Posted: Jul 30, 2012 6:09 pm
by Zwaarddijk
Badly written, again, but sums up some kind of important things in way too simple a manner. Ask for clarification and it will be given. This is mainly just maybe a starting point for a discussion if there's questions. I've been a bit too busy again :|

It can trivially be shown that languages change. Just compare, say, a few newspaper articles from the 19th century and some from last year, and you'll find differences. There will be words in the older articles that have fallen out of use, but there'll also be words in the newer articles that weren't coined or whose usage was marginal during the 19th century. Some words will also have changed meaning in different ways.
But other changes will also have taken place besides the vocabulary-related ones. The style will have changed. That's mainly a kind of social thing: in the 19th century, newspaper articles were expected to be phrased and structured in a certain way, now they're expected to be phrased and structured in ways that are slightly off from that. The newspaper article is a kind of context in itself, where language is used in some specific ways that differ from the ways we use language when talking to our colleagues over lunch, when we chat someone up at a bar, when we confess to someone that we've done a terrible mistake, etc. That's a kind of meta-grammar, though - it specifies how we wrap things together in a discourse. This meta-grammar (if I may use such a nonstandard term) is studied in pragmatics.

If we go back a bit further, we find more obvious changes in things: word order may seem odd at times, there are unfamiliar endings and prefixes on the verbs (but some may seem somewhat familiar if we know, e.g., German), and we'll also find that some weak verbs were strong - as well as possibly the opposite. We may also find more and different suffixes on the nouns. At some point we also find things referred to by gendered pronouns as well as the neuter pronoun - and that the gender used for any given noun will be rather invariant through a period, but also that the gender used with a noun seems rather arbitrary.

It's easy to understand why the meanings of words change and why words come and go. (There are, of course, complications - generally complications that best are understood through sociolinguistics, but I am not opening that can of worms just yet.) New things are invented, old things fall out of use, new social structures appear that need describing, old structures fall out of use, words are generalized, become more specific ("meat" formerly included all foods, "deer" formerly included animals in general), or just switch around by some kind of association (the evolution of the word "gay" is a good example, the Scandinavian word "arg" is an example that kind of goes the other way around, it started out meaning something along the lines of "unmanly", developed to meaning"the submissive part in homosexual sex, or the victim of homosexual rape", then "someone who has been thoroughly disgraced", then through some other steps ended up meaning "angry" with no sexual connotations whatsoever.

It's also easy to understand why grammatical affixes disappear - they simply [well, not really "simply"] drop out.

Let's look at an English noun about a millennium ago:
- it inflected its nouns (and adjectives) for case. Any noun that was used got marked for the nominative, dative, accusative, genitive or instrumental (which was increasingly being conflated with the dative). These cases reflected different grammatical roles.
- the verbs were all inflected for person, even in the past tense.

Why's that drop out? Well, it's easy to attribute it to some simplistic explanation like, say
people don't bother with tedious affixes[/b], or [i]people didn't learn the rules properly or language tends to simplify over time. Consider, for instance, the verb suffixes - usually, you'd also have the pronoun present, so it was somewhat redundant. Another popular explanation of (apparent) simplification in grammar is foreign influence. Yes, foreign influence does both give and take in linguistic change, but it seems to me it's too often used as a recourse to explain any change. Languages do change even without foreign languages involving themselves in it. The polynesian languages, for instance, wouldn't have had almost any variety in them before the arrival of the Europeans if outside influence was a necessity for linguistic change. I will get back to why change occurs later in this post.

How that happens isn't difficult, but how does a language acquire or shed grammar? There's a few processes that change the grammar of a language. But before going there, let's refine our notion of what grammar is. Let's go so far as to try and come up with some kind of notion of what form grammar exists in!

Let's consider a bit what it means for something to be grammatical.

There's several ways of defining a language's grammar. We could, for instance, decide that what some authority decides is the grammar of English. We'd have some arbitrary group of people - say a bunch of English professors or somesuch, and whatever they say goes. Let's say a situation appears where you need to express something that isn't possible to say with the current rules of the language, or that isn't regulated whatsoever. Before we know if you said it right, the committee that decides would have to decide how that particular need is handled. Verifying whether something's grammatical would simply mean checking the most recent version of the grammar. Constructing a grammatical utterance would simply mean applying the given rules constructively. If the committee is real kind, maybe we're even given algorithms for these two problems.

So, who decides to give this committee this power? The queen? The congress? The joint heads of state of all anglophone countries? Or should there be separate committees for Canada, England, Scotland, Australia, ... ? Now, I am sort of highlighting here the absurdity of such an idea, yet it is true that some countries do have such committees, and these committees tend to fight a losing battle against evolution, although on occasion they do manage to change the course of the entire river that the language is. As a speaker of Swedish, I can point out that until relatively recently, the grammar of Svenska Akademien forbade a certain construction based on a reasoning that should have forbidden a few others, yet the grammar permitted those for no given reason whatsoever - it was inconsistent (other inconsistencies undoubtedly exist today). This particular rule has been omitted, but prescrivists still occasionally like sneering at you if you use that construction. Also, this organ that defines and refines the grammar of Swedish has not had a member to represent the 300 000 strong minority of speakers of Swedish from Finland.

In this case though, the grammar exists as a well-defined formal description. How people actually speak and write is not interesting, and speaking grammatically means adhering to the rules set fort in some formal description. A change is an order from the top to change some rule. You can't 'really' speak until you've learned a sufficient part of the rules.
If this were how grammars are defined, the majority of languages would not have grammars. Language change would come from the top down in an explicit form, and so on. Quite a bizarre and unrealistic description of the situation.

Another misconception of the situation that is rather popular is based in some kind of, I dunno, naive platonism or something. The kind of idea that "this is how English is, and that's that. There's no mystery to it, that's how it is."
But no god has stepped down on Sinai or any other mountain to lay down the law of English grammar - there's no external intangible arbiter or essence of English that exists in some ethereal manner that decides what English grammar is like. Believing in such a thing basically isn't that far from believing in ghosts. Yet lots of people who'd never be caught believing in something so preposterous, do believe there's a grammar of English existing in Plato's Cave. No different, really, from believing that God somehow works as an external guarantee of ethics or something, and somehow has revealed this to us in a collective revelation that some of us decide not to adhere to. Just as superstitious.

Let's try to think about it more rationally. Where does grammar reside? It resides in brains. Now, you can of course object that many speakers don't know when something's an object or a subject, or explain how anaphora bind, or even analyze their own usage of morphology reliably. That's not a problem, though: labels like object, subject, anaphora, 1st declension, etc, are labels, used to describe the processes that go on in our heads. They classify the different building blocks we use, and are then used to figure out what patterns we adhere to when we speak.

How does knowing where language resides help us figure out what the grammar of English is? Now, you probably share quite a significant bit of grammar with those around you - but there may very well be variations between you. One of you might not use a certain verb intransitively at all, another might permit making it intransitive without any care in the world. Someone might have another adposition with a specific phrasal verb, and someone might drop the adposition with that phrasal verb altogether. So we can't just study one person's grammar, and then say we've learned all of English. The reason for this, really, is that as pattern-matching machines, our brains sometime overanalogize, and sometimes, they miss patterns as well. Some time, you've undoubtedly uttered a strong verb as though it were weak or vice versa. And these kinds of things sometimes happen on a scale grand enough to make most speakers of the language change which of those two kinds of verb a verb belongs to. We need to get some kind of idea of what features unite the entire speech community.

Well, we could come up with a definition of English as some kind of weighted average (with some permitted deviation) of some kind of "sum" of all grammars of those who speak English. Of course, where do we draw the line for what English is, we still haven't got a grammar by which to decide! When is some dialect divergent enough no longer to qualify as a form of English? We look at how people identify what they speak, the history of how what they speak has come about, how well it's understood by people closer to the average, and to what extent there's chains of transitional forms of the language. Often there'll be borderline cases that can't be answered absolutely, and we can often also find subsets where people cluster towards some kind of centre as well - Scots might qualify as that in the case of English. Of course, we can't peer into everyone's brains and pick the grammar out and compare them. We need to use some other methods. Luckily, we can just sample. And we try to find patterns in an intellectual manner in the corpus we get, and since we have live informants, we can produce intentionally misproduced sentences and see if our hypotheses are right: a misproduced sentence should produce some kind of feeling of error in the native listener. And if most people seem to grasp what we say and write well enough, and it patterns well with what others utter in our vicinity, we can be sure oour grammar is fairly correct. Many common pet-peeves are actually mistaken analogies: 'begging the question', if you apply analogy and read it as though it were just a normal phrase would mean exactly what people mistakenly think it means. (Or something along those lines - it might actually be borderline malformed for some speakers, and this might set off a warning bell for some that there's something odd about this expression. If you try to find a meaning for it without having looked it up, the only meaning you would come up with is the meaning people do come up with.)

When it comes to figuring out how language works (and everyone does that at least for one significant bit of their life), we're in a bit of a double bind: language is used to communicate, among other things, what we think, but at the same time, to reliably construct a model of the language in our own head (esp. when learning our first language), we would need to know what the speaker thinks! To some extent we can, but there's still room for doubt about it. We're after all not telepathic beings, and we will parse an utterance the wrong way at times. Since our neural network doesn't get entirely static at some point, but keeps adjusting, this will also affect the language model stored in our brain. Our analogizing can fail simply because we're not smart enough to quite get what the other guy is saying, or we can project too much smarts on him, and fail that way as well.

This is, a kind of abstract, kind of concrete look at how the most realistic definition of grammar for real, spoken languages would work out.

So, how does grammar change, and where does the grammar this speech community share come from if it hasn't been handed down from a committee or God-on-a-mountaintop?

There's several processes we need to be aware of.
  • analogy
  • grammaticalization
  • sound changes merging or deleting elements with grammatical roles

Analogy is a very important process in language. Every now and then, we are going to utter a sentence never uttered before. We can't look such a sentence up in a list of accepted sentences of the English language - no such list exists. And none of us have ever seen an exhaustive list of all rules of English. And when we use a new word, it's seldom we look it up in some list of approved forms in order to inflect it right. Our brains probably don't generally do this by implementing a list of rules in some form, but by analogy to other things we've heard. Our brain is a neural network, and neural networks are good at pattern matching to some extent.

As it happens, English inflection nominal morphology is rather simple: you have a singular form and a plural form for all count nouns. We have a few occasional exceptional nouns - ox, oxen; man, men. A rather famous experiments showed that children relatively early have analogized the plural allomorphs of English, -/s/ after voiceless consonants, /-z/ after voiced consonants and vowels, and -/Iz/ after some consonants. The test asked children about plurals for various words that don't exist in mainstream English. The wug has kind of remained a fixture in linguistic lore ever since:


But as I said, analogy can draw different conclusions: someday, the strong verbs may very well disappear due to analogy with the weak verbs. Already, most colloquial English has reduced the modal forms of 'to be' to rubbles. Meanwhile, it's also likely a lot of syntactical assumptions that English had 600 or even 400 years ago may have been lost without anyone even caring to complain about them, since there were no scholars that had identified that kind of grammar back then, so no one even knew to whine.

Sound change is another important source of linguistic change. Not only does it shape our words - e.g. turning *skuldra into sholder into shoulder. It also changes our grammatical morphemes. For one, the loss of most un-stressed word-final vowels in English pretty much spelled out the end of the case system. Similar changes happened throughout western Europe, reducing the case systems of Latin and Proto-Germanic to just some occasional remnants in the pronominal systems. Similar changes also reduced the number of distinct verb forms. Keeping that in mind, it'd seem natural that morphology is doomed, over time, to simplify. But that's an untenable position: if that were the case, language should've been at its most morphologically complex when it began being spoken! A highly odd position, no?

Turns out we know a process that does the opposite (but also through the reduction of elements). Grammaticalization. One important form of grammaticalization is when worn-down function words get their actual meaning semantically bleached - they no longer carry any actual meaning as words, only their function remains. In English, the article system has developed since medieval times. Articles did not exist in the year 600 or thereabout. A/an is a reduced version of "one". I am not sure about the origin of "the", but I'd wager some demonstrative or such whose meaning got less demonstrative and only the definite part remained. And there's quite a bit of grammar in the use of "a/an". (C.f. the previous sentence, in some related languages, you'd say the equivalent of "a quite bit"; there's also a lot of grammar as to when you should omit the articles, which kinds of nouns don't take the indefinite article, etc. There's a whole bunch of grammatical rules in there, rules that probably no one even had a reasonably complete description of until maybe the 1950s, yet most speakers of English had used those rules fairly consistently with the occasional outlier for over a century.)

Grammaticalization may go as far as to create new morphology. The Scandinavian definite articles - which have also evolved roughly during the same time as the English articles - are suffixes. a car / the car translates as en bil / bilen in Swedish. This suffix did not exist when the rune stones were written. It's been worn down, and assimilated into the preceding word. Finnish with its 16ish cases descends from proto-Uralic, a language that probably had six cases. The Scandinavian passive verb originates with a grammaticalization of a reduced form of the pronoun "himself/herself/itself", "sig", so now, to passivize a verb in Scandinavian, you can just tuck an -s onto its end (instead of bothering with this tedious "is/.../was/were x:ed" periphrasis.)

All this happens through people either being a bit lazy and simultaneously doing what they have to do to keep up with language: analogy. And this all happens in a fluid democracy-like thing where majority vote (by usage) is kind of powerful, but minority votes can be maintained for long times within a speech community as well.

(that in kind of ~evolutionary~ terms describes how redundancy comes about)

It's easy to think that redundancy is somehow a flaw in a grammar. It's common among people who like to improve things to try and come up with improvements for human language that get rid of the redundancy, and also build it up more logically. This seems to be popular among engineers who are also into conlanging (the art of constructing languages.) There is an evolutionary reason for the redundancy though, and there's also an evolutionary reason why redundancy sometimes is dropped as well. We live in a noisy channel - at times, information gets lost en route. If there's enough redundancy, the likelihood that the mind can reconstruct the message is increased a bit. Sometimes we parse wrong even if we hear every word of the message clearly - maybe we assume the wrong context or the wrong intentions behind it or whatever. This kind of redundancy is encoded in many ways: in some languages, you have gender agreement, in some you have person agreement on the verb + mandatory subject and object nouns or pronouns present, in some, all adjectives are marked for the same case as the noun they modify, in some you have all objects of negated verbs appear in one specific case, in other languages double (and even triple or quadruple) negation is a way of ensuring that the listener doesn't happen to miss the negation particle. Almost all languages have some level of redundancy in how they encode stuff. Some sentences may very well lack that redundancy, and some may even permit ambiguity, but there'll usually be some strategies that introduce a bit of redundancy. And why is this? Well, imagine if your language is optimized to squeeze too much information into this noisy channel. People will increasingly often ask "excuse me, what're you saying?", until you get quite irate with it. You'll probably - possibly subconsciously, start repeating salient parts "the car mechanic, he did repair the car quite well he did'. Speaking in a way where people start misunderstanding or asking for repeated statements a lot will make people annoyed, and that's the battle of the fittest being waged here: a more efficient way of speaking will probably win out (if it's also easy enough to learn by hearing). Younger learners pick up this way of repeating some information that may help out and at some point there's sufficient amounts of redundancy, so no one keeps adding any more of it as they'll sound like someone who can't shut up and can't keep quiet and really sucks at succinctness. At some point, sound changes start wearing it down and all of a sudden a former pronoun is now a part of the verb as a subject or object marking on the verb, maybe an adposition has merged with a noun in forming a case inflection, and so on. This is of course just a narrative, but similar narratives easily can explain most of the ways how languages gain and lose morphology and redundancy and other features. Of course, normally this'll be driven by some typological tendencies, e.g. a language where the verb tends to go first will almost invariably have the subject go before the object, permit that the verb goes in the middle between them on occasion, will have prepositions instead of postpositions, will almost invariably have adjectives go after the noun, and so on. Why this is is a more difficult thing to explain, but it tends to influence how a language changes over time. Of course, a language with verb-subject-object order easily can change into a SVO language (as that order almost always is permissible in the language, it just needs to become a strong enough fad). We know of languages that have gone from SOV to VSO, and vice versa. Languages with the object preceding the subject are unusual, but all three variations of that are attested as well.