Posted: Jul 05, 2012 10:31 pm
by Zwaarddijk

Why do people tend to equate it with morphology? Grammar is an often misunderstood part of linguistics. Some people equate it with speaking properly, some people equate it with rich morphology, etc.

It's not unusual to find claims - even on fora devoted to languages - that e.g. Chinese lacks grammar. It is true that Chinese has a relatively minimal morphology, but morphology is not all there is to grammar. Morphology just happens to be the most visible form, and the easiest form to tabulate and show people.

Consider Latin grammar, what's the thing most people who have learned Latin in school associate with grammar? Generally, it's big tables of inflections. Some other languages with rich morphologies are Classical Greek, modern Russian, Church Slavonic, Sanskrit, Finnish, Arabic, Hungarian, Turkish, ...

Now, if we compare the words "mother" in Finnish and Chinese, we get

媽 / 妈 (ma), vs.

sg. nom äiti pl.nom äidit
sg. part äitiä pl. part äitejä
sg. acc äiti/äidin pl.acc äidit
sg. gen äidin pl.gen äitien
sg. transl äidiksi pl.gen äideiksi
sg. ess äiti pl. ess äitei
sg. ela äidiltä pl. ela äideiltä

where you get about 28 basic forms, then you can add suffixes to pretty much every one of those encoding possessor:



On top of this, you can add some discourse particles like
talossanihan ≃ in my house, though, ...
talossanikin ≃ even in my house
talossaniko? ≃ in my house?

This looks like a lot of grammar, whereas the Chinese example is just quite unimpressive: one unchanging syllable. And of course, the Finnish example is quite like the traditional examples we get in Latin or Greek classes, a kind of grammar that still has etched itself in the mind of grandkids of people who studied Latin back in the early 20th century or whenever. The reason this kind of grammar tends to get our attention is that it's very easy to tabulate it, it's easy to explain it, and due to the often combinatorial nature of it, it's easy to calculate a fine round number and say something along the lines of ~Finnish nouns have six thousand forms~ (note: that number is wrong. I'll also try to show you why such a number is a bit irrelevant).

First, we need a concept: a unit of "part of a word that carries meaning". In English, "cars" consists of two such units - we can subdivide it into "car" and "-s", such that both carry some meaning relevant to the word, whereas "ca"-"rs", doesn't, nor does "c" - "ars". We call "car" and "-s" morphemes. A word such as antidisestablishmentarianism has several morphemes to it:
anti - dis - establish- ment- ari- an- ism. Some of these can occur in isolation.

The absolute majority of plurals in English use the -s desinence to mark plurality. However, there are some exceptions: oxen, fish, mice, men. So plural morphemes in English are {-s, -en, -Ø, -[some weird things that change vowels in the word]}. The different morphemes here don't correlate with any grammatically meaningful categories of English, nor are they triggered by the phonetic environment they occur in. English does have at least one desinence where variant forms are triggered by the context: the past tense -ed. Anyways, variant morphemes that express the same thing but in different contexts (e.g. in different words, in different phonetic environments, etc) are called allomorphs. Finnish is rich in allomorphs due to its vowel harmony rule: words with {ä,y,ö} in it will take no suffix with {a,o,u} and vice versa. So, e.g. -lla will occur with words where there's a, o or u, -llä will occur with words where ä,y,ö occur.
There's another allomorphy going on in the Finnish examples I gave: äiti- vs. äidi-. The conditioning factor is mostly whether the final syllable is closed (although the possessive suffixes don't trigger that for some reason). We can't just count the number of different endings: we need to identify which allomorphs belong to the same morpheme.

(In Finnish, there's 5 possessive suffixes + no suffix, 14 cases, two numbers for a total of 168 combinations, but not all cases permit having the empty suffix. In addition, one of the cases does not have singular forms. In addition, there's a number of almost-cases, some nouns have two forms with slight differences in connotation but the same basic case (e.g. lapsena, lasna - both essives of child, one a bit more poetic and cutesy), etc, so calculating it is not as simple as taking the size of a cartesian product of the categories expressed morphologically. The discourse suffixes as well multiply the number of available forms, and we could make huge tables for these. Finnish verbs would permit even greater variety, with a rich number of regularly derivative morphology. As for marginal cases, Russian is a good language: there's any number of cases that only seem to occur with a small set of nouns, and which basically are conflated with some other case for the vast majority of cases. The most standard one of these is the 2nd prepositional case for some nouns such as port, aeroport, les; normally, the prepositional singular ends in -e, but for those, the alternative case ends in -u.)

Now, languages with morphology - "synthetic languages" - come in two main forms: agglutinating and fusional languages.
In Finnish, case and number agglutinate. They have separate suffixes each, and you simply "glue" them on top of each other. auto-ssa = in car, auto-i-ssa = in cars, auto-i-ssa-ni = in my cars. Russian, however, merges number, gender and case - those caetgories are fusional in Russian. Essentially, a plural case ending may not bear any superficial resemblance to the corresponding singular ending, and feminine singular dative may have a suffix distinct from the masculine singular dative.

It's a bit stupid trying to compare the number of forms of Finnish and the forms of Latin or Russian for that reason: Finnish lines morphemes up, Russian kind of merges them.

In Finnish, you basically get something along the lines of (using set notation)
{sg (Ø), pl (-i-, -ei-)} x {nom (Ø, -t), gen (-n), part (-a, -ta), all (-lle), ill (-:n, -in, -hVn, -sVVn), abl (-lta, -ltä), ela (-sta, -stä), adessive (-lla, -llä), inessive (-ssa, ssä), trans (-ksi), ess (-na, -nä), abessiva (-tta), instructive (-in, ...), comitative (-ne-)} x {Ø, -ni, -si, -nsa/-an/-nsä/-än, -mme, -nne}.

In Russian or Latin, you need to look at it one step more abstractly, as there's less of a direct connection between the function of a form and its surface realization:
{nom, acc, gen, dat, instr, prep} x {masc, fem, neut} x {sg, pl}
However, there's also reason to think of this in terms of
{nom, acc, gen, dat, instr, prep} x {masc, fem, neut, pl}.

Some genitives and accusatives are identical: {acc, masc} is identical to {gen, masc} for animate masculines, whereas {nom, masc} is identical to {acc, masc} for inanimate masculines (and all neuters). {prepositional, feminine} and {dative, feminine} are also conflated. Some desinences (endings) are also shared between genders and cases: masculine and neuter genitive has the same ending, essentially, as feminine singular nominative. All singulars have the same locative: -e. The feminine genitive and the plural nominative are identical.
So, ultimately, we get a situation where the number of "functional forms" - e.g. actual case-number combinations - and the number of actual visually distinct forms are different.

The situation is further complicated by no noun having all the desinences appear on it. So, unlike the Finnish example, we can't just count the endings, put them in distinct sets along with what position they take and multiply together the sets to get a rough idea of the number of cases.

Having looked this much at the Finnish noun, and a little bit less at the Russian noun - not even getting into derivational morphology, one would think it's quite obvious Finnish and Russian have more grammar than Chinese - an unchanging 媽 / 妈 (ma) isn't comparably really impressive in any way.

What use do the Finnish forms have?
äidille - to mother, äidillä - by mother, at mother, in mother's possession, äidin - mother's, mother(object), äitien - mothers', äidiksi - (turning) into a mother, becoming a mother, äitinä - as a mother, being a mother, in the role of a mother, äidiltä - from mother, äitiä - (part.obj, neg.obj, ) mother, etc. Basically, these suffixes have roles fairly similar to prepositions in English or Chinese.
Finnish does have adpositions (both pre- and postpositions, in fact), so Finnish hasn't just stuck all their adpositions at the end of words, but what's historically actually has occured is that some adpositions have merged with the noun.

Russian cases likewise - but to a lesser extent - cover uses that syntax and adpositions do in English. E.g. the dative either replaces something along the line of "to (someone)", as in 'we gave a gift to her', or the adpositionless 'we gave her a gift'. In the English translations here, we have grammar as well - adpositions carry a significant bit of the grammar in English (but also in Russian and Finnish!). The Russian prepositions, as well as those in German collaborate with the cases, so e.g. Russian v + prepositional = at, v + accusative = to (location), na + locative = on, na + accusative = onto, German an + acc = onto, an + dat = on.
This too is grammar - but it's not as easy to tabulate this kind of rule-based grammar. Let's take another example of grammar that is not easy to tabulate:
In English, there's two kinds of phrasal verbs.

The first kind includes verbs like "to wait for", "to look at".
The second includes "to run up (e.g. a bill)", "to bring in", "to blow out".

Let's consider two example sentences:
John waited for her.
Bring in the guests!
John blew out the candle.
Eric walked out the door.
I looked at the painting.

We can rearrange some of them:
John blew the candle out
Bring the guests in!
but this can't be done with some of the other sentences:
*John waited her for
*I looked the painting at
*John walked the door out

Apparently there is some kind of difference here in what reorderings are permissible and which are not. Which ones are permitted and which ones are not is lexically determined - e.g. there's some phrases in the lexicon which are fairly well established as such verb-particle pairs, and some that aren't. Apparently, in English grammar there's a bit that says the phrasal verbs can rearrange like that, and verbs + particles that aren't phrasal verbs don't permit that kind of reshuffling. How do you quantify a bunch of grammatical rules like that? It's not easy to tabulate, so it can't be made to look impressive the way verb and noun endings like Latin or Finnish can. And it's less easy to spot mistakes in a language you only barely know if it's in those rules than if it's in the morphological bit - you simply often don't know the relevant rules, as some very far-out rule might be involved in making your example sentence wrong(ish), whereas in morphology, you usually have your paradigm, your declensions and conjugations, and sometimes a small list of specific exceptions. With rules regarding reorderings, what elements can combine, what happens when elements combine, etc there's just so much that can happen - and we can be fairly sure no grammar of any human language is fully described anywhere.

It turns out - and this is an important thing in some modern theories of grammar - that grammar and the lexicon interact quite a bit; in fact, it can be reasonable to even say that a fair bit of our grammar resides in the lexicon, basically words are not just sequence of sounds + [something that tells us which suffixes, infixes and prefixes it can take] + [meaning], but there's also a fair bit of + [things you can do with the word]. It's quite unlikely every speaker of a language has the exact same grammatical lexicon in their minds, sufficient matches are enough.

Another instance of grammar that might not seem quite obvious in a language that has it - and this is known from many languages, and reconstructed for proto-Indo-European: in some languages, subjects and objects are differentiated by morphology (i.e. a case) (English does this in the personal pronouns). In other languages, word order (English, Chinese, Swedish, loads of other languages) or some particle (partially in Spanish, Biblical Hebrew, ...) serve to distinguish the subject from the object. In some languages, however, neither of these strategies is used, yet speakers can always pinpoint with very great accuracy which constituent is the subject and which is the object.

So what's used to determine which is which? Turns out many such languages have a hierarchy of words. In a sentence with two constituents that 'vie' for subjecthood, the one higher in the hierachy always wins. This hierarchy often correlates with some notion of animacy, but there may be exceptions. If the verb is intransitive (i.e. cannot take an object), you'll only have one candidate, and no such strategy is needed. Naturally, oftentimes the more animate noun will be the subject - although this to some extent also depends on how things generally tend to be expressed in the language and on what other
The word 'water' comes in two inherited versions in Indo-European languages, some are cognate to the English word - water, vod, vatten, ὕδωρ (hydor), udens (Latvian), some are cognate to Latin aqua, eau, ujë, å, ...
There's good reason to think one of these were higher in the hierarchy, and hence used when it was a transitive subject, and the other used when it was the object. This is a situation not unusual in languages with such hierarchies, if some noun low in the hierarchy also happens to actually be perceived as a subject every now and then. Water is a thing that easily can be perceived as a thing that does things: it covers things, it takes people, etc.

Let's play a bit with an idea here. Let's say we have two typologically quite different languages, and we try to figure out which one has more grammar. How many noun pairs like the above does it take to be as much grammar as three case forms? As two genders? As a perfective-imperfective distinction in the verb?

Finnish marks telic, nonnegative transitive verbs' objects with its accusative case. All telic negative, atelic negative or atelic nonnegative verbs take objects in the partitive case. How does this compare to a language that has, say, distinct perfect and imperfect verbs, differentiated by some regular affix or stem change or auxiliary verb or particle?

Despite the fact that the Chinese word for mother - 媽 / 妈 (ma) - does not possess the same amount of distinct forms as Finnish äiti does, does not mean we can be sure Finnish has more grammar than Chinese. The kind of grammar that is difficult to tabulate is often also more difficult to research, and it's especially difficult to showcase to people that think Chinese lacks grammar.