I've written before about the strange distribution of the vowels in Arabic, and then especially the difficulty one has to find the /u/ and /i/ contrasting. These two vowels cannot appear in one word right next to each other except in some very marked and probably also very artificial verbal formations.
The /u/ and /i/ in Classical Arabic are clearly distinct sounds but finding minimal pairs in many contexts is impossible. And often there is free variation between the two vowels. In a way these phonemes are 'sort of' phonemic. But the functional load is extremely slim.
Owens discusses this strange situation A linguistic history of Arabic at length. It's a great book in many aspects presenting a very sober view of the convoluted relationship between Old Arabic 'Neo-Arabic' and written and spoken Arabic.
I have always found phonemes that are clearly distinguished by speakers, even though they carry no functional load extremely fascinating.
If one can only show a phoneme to exist by finding minimal pairs, how does one account for these cases?
It is interesting to see that this system of Arabic where the distinction between V[+high] ~ V[-high] is far mor important than the distinction between the two high vowels is also found in other Afro-Asiatic languages.
In Berber a system very similar to this arrives. Sadly much of it has collapsed in all but Zénaga Berber since all Berber languages except for Zénaga lost the distinction between the short high vowels /i/ and /u/. But even with long vowels this is still relevant. In both verbal and nominal morphology alternation never occurs between /i/ and /u/ but always between /a/ and /i/ OR /u/. The vowels /i/ and /u/ perform somewhat the same role.
Did Proto-Afro-Asiatic have free variation of the /i/ and /u/? Or conditioned variation that was lexicalised due to extensive use of vowel patterning inside roots? Or maybe we shouldn't see a problem at all in phonemes with a low functional load and just assume that even in Proto-Afro-Asiatic the /i/ and /u/ already had a low functional load.
Although these phonemes of low functional load may seem very exotic to some, the situation arises in Dutch too. The distinction is between [s] ~ [z] and [f] ~ [v], the four fricatives of the Dutch language.
Some things to know about these fricatives:
- Word finally the voiced fricatives devoice.
- In coda, the fricatives generally assimilate their voicing to the next consonant.
- Exception to the above rule is with the marker of the past tense t~d which assimilates to the voicing of the fricative in front of it. kleven ~ kleefde but kleffen ~ klefte. The voicing of these fricatives is determined by the preceeding vowel. If it's a 'tense' vowel ([a], [e], [i], [u], [y], [ø] or a diphthong) the fricative is voiced. If it's a lax vowel (all other vowels) the fricatives is voiceless.
To sum these rules up, the voicing distinction of fricatives is non-existent in coda and intervocalic position.
Then we get to the real stuff, the fricatives in initial position.
It's worth noting that from a purely historical perspective there should not be a distinction. Proto-Germanic *f, *s simply shifted to v and z in initial position. But of course, as it goes, we Dutch people love to loan words.
v ~ f distinction is found in one minimal pair: vee 'cattle' (from PIE *peḱu-) ~ fee 'fairy'. The s ~ z distinction exists but knows no minimal pairs.
In many dialects either of the two versions has been generalised. North-western varieties (as in Amsterdam) have the voiceless variant unconditionally, while the more southern variants tend to have the voiced variants unconditionally.
But in my speech, and of plenty of other more centrally spoken dutch varieties (and in the 'official' dialect which is rather artificial) there is a distinction.
This distinction usually corresponds with the orthography.
I pronounce vee as [ve] and fee as [fe].
I pronounce soep 'soup' as [sup] and zoek 'search!' as [zuk].
But there are discrepancies, and this is common among speakers.
I pronounce vijf 'five' as [vɛif] but vijftig 'fifty' as [fɛiftɪχ]. The same thing happens in the words for six and sixty zes, zestig pronounce respectively as [zɛs] and [sɛstɪχ]. Pronouncing this any differently will sound wrong.
In less careful speech I'm more prone to drop the voicing distinction between [v] and [f] than [z] and [s], which you will never catch me doing. This is interesting, because the [v] and [f] have a minimal pair while [z] and [s] do not. So why do I feel like I should keep that distinction while I should drop the other?
So to sum up. The distinction of voicing in the Dutch fricatives s, f, z and v are extremely limited. Yet, the distinction is clearly upheld by some speakers despite the functional load being virtually non-existent.