Posted: Apr 14, 2012 3:29 pm
by Zwaarddijk
HISTORICAL LINGUISTICS I

There's a few different ways by which languages change over time. Words and set phrases change meaning, sounds in words are changed, and grammar is reanalyzed, acquires new building blocks and loses old ones.

I will start out looking a bit sound change, some of the driving forces and some of the results, as this tends to be the most regular part of linguistic change over time.

Now, it's quite usual that sounds affect sounds in their vicinity, in part because the mouth is a bunch of muscles and bones and teeth where the previous positions of everything may spill over a bit into the next position of things. I never explained all the secondary articulations and different things we can mark by diacritics when doing a narrow (phonetic) transcription of speech sounds. I will explain the ones I take as examples whenever they turn up.

- some examples of sound changes that have occured in different languages -
One very common sound change is that consonants between vowels get voiced - vowels are almost always voiced, and stopping the phonation for a fraction of a second might easily seem somewhat superfluous or even difficult to time right or a needless effort or whatever. So we get changes like Latin to Spanish vita → vida, lupa → loba, caeca → ciega (all of which are taken from Wikipedia's article on lenition, fyi), although these later have gone even further and the voiced stop has turned into a fricative.

Fortition - the removal of voicing and various other changes also can happen, and both conditional and unconditional varieties of it happen. An unconditional change happens everywhere to the sounds involved, a conditional one happens where some surrounding factors are satisfied.

One sound change that often happens is that g > dʒ and k > tʃ. This is even more common in front of front vowels (and sometimes, it's also happened after front vowels), so a word like /gin/ > /dʒin/ (English) or /kil/ > /tʃil/ (In Eastern Swedish, that is. In Swedish they speak in Sweden, this became some kind of variety of [ɕ] or such, although the exact pronunciation of the fricative there is under dispute... ). In both of these languages, in addition, /sk/ before front vowels have been turned into /ʃ/, c.f. proto-Germanic *skurtijon and modern English shirt. Funny enough, Scandinavian and English did this sk>ʃ change at different times, and in the time after English did it, English also borrowed skirt from Norse - which originally was the very same word. Nearly all /sk/ before front vowels in English are of Norse origin. And nearly all of these have turned into ʃ or similar in Scandinavian dialects (although dialects that have conserved the sk exist, as well as ones that have had different changes resulting in [ʃtʃ] or [stʃ] or [sts] or [stɕ]).

That change is rather obvious, front vowels pull back consonants forwards, and along the way the manner of articulation slightly changes. One might assume that the /i/ in part acquires more closure or whatever, and this is in part the cause of the ʃ/s/ç (postalveolar|alveolar|palatal) fricative.

Similarly, ti (especially if the i is relatively weak and part of a diphthong or hiatus (i.e. two vowels forming a sequence with no consonant between), the t tends to be slightly retracted and the i turned more consonanty, giving cç or tʃ or even just ʃ or ç. English orthography carries traces of this: all -tion word have gone all the way to -ʃən due to reduction of the vowel as well.

A less obvious change that is fairly well attested around Europe and elsewhere is the following:
ku > pu
ko > po

k and p are fairly far apart, so what's up with that? /p/ is obtained by closing the lips together during the air-stream, /k/ is obtained by closure at the soft palate. Now [u] and [o], which are present in very many languages, are rounded vowels. This rounding easily spreads to surrounding consonants, so although we perceive these as /ku/ or /ko/, what's really happening tends to be more like [kwu] and [kwo]. This rounding is a labial coarticulation, w marks on a consonant that it's pronounced with rounded lips. Turns out [kw] is articulatorily and acoustically somewhat similar to p, and at some point the speakers start confusing them or misproducing them often enough that what previously was /k/~[kw] now becomes /p/~[kw] (somewhat home-made notation - it marks that they start perceiving kw as an allophone of /p/, rather than as an allophone of /k/; so the other route this can go by if we assume a maximally simplistic model is mispronounciation and mishearing, in which case [kw] is replaced by [ p ] for whatever reason - that is, they genuinely don't reassign where [kw belongs, they replace it throughout.)

Another change that can happen to secondary articulations is that they become independent sounds: kw can become [kw]/[kv], tj can become [tj] or [cj] or tʃ. j marks patalization - a secondary articulation where there's extra closure along the back of the tongue/palate. In many Slavic languages, palatalized consonants and non-palatalized form distinct series, so pj vs. p, rj vs. r, tj vs. t are heard as distinct sounds and can distinguish two words. Spanish <ll> in some dialects is lj, and ñ is nj, but in many varieties, the l-part has been lost and the j-like part has turned into an actual j instead. Likewise, in some dialects of English, Polish and some south Slavic languages, lɣ - a velarized ( a similar secondary articulation, with quite a distinct acoustic effect) l, has lost its l-part and turned into more of a vowel or a w (which, despite being bilabial often has a velarized component in most languages that has it). Oftentimes in languages with a palatalized series, the non-palatalized sounds are slighty velarized to increase acoustic difference between them.

This [kw] > [p] change has happened in some Celtic languages, so you have mab / mac as a pair of cognate between P-Celtic and Q-Celtic (even though that division of the Celtic languages apparently is under some dispute, the Q-Celtic are the ones that retain a k-like pronounciation, and the P-Celtic the ones that turn kw into p, and gw into b; later sound changes may do things to p/b or kw/gw too, so that's why it's mab vs. mac, not map vs. mac.) [kw] can of course also lose its labialization, or keep it unchanged for a good while.

Another sound change that might be a reasonable example is z → r, which occured pretty consistently in the Scandinavian languages (to the point that they actually lack z now), c.f. English bit<->bits from Old English bita<->bitaz (or somesuch), Swe. bit<->bitar. Scandinavian has had pretty much the same plural morpheme as English -s on some nouns, bitar has been something along the lines of bitaz there as well. But this has consistently changed into r all over the place. Similarly, we get cognate verbs such as was vs. var. English has had a bit more restrictive rhotacism, c.f. was, *were* - where the s has been intervocalic, and therefore voiced, whereas it probably wasn't voiced in was at the time the rhotacism took place. Likewise, intervocalic s in Latin turned to r, and this also was conserved in the morphology: flos - florem, genus-generis ...

A thing with sound changes is that sometimes, the conditions that cause them are only present in some forms of a word. Here, a grammatically conditioned kind of change sometimes comes into action. Analogy finds patterns and strengthens them. This can cause the new sound to turn up in forms that shouldn't have obtained them by regular sound change. Or it can erase the new sound from forms where it should have turned up. Or it can even insert it into new words - c.f. how some strong verbs in modern English haven't historically been strong, and how some nouns with vowel changes didn't have them historically. Originally, forms like mouse > mice, man > men originate with sound changes. The factor that caused them has since been omitted, but we've learned these forms like they are, and so use them as such.

Important things here are:
- sound changes tend to be regular -
I.e. if we have a change where intervocalic voiceless stops are voiced, you won't find individual exceptions (unless such an exception comes about by analogy or such). Most of the time, you won't have "the sound change happens in these words, but not in these". This is just a rule of thumb, though, as we can find exceptions, and even pretty weird ones. Grammatically conditioned exceptions seem to be somewhat common, e.g. why Russian -ogo/-yego has turned into -ovo/-yevo if they're part of the masculine genitive suffix, but not otherwise. The neogrammarians of the late 19th and early 20th century rejected even grammatically conditioned sound changes, and attempted to find phonetic justification for all sound changes they accepted. This is probably a misguidedly strong stance, but there's good justifications for at least trying to assume maximally predictable and mechanistic changes. There's some very good results obtained this way as well, although it seems clear today that some changes cannot be assumed to have been entirely regular.

Basically, you could have a simple piece of deterministic software go through a list of words (and for most changes, the piece of software should even be ignorant of what part of speech the sequences of sound are part of), and apply the change whenever the right conditions are satisfied. Sometimes, the conditions may be fairly complex though. Latin rhotacism of s apparently was blocked by the presence of a rhotic in the word, yet there was a further rule that could unblock the blockage.

- sound changes are opaque to speakers -
A sound change can't depend on things that were present in a word prior to a previous sound change. You can't have a sound change that says "voiced stops that originate with previously unvoiced stops that were voiced during the intervocalic voicing now turn into voiced fricatives, whereas other voiced stops remain stops". Sound change has no memory.

This too has some exceptions. Different dialects go through slightly different sound changes, and a word we find in a language that seems not to have adhered to an expected change may have been borrowed from some dialect that never had the change, or vice versa, a word with unexpected changes in it can have gone a similar route. (I leave talking about what a dialect even is to a later stage, the concept of dialects is of minor interest for this topic.)

- sound changes generally are acoustically or articulatorily motivated -
But chains of sound changes can obscure such relations, as the next sound change moves it further down the chain. The strangest regular correspondence I know of is Armenian erk, from Proto-Indoeuropean *dw. What we perceive as acoustically similar might be language-specific, though: in a language with very few consonants, a lot of the available space for consonant variations might sound similar to other consonants because the speakers are not used to spotting minute acoustic differences.

- sound changes can merge sounds, move individual sounds in a word from one phoneme to another or create new sounds -
If we know a language has had a sound change that turns voiceless consonants intervocalically into voiced consonants, and we have a word "aba", we can't know if this was apa or aba before the sound change unless we have some other evidence of it - a dialect that hasn't had the sound change, some inflected forms that don't have it, a written source that predates it (and is in a reasonably pronunciation-like spelling), a conservative orthography (although this is not necessarily accurate either), recordings, or other more indirect proof. A sound change can fully merge two sounds that previously were considered distinct in all positions.

- sometimes, sound changes cause chains of changes -
This is especially common with vowels; one vowel phoneme migrates a bit in the vowel trapezoid, and another vowel moves away from it not to be confused with it (and possibly another moves slightly into the space it left behind as well), and a chain of changes like these pull each other around the trapezoid's surface. The same does happen with consonants as well; often, similar changes happen to similar sounds. So e.g. dental/alveolar consonants followed by front vowels easily all get a palatal coarticulation. Since we want maximally different sounds AND maximally few different manners of producing them, we'll usually end up with some kind of compromise.

- sound changes can also entirely drop sounds out of words-
The most radical example I've heard of is complete drop of all non-stressed syllables. English and many other languages have weaker examples of this. Have you ever wondered where all silent vowels at the end of any number of English words come from? They've been there, they've been lost. Likewise, /h/ has been lost in some dialects of English in many positions. In fact, all varieties of English has had some extent of h-dropping, as the pronoun it originally was "hit". Conversely, hypercorrection has caused h to be added to words that never had it under the assumption that it was restored. Hypercorrection will be described more carefully later.

The sound changes given above are only a small set - sometimes, even the opposite change occurs, and these are far, far from all known sound changes - and there's probably way more possible changes around. (Sometimes, sounds come from nowhere as well - Spanish for instance, has inserted vowels ex nihilo into words that begin with certain clusters, c.f. Italian scuola, Spanish escuela.)

With these basic ideas we have some basic footing in it, although far from enough. Next time, I'll look at semantic changes and grammar changes, and then I'll look at Indo-European and the Uralic languages, how we know the IE languages are related, how we know the Uralic ones are related, and other similar things.