı-, the second syllable was long until 1300 or so and there is actually and old affix in the end of uzun. Maybe this is the reason.

I wrote my original question for practical reasons  -  I felt I couldn´t derive the words  -  but it has turned out to be an interesting one. Now it seems that they are more a rule than an exception or am I wrong again?

Well,

Set the Arabic loans aside. Turkic words are all old words and very specifically are used for body parts so it´s a long development over years.

Alın - Aln-ı

Omuz - Omz-u

Karın - Karn-ı

Ağız - Ağz-ı

Burun - Burn-u

Boyun - Boyn-u

Koyun - Koyn-u

etc.

Why should I set them aside if they follow the same pattern? Speakers hardly make these distinctions.

I have understood phonetic laws  -  if this is one, it´s my guess  - work in a language for some limited time and go. That´s why I suggested it happened while the Arabic loans were already there. A phonetic law doesn´t ask where the word came from, it looks for certain phonetic combinations and operates in them like a machine.

You don´t seem to get my point.

They are different developments. Arabic loans come from Arabic without vowels in between. We add them according to Turkish phonetics. On the other hand Turkish ones drop their vowels as a result of some historic development of some specific class of words.

They come from different ends and they meet in the middle.

Yes. I really missed it. That the Arabic loans didn´t have the vowel in the first place and there must have been more steps in their development.

But my point is that phonetic changes don´t happen to certain groups of words, they happen to certain queues of phonemes.

Whatever the order of events has been a phonetic law has come and swallowed them all. Not one by one or group by group.

To the contrary of what some people think on this board, the rule is not limited with the old Turkish words for "body parts", nor with some of the exact Arabic origin words in Turkish..

And despite what they claim, koyun is not a body part. 

koyun - koynu

kayın, oğul, çığır are not body parts either but they follow the rule:

kayın - kaynı / kaynata, kaynana

oğul - oğlu

çığır - çığrı

Some of the passive forms of the verbs formed with factitive -r- follow the rule:

ça-ğı-rıl-mak => çağ-rıl-mak

bu-yu-rul-mak => buy-rul-mak

çe-vi-ril-mek => çev-ril-mek

ka-vu-rul-mak => kav-rul-mak

Derivatives of such words follow the rule as well:

ça-ğı-rış => çağ-rış

bu-yu-ruk => buy-ruk

çe-vi-rik => çev-rik

kav-vu-ruk => kav-ruk

Right scalpel. But it´s not a rule right? They are all old words and mostly the body parts (because most of them are old words). And they are the sound changes developed over a long period (I am talking about Turkish words). They are exceptions to be memorized and given in dictionary as such (usually where you see karın you also see karnı for example).

On second thoughts what I suggested about (C)VCVC stems can´t be true. From my lacking vocabulary I find plenty of (C)VCVC nouns which don´t lose any vowels but keep them tight. Such as okul, bedel, keder, kadın, bahar, besin, onur, ışık, atık, katır, umur, iman, beyaz. So much for that.