This is the fifth part in a series on historical linguistics using Terry Crowley and Claire Bowern's Introduction to Historical Linguistics as a base. Part I is here. Part II, here. Part III, here. Part IV, here.
So, on with feature. Features, as we saw last time, are what phones break down into. [m] has a set of components, including [+ consonantal] and [+ nasal], and so does [a] and [x] and [ç] and every other phone in every human language. Of course, we all share the same phone-production equipment, so all sounds have to be able to break down into features that are potentially parseable by any other human brain or vocal apparatus. In speech, the presence of two phones in close proximity can make them fuse together through the removal of features from one phone and their coming together with other features from the other phone.
Crowley gives the example of a common sound change in ancient Greek: the replacement of [gʷ] by [b] due to the fusion of their features. [gʷ] is two sounds - a velar stop [g] and a bilabial semi-vowel [w]. These fused to become a bilabial stop. (I believe this also happened in Celtic, where [gʷ] also became [b] - cf. Irish bó, 'cow'. I'm not an expert on Indo-European, by any means!)
*gʷous > bous 'cow'
*gʷatis > basis 'going'
...and so on. The voiced velar stop in [gʷ] turned into [b] (a voiced labial stop); in the case of the voiceless velar stop with the same bilabial semi-vowel [kʷ], it turned into [p] (a voiceless labial stop). The [k] in cow is the result of a different sound change in Germanic from the same [gʷ] phone, so even though cow and bous don't look all that similar, they are in fact genuine cognates.
I like this example a lot, primarily because it's so counter-intuitive. It was only worked out because linguists took the time to establish the features of these sounds when comparing Indo-European languages, and it feels a little like we're unlocking the deepest mysteries of human sound production to make sense of this historical connection. Check out the Wiktionary list of words derived from the same Indo-European root for cattle, by the way. Look at guey, in particular, which functions as a Mexican slang word for 'dude', but exhibits a similar sound change as the one above, but in reverse, sort of. Sound changes are crazy. Crazy, but retrodictable.
Crowley uses an example from Old Irish to demonstrate compensatory lengthening. That's a kind of a fusion that depends on one of the features of the phoneme being its length, or the space it takes up. Let's say that each sound takes up a certain amount of space, and that the reduction or absence of a sound removes a certain amount of space from the word. Now imagine that this space feature of the phone has been added to the phone that has been left behind. This is compensatory lengthening.
To make this clear, imagine that you've got:
[+ short] (e.g.)
Then you'll end up with:
*magl > ma:l 'prince'
*etno > e:n 'bird'
I'm much more familiar with Mongolian than Old Irish (I studied it at university for three years), and it shows a similar trend when compared with the classical Mongolian found in the Secret History:
ulagan > ula:n 'red'
xagan > xa:n 'Khan'
You may also note that the pronunciation of 'Magdalen' as ['mɔːdəlɪn] - as in, the Oxford college - includes an example of compensatory lengthening, the same one as occurred in Old Irish.
This is a kind of fusion, even though it doesn't seem like it. It's the fusion of a feature that doesn't seem strictly phonetic, although of course it is (and many languages have a distinction between long and short vowels, indicating that phone space is an important feature). Fortunately, it's one of those sound changes that is obvious even to people who know nothing about phonetics. In fact, in reading modern Mongolian in classical Mongol script, the [g] is left in. In classical script, which is seldom used in Mongolia, Ulaanbaatar is Ulaganbagatur ('red hero').
Fission is the inverse of fusion. Where fusion takes the features of phones and jams them together, removing some of the features, fission splits the sounds and puts features back in. Crowley's first example is, unsurprisingly, from Bislama, Vanuatu's lovely creole and subject of much of Crowley's research. The French word camion, meaning 'truck', ends in a nasal vowel and has a palatal approximant before it. In its borrowed form in Bislama, it ends in a nasal consonant (a velar nasal) and the palatal approximant has become a vowel, [i] - [kamioŋ].
One common type of fission is vowel breaking. This involves the change from a single vowel into a diphthong due to the addition of a glide, or semi-vowel ([j] or [w] or similar). When the glide goes before the vowel, it is apparently referred to as on-glide, and when it comes after, off-glide. It is especially noticeable in some dialects of American English, which break vowels I never would - e.g., [bæd], 'bad', becoming [bæʲəd]. Crowley also notes that Barbadian English typically employs an on-glide before the vowel [æ] - [kjæt] for 'cat', etc.
Crowley's example comes from Kairiru, a language of Wewak, New Guinea.
*pale > pial 'house' (proto-Austronesian *balay, cf. Hawaiian hale, Samoan fale, Lamalera bale)
*manuk > mian 'bird' (cf. Filipino manok)
*ɲamuk > niam 'mosquito' (cf. Indonesian nyamuk)
You'll note that niam also exhibits apocope.
I was going to continue on with assimilation and dissimilation today, but that would probably be counter-productive. So assimilation and dissimilation will have to wait for the weekend. In any case, fusion and fission are important principles, and it's good to get them right. Assimilation might detract from that.