Sunday, May 12, 2013

Non-Linear Morphology and Language Change

I've been sitting on this one a while, trying to do some more research for it, but, as ever, the time just isn't available. Maybe one day someone will pay me to do this, and I can write more substantive blogs. Nonetheless, I will post what I have so far, which aims to dig a little deeper and get more substantive, addressing one of the central theses of this blog's existence: that conlanging can be of value to Linguistics. 

As a matter of fact, I recently listened to the conlangery podcast episode "How to Read Linguistics Papers", and it made me realize what should have been a very obvious fact: what I mean when I say that conlangs can be of value to Linguistics is that they can be of value to Linguistic Theory. The entry pasted below suggests one way that might happen.


The main constructed language, or conlang, I have worked on is the language of wider communication throughout the nations descended from a former empire, a bit like Latin in the centuries after the Fall of the Roman Empire. This naturally implies that there are other languages spoken at the time my stories take place, many of them presumably descendants of the main language, Paiodd. I have sketched out bits of notes considering what kind of changes might have occurred in each language, and how they might differ from the parent language and each other, but one of the issues that I have come across in doing so is what Paul Kroeger (2006) refers to as 'non-linear morphology'. That is, indicating grammatical relationships in words not by affixes or word order, but by changes to the stem word itself. We see this in English with many of our irregular verbs: run - ran, see - saw, etc. In other languages, this might take the form of a tone change, nasalization of a vowel, voicing or de-voicing of consonants, and other changes of features that do not involve adding or subtracting from a word stem.
In fact, the English system was the inspiration for the Paiodd system, prompting me to wonder what a language would be like if all verbs were 'irregular'. To be fair, the system in Paiodd is in fact regular, based on a vowel hierarchy which is always consistent. But I went a step beyond that as well, since nouns also undergo vowel changes, based on the same hierarchy, and moreover also experience voicing and devoicing, as well as fricativization of final consonants. All of these changes indicate different grammatical relationships. When we consider this historically, however, we have a bit of a problem.
In English, and perhaps in most languages that exhibit non-linear morphology, these changes were originally motivated by the effects of neighboring sounds. The i-umlaut of English, which is where our irregular plurals men and women come from, among others, occured due to the presence of a palatal vowel in the plural suffix, which in Proto-Germanic was *-iz (Smith 1896). This vowel caused the vowel in the stem to move towards its own place of articulation, changing /a/ to /ε/, ultimately resulting in our modern pronunciation of men. Likewise, I have heard of work going on in some of the Mixtec languages of Mexico analyzing their tone systems historically. It may well be that apparent irregularities in the tone system there are due to the phenomenon of 'floating tone', or tone sandhi, which in this case results in a different phonetic form for a different grammatical meaning, even if the difference is only one of tone. This raises the question of whether all such non-linear morphology is the result of historical processes. I'm not sure if we'll ever know the answer to that, since many of the languages exhibiting such features are minority languages either without any written tradition, or with one which has been lost or suppressed.
When it comes to my fictional language of Paiodd, however, the question is reversed. As far as I know, these non-linear forms are not the result of historical processes but simply the method which the language employs to show grammatical relationships. Thus, considering what kind of changes might occur in the daughter languages of Paiodd leads me to believe that these vowel changes would in fact be fairly stable. That is, the daughter languages would still show vowel alternation in stem words indicating grammatical relationships or meanings.
However, what those relationships or meanings are could very well change. If a 'weakened' vowel in Paiodd verbs indicated a very general irrealis state, perhaps a daughter language uses a weakened vowel to show only one specific irrealis state, say the future, as in Ex. 1 below, where PO represents Paiodd, and PoD represents a Paiodd Daughter language:

1) PO ta 'put', te irrealis 'may put' > PoD ta 'put', te 'will put'

Another thing that could change is the nature of the hierarchy itself. That is, the daughter languages might have merged some of the vowels that Paiodd considered distinct, which would have two possible results: either the relative strength and weakness of the vowels would shift to accomodate the missing vowels, as in 2a, or the stem would no longer change at all, as in 2b.

2a) PO hierarchy: ludd 'dog', uninflected > lod- weakened, lut- no change
PoD hierarchy, u merges with o: lodd 'dog' > lad weakened, lot no change

2b) PO hierarchy: vur 'speed, haste' > vor- weakened, vul- no change
PoD hierarchy: vor > vor, vol no vowel change in either form

In the first scenario, whatever the original vowel was would now change to an entirely different vowel from its ancestor language, as indeed happens in 2a. The second scenario would have to result from a merger of two vowels that were stronger or weaker than each other into one, so that the vowel simply doesn't change at all, because both historical vowels have changed into the one modern vowel, as in 2b.
Examples 2a and 2b were both examples of nouns, which also show vowel change to indicate grammatical relationships, and which would also change in the daughter languages. So where Paiodd uses a weak vowel and a voiced consonant to indicate adverbial force, this form could come to have a more specific meaning, such as only indicating the manner of the action, as in 3a, or a more general one, such as becoming the form for nominative and accusative nouns, rather than only adverbial.

3a) PO: levodía daemm sem armes. dagger.VBL-INSTR bread 3p cut
He cuts bread with a knife.
PoD: *levod daemm sem armes. dagger.VBL bread 3p cut
He cuts bread daggerly.

3b) PO: Piviría tuoss dhaia. king.VBL-with/of poem sing
The poem sings of the king.
PoD: *Pivir tuaz dhaia. king.NOM/ACC poem.NOM/ACC sing.
The poem sings the king.

Another thing that is quite likely to happen in the development of the Paiodd noun system is the loss of suffixes and affixes, at least in certain cases, so that the inflected forms of nouns would actually result in creating new parts of speech: adjectives and adverbs! We've seen this to some extent in Examples 3a and 3b. But here are some further examples:

4a) PO: lodd hapía. man size.NML-with
The man is big.
PoD: lodd hap. man big.
The man is big.

4b) PO: hapía lodd dhaia. size.NML-with man sing
The big man sings.
PoD: hap lodd dhaia. big man sing
The big man sings.

These changes don't even begin to touch on things like semantic drift, other phonological changes, or innovations. But most of these lie outside the specific concern of this blog post, which is non-linear morphology, and how it might affect and be affected by historical change.
In sum, then, the point is this: in some natural languages, grammatical relationships or meanings are indicated by changes of stems that do not involve adding prefixes or suffixes, but some other change to a word's stem. In at least some of these cases, these changes are historically motivated. However, in the case of my constructed language, there is no historical motivation for such stem-changing, non-linear morphology, and the language later split into several daughter languages. Such stem changes being so important to the grammar of the parent-language, it seems likely that they would remain somewhat stable, in form if not in meaning.
Considering such issues also raises the question of whether non-linear morphology in natural languages is always the result of historical phonological processes. As far as I know, this is not a question many people, if any, have pursued. Of course, there at least two good reasons for this. First, many languages that exhibit this are minority languages about which little is known today, and even less the farther back in time we try to go. Due to suppression and the influence of more dominant languages, these tongues may not have a long written tradition, and speakers may not value their own language at all. This makes it difficult to learn anything about the language at all, much less its historical development.
The second reason is that it's not an easy task to collate and analyze data from multiple, unrelated languages. The old saying, 'Jack of all trades, master of none' applies quite well here. Most historical linguists focus on a language family, for obvious reasons, while the kind of work it would take to answer this question requires vast knowledge of very unrelated languages. It may not be impossible, but it is certainly difficult, and would really have to be the result of collaboration between experst in multiple fields. There is the field of typology, which makes some attempt to do this, but I'm not as familiar with typological methods as I would like to be, so I can't sufficiently evaluate their effectiveness.
Nonetheless, the question of the origins of non-linear morphology is an intriguing one, and I think it should be pursued, if it hasn't already been. If anyone knows of studies pursuing this kind of question, or even just work being done on non-linear morphology or languages that exhibit it, I'd love to hear from you! I'd also love to hear about examples from other natlangs or conlangs of non-linear morphology, so feel free to comment with any you know about!
I close with this: it is an ongoing central thesis of this blog post that 'Hypothetical Linguistics', as I call it, can prove useful to the field of Linguistics as a whole. I hope this entry serves as an example of how that might work. By thinking about the historical implications of a particular feature in my conlang, I've come across an intriguing, and, so far as I know, unanswered question in Historical Linguistics. This is precisely one of the most important ways that I believe Hypothetical Linguistics can serve General Linguistics - simply rephrasing questions from the opposite direction. That is, rather than asking 'What do languages do?', we can ask 'What could a language do?' and from that arrive at further questions, which could serve to shape future research in Linguistics.

Works cited:

Kroeger, Paul R. 2005. Analyzing Grammar: An Introduction (Cambridge Textbooks in Linguistics).
Smith, C. Alphonso. 1896. An Old English Grammar and Workbook (Kindle Edition)