Linguist Forum

Specializations => Morphosyntax => Topic started by: zaba on March 22, 2014, 09:11:59 AM

Title: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 22, 2014, 09:11:59 AM
Let's say I hedge that a language X lacks adjectives. I show you that they adjectival forms actually decline like nouns (when they are heads in a NP -- but otherwise are undeclined). I also show that if they take tense, they must verbalise first.

e.g. 'good'
GOOD-comparative.case = 'like'
1subj-3obj-make.verbalizer-GOOD = 'I heal her.'
GOOD-pl = 'good (ones)'

Is this a convincing analysis? If not, what would you need?



Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Jase on March 22, 2014, 11:52:42 PM
Let's say I hedge that a language X lacks adjectives. I show you that they adjectival forms actually decline like nouns (when they are heads in a NP -- but otherwise are undeclined). I also show that if they take tense, they must verbalise first.

e.g. 'good'
GOOD-comparative.case = 'like'
1subj-3obj-make.verbalizer-GOOD = 'I heal her.'
GOOD-pl = 'good (ones)'

Is this a convincing analysis? If not, what would you need?

You've completely lost me. Are you saying that verbalizing "good" becomes "heal"? That's odd. Why wouldn't it be "do good," "treat well" or "perform good deeds (for)"?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 23, 2014, 12:05:10 AM
sure, Jase. GOOD + the MAKE-Verbalizer (a sort of factive) transforms `good' to `make good' = in the sense of heal.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Jase on March 23, 2014, 01:07:15 AM
sure, Jase. GOOD + the MAKE-Verbalizer (a sort of factive) transforms `good' to `make good' = in the sense of heal.

What language are you talking about? You must be referring a specific language that does this.

I’m a speaker of Hebrew, and in Hebrew we have causative structures that can be added to roots to create this sense. For example, the root ט.ו.ב (ṭ-w-b) forms the adjective טוֹב (ṭôḇ “good”), the adverb הֵיטֵב (hêṭēḇ “well”) and the verb הֵטִיב (hēṭîḇ “he did well/good”). All three are tied into the single root, and this happens a lot in the language.

However, the concept of “healing” is tied into a different root (ר.פ.א r-p̄-ɂ). In what language is the root associated with “good” also associated with the concept of “healing”?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 23, 2014, 01:47:24 AM
it's not so surprising if one considers good/well + factive = make-well (as in 'the doctor made me well again')... anyway, this conversation is getting derailed from my original question!
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on March 23, 2014, 03:08:35 AM
How would you say something like "Pass me the blue book." or "The book is blue."?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 23, 2014, 04:54:15 AM
Quote
"Pass me the blue book." or "The book is blue."?

Below!
Blue book-accusative 2subj-1obj-pass
Book blue

Note how in the first sentence, the word book is accusative marked, but not blue. In a np, the head is obligatorily marked for case and the modifier is optionally so.


Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on March 23, 2014, 06:00:39 AM
I'm not really following, I must confess. That looks like a pretty run-of-the-mill adjective to me.

(And, just as a side note: it's sometimes very tough to read your glosses because (a) you're leaving out the source language transcription and (b) you're using non-standard labels. The Leipzig glossing conventions (http://www.eva.mpg.de/lingua/resources/glossing-rules.php) are what most disciplines use.)
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 23, 2014, 06:53:19 AM
My argument is simple (even if my glosses aren't!)

1. Adjectives can receive nominal morphology (e.g. plural and case declension); for example big-PL 'the big (ones)'
2. Adjectives can also receive verbal morphology (e.g. tense) but must first receive verbal affixes, for example, big-PL-VERBALIZER-3sg 'they are small'

Therefore, adjectives are actually nouns, i.e. there is no separate category of "adjectives" in this language.
Is this a cogent argument?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Jase on March 23, 2014, 07:10:46 AM
My argument is simple (even if my glosses aren't!)

1. Adjectives can receive nominal morphology (e.g. plural and case declension); for example big-PL 'the big (ones)'
2. Adjectives can also receive verbal morphology (e.g. tense) but must first receive verbal affixes, for example, big-PL-VERBALIZER-3sg 'they are small'

Therefore, adjectives are actually nouns, i.e. there is no separate category of "adjectives" in this language.
Is this a cogent argument?

I have to agree with the user above. big-PL is certainly not English. What language is it from? Hebrew has adjectives that can act as substantives (NPs) without fillers like one or ones in English. Could you lay out exactly what language you’re referring to and give examples in the language itself along with glosses to explain what you’re trying to describe? Are you attempting to play silly games with us? What’s up
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on March 23, 2014, 07:21:38 AM
Most languages provide ways to shuttle morphemes between word categories. Sometimes the transformation adopts a phonetic exponent ("small-ness"), but sometimes not ("three smalls").

To argue that there aren't actually lexical adjectives, on the other hand, requires quite a bit more information. Specifically, you want to analyze the distributional criteria of semantic "things" and "properties". I suspect you'll find some restrictions, though it's possible you won't.

In any case, if you have phrases like "the blue book", the fact that you might also have a phrase "the blues" doesn't actually suggest that "blue" was covertly a noun all along. That's where I'm losing you I think. If this language doesn't have adjectives, and consequently if "blue" in "the blue book" isn't an adjective, what is it?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on March 23, 2014, 07:25:05 AM
I have to agree with the user above. big-PL is certainly not English. What language is it from? Hebrew has adjectives that can act as substantives (NPs) without fillers like one or ones in English. Could you lay out exactly what language you’re referring to and give examples in the language itself along with glosses to explain what you’re trying to describe? Are you attempting to play silly games with us? What’s up

Easy there...no need to start slinging around accusations.

Many languages (including English) nominalize adjectives. In fact, big-PL ("the bigs") is perfectly familiar English among aspiring baseball players.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Jase on March 23, 2014, 10:20:49 AM
Easy there...no need to start slinging around accusations.

Many languages (including English) nominalize adjectives. In fact, big-PL ("the bigs") is perfectly familiar English among aspiring baseball players.

I just feel odd that we’re not being clued in on what language the author of the post is talking about. He’s asking in generalities that would be better handled if we knew what language it was and could tackle specific examples of phrases from the language. It’s certainly a valid question that’s been evaded until now.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: ibarrere on March 23, 2014, 02:01:07 PM
I just feel odd that we’re not being clued in on what language the author of the post is talking about. He’s asking in generalities that would be better handled if we knew what language it was and could tackle specific examples of phrases from the language. It’s certainly a valid question that’s been evaded until now.

There is a benefit to talking about language in a general sense, too. The point of the exercise, I think, is to address how to talk about such a phenomenon without being bound to a particular language's functionality. Much of theoretical linguistics attempts to define universals, so I don't see it as a problem to omit real world examples, hypothetical examples can be just as powerful.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Jase on March 23, 2014, 03:25:05 PM
I just feel odd that we’re not being clued in on what language the author of the post is talking about. He’s asking in generalities that would be better handled if we knew what language it was and could tackle specific examples of phrases from the language. It’s certainly a valid question that’s been evaded until now.

There is a benefit to talking about language in a general sense, too. The point of the exercise, I think, is to address how to talk about such a phenomenon without being bound to a particular language's functionality. Much of theoretical linguistics attempts to define universals, so I don't see it as a problem to omit real world examples, hypothetical examples can be just as powerful.

Do you think it’s as useful as pulling examples from several languages (for the purposes of demonstrating a universal character of grammar) or from a single language (to demonstrate its uniqueness from other languages)? That seems more powerful and more meaningful. I mean, we are all apparently having a hard time grasping what the poster’s point is, since he is using examples that are not true for English (such as adding a verbalizer to “good” to mean “heal”). It is hard to imagine that a language actually does that without receiving an example of a language in which that is true.

Perhaps it’s just a lack of clarity in the posts. Does he mean to say, for example, that a doctor “makes someone well” as a way of saying that she “heals” the person? In that case, we should probably take “well” in a homonymous sense, seeing that it can mean both good.ADV or “healthy.” Does the poster mean something else? I’m having a hard time following the line of reasoning.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 23, 2014, 11:12:22 PM
Quote
In any case, if you have phrases like "the blue book", the fact that you might also have a phrase "the blues" doesn't actually suggest that "blue" was covertly a noun all along. That's where I'm losing you I think. If this language doesn't have adjectives, and consequently if "blue" in "the blue book" isn't an adjective, what is it?

Well, there are adjectives in the sense that if a NP is comprised of N+N, it is always the first N that is the modifier or adjective -- but I think this is syntactic and contextual, not by virtue of lexical categories.

So any "adjective" can be a head, can take any nominal morphology, and can do anything a noun can do.
Apparently, however, this is a weak argument.

Quote
To argue that there aren't actually lexical adjectives, on the other hand, requires quite a bit more information. Specifically, you want to analyze the distributional criteria of semantic "things" and "properties". I suspect you'll find some restrictions, though it's possible you won't.

Can you give me a criteria that would get me started. I'm not convinced of this argument anyway, so I'd be eager to be proven wrong in my hypothesis.  ;D

Thanks!
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 23, 2014, 11:16:46 PM
Quote
I just feel odd that we’re not being clued in on what language the author of the post is talking about. He’s asking in generalities that would be better handled if we knew what language it was and could tackle specific examples of phrases from the language.

Tough, isn't it? Welcome to my world. I'm working with a couple different languages these days. This (potentially) adjective-less one is an Amazonian isolate. No description available. It is an agglutinative language surrounded by highly synthetic tonal ones.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 23, 2014, 11:27:32 PM
Jase,

Quote
Perhaps it’s just a lack of clarity in the posts.
Forgive me! It's tough stuff to understand for me, too.

Quote
Does he mean to say, for example, that a doctor “makes someone well” as a way of saying that she “heals” the person? In that case, we should probably take “well” in a homonymous sense, seeing that it can mean both good.ADV or “healthy.” Does the poster mean something else? I’m having a hard time following the line of reasoning.

As for that example, it's not strange at all if one considers (1) that there may be no adjectives in this language so GOOD and WELL are the same words; and (2) that this verbalizer is a factive. Thus good/well + factive = 'make good/well', which, when said of animate nouns means 'heal'. You see, translation isn't always so straight-forward in languages which differ from your native one!

Anyway, again, I beg your pardon if my examples are difficult to understand or if it's hard to get my point. I am using natural language data, not textbook examples so that makes things more complex perhaps.

In the subject to this forum I asked if there are nec and sufficient conditions for a language to fulfil  in order for it to be considered as lacking adjectives. Many languages of the world lack adjectives. In some, the adjectives are verbs and in others nouns. This is, at least in part, a typological question.

I've always been happy to give other examples or provide additional data. I really enjoy getting the opinion of experts like yourselves and learning from you. However, not everything is unambiguous in this field -- especially when it comes to meanings ;)
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on March 23, 2014, 11:30:22 PM
Can you give me a criteria that would get me started. I'm not convinced of this argument anyway, so I'd be eager to be proven wrong in my hypothesis.  ;D

Thanks!

I'll address your first part at some point later if I have a moment, but here's a simple test for the second.

Your skepticism about adjectives seems to come from the fact that semantic "adjectives" can function syntactically as nouns. In other words, you can have "a book" and "a blue book", but also simply "a blue".

The important question then becomes: can you have "a blue that is book"?

Edit: type-o
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 24, 2014, 12:53:23 AM
Quote
The important question then becomes: can you have "a blue that is book"?

Hard to imagine such a context, you mean for colour description, a construction a kin to: `grass-ish green' (as opposed to grass-green)? If so, yes (though for this lang it'd simply be grass-green)
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on March 24, 2014, 09:17:39 AM
Malfet's point is a good one. In languages said to have no classes, anything can be an argument or predicate. So the two sentences are possible:
Man student.3SG 'the man students'
Student man.3SG 'the student mans'
Nonidiomatic translations to get the point across, meaning of course "...is a man". Likewise:
Eat man.3SG 'The eater mans'
Happy man.3SG 'the happy one mans'

So if you want to show lack of classes show that all nouns can be adjectival and all adjectives can be substantival. If so, there's no distinction at all.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 24, 2014, 09:58:13 AM
OK, I'll take it:
man student-verbalizer-3sg 'the man is a student'
student man-verbalizer-3sg 'the student is a man'
eat-agentive man-verbalizer-3sg 'the eater is a man'

But that's not the point: nouns and verbs are certain separate classes (after all, there are suffixes that nominalise and suffixes that verbalise) -- but the point is ADJECTIVES are NOUNS.

Can you help me dis/prove this hypothesis?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on March 24, 2014, 10:13:45 AM
I understand. You'd use similar tests for those classes. That's just what came to mind. There's an exercise in Carnie's textbook (from Salish?) that shows sentences like that. You should be able to do the same for A/N. Mallet gave one example.

Also, you could look at David Gil's research on Riau Indonesian, which he says has no classes at all.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 24, 2014, 11:30:52 AM
I'll check it out -- I'm not familiar with Carnie's textbook though. What's the title?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on March 24, 2014, 01:38:23 PM
Syntax: a generative introduction.

I'm thinking of the slightly older edition not the new 2013(?) one.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 24, 2014, 11:00:07 PM
oh sheesh does this go into some generative account of semantics? to be honest, i'm kind of troubled that no one on this forum can give me necessary and sufficient conditions to determine if there are a category of adjectives in a language. Nothing personal guys, but I didn't think it was a tough question.  You can ask anything about the language and I can provide working translations... think of it as a fieldwork exercise!

Or is the training these days so much more theoretically-orientated?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on March 24, 2014, 11:20:37 PM
Quote
oh sheesh does this go into some generative account of semantics?
Syntax? And, no, it doesn't. It's in a theoretical textbook, but that's what you're doing-- theoretical linguistics, based on distributional criteria for word classes. If you don't understand that much (whether or not you know/accept the rest of the theory), then you cannot possibly answer the question you're asking.

Quote
to be honest, i'm kind of troubled that no one on this forum can give me necessary and sufficient conditions to determine if there are a category of adjectives in a language. Nothing personal guys, but I didn't think it was a tough question.  You can ask anything about the language and I can provide working translations... think of it as a fieldwork exercise!
Translations are far from sufficient for grammatical analysis. If you use translations, you'll find that mysteriously all languages have the same word classes as English! Further, if you rely at all on semantics, you'll also run into major problems ("destroy" vs. "destruction" for example). This is covered in an introductory textbook like Carnie's (among dozens of others).

Quote
Or is the training these days so much more theoretically-orientated?
What do you think you're doing? Where do classification criteria come from?


In the end, here's where we are: you aren't using a theory, but you want to know what the arbitrary words "noun" and "adjective" mean enough to determine whether a certain language has those classes. How can that be determined except by theoretical criteria?

Regardless, a simple answer has been provided: word classes are distinct if there are situations in which a word from one class cannot be used in place a word from another class. For example, prepositions and conjunctions are distinct in English:
I walked in the store.
*I walked and the store.
(That should be obvious, but also a clear explanation of why.)

Now you just need to look at the language(s) you're working with and see whether nouns and adjectives (based on an initial, rough approximation from semantics) are interchangable in all cases. Are there times where only an adjective can be used?


There is no better answer than that, unless you're working in a certain theory (in which case that will still apply, so you may as well rely on it, given that the theory could be wrong). There aren't universal rules out there for what a noun is or what an adjective is.
And that's the problem: we can't do anything (like come up with N&S conditions) for nouns or adjectives unless we know what they are; to work around this, we can assume that such classes exist and then test whether they're distinct.


Personally, I'd recommend reading Carnie. The first couple chapters aren't especially "theoretical" in the sense you're using the word, and they cover all of this in detail.



Finally, on a related note, you may want to consider whether you're talking about syntactic function or lexical nature. Do these words have classes in the lexicon? Or do they just function one way or another in a sentence? This gets very complicated with affixes like verbalizers and so forth.
In a sense, we can say (not sure how valuable this is) that all languages have nouns and adjectives-- things, and words that describe those things. That's on the surface. But whether a word like "blue" is technically two words (blue.ADJ and blue.N) or just one word (blue.N/ADJ) is a major question that depends on how you analyze it.

What appears to be the case is that there are roots and those roots get used in various positions with affixes. This suggests that there is a lot of morphology and possibly no/few distinctions in the lexicon for those roots. But then you get into complicated questions of whether these are grammatical or derivational morphemes and so forth.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: jkpate on March 25, 2014, 01:23:56 AM
I just want to chime in and reinforce djr33's point that "adjective" and "noun" are ultimately going to be theoretical concepts, and so the necessary and sufficient conditions will come down to what your theory of word categories says, and your choice of theory will come down to what you want to do with that theory. You seem to be building your intuitions from two observation: 1) words that typically refer to entities can receive the same suffix as words that typically modify words that typically refer to entities, and 2) words that typically modify words that refer to entities can also themselves refer to entities. If these are the only kinds of generalizations you want your theory to capture, then it might make sense to use only one category for these words.

As an example of another theory, CCG has only two simple syntactic categories, N(oun) and S(entence), and an infinite number of complex categories that encode how words combine with each other. The categories are defined this way to build a close relationship between syntactic structure and semantic structure.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on March 25, 2014, 08:50:59 AM
I need to look into CCG. Is there a good intro you can recommend?
Sorry for dragging this a bit off topic, Zaba!
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 25, 2014, 01:27:43 PM
Quote
In a sense, we can say (not sure how valuable this is) that all languages have nouns and adjectives-- things, and words that describe those things.

Yeah, I may be a novice, but I kinda got that already...

Quote
That's on the surface. But whether a word like "blue" is technically two words (blue.ADJ and blue.N) or just one word (blue.N/ADJ) is a major question that depends on how you analyze it.
Sure, but that decision is obviously based on some criteria, no? That's what I'm asking about.

Quote
What appears to be the case is that there are roots and those roots get used in various positions with affixes. This suggests that there is a lot of morphology and possibly no/few distinctions in the lexicon for those roots.

Sure, but what I keep saying in nearly every post is that there are verbalizers and nominalizers.
Here's my logic in saying that there are only two categories:
- A noun must be verbalised to take verbal suffixes like tense, mood, or aspect.
- A verb must be nominalised to take nominal suffixes like declension
- A noun can never be nominalised
- A verb can never be verbalised
- An adjective must be verbalised to take verbal suffixes like tense mood and aspect.
- An adjective can never be nominalised.
Therefore, adjectives pattern with nouns. The only reason to postulate a category of adjectives is if we are influenced by indoeuro languages.

Quote
But then you get into complicated questions of whether these are grammatical or derivational morphemes and so forth.
OK, let's not go there ;)
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Jase on March 25, 2014, 01:38:45 PM
As an example of another theory, CCG has only two simple syntactic categories, N(oun) and S(entence), and an infinite number of complex categories that encode how words combine with each other. The categories are defined this way to build a close relationship between syntactic structure and semantic structure.

Could you just clarify what CCG is for me? Thanks.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on March 25, 2014, 05:15:07 PM
http://en.wikipedia.org/wiki/Combinatory_categorial_grammar
But jkpate can probably provide us with a better source to learn about it.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on March 25, 2014, 05:19:06 PM
Zaba:

Quote
Yeah, I may be a novice, but I kinda got that already…
Either way. I'm not judging you for it. But it is worth discussing, because you're not being very clear about this. You're asking questions about undefined concepts "noun" and "verb"-- we need a theory to even know what those words mean!

Quote
Sure, but that decision is obviously based on some criteria, no? That's what I'm asking about.
It's based on whatever theory you choose-- based on personal preference, audience, experience, etc.

Quote
Sure, but what I keep saying in nearly every post is that there are verbalizers and nominalizers.
Here's my logic in saying that there are only two categories:
- A noun must be verbalised to take verbal suffixes like tense, mood, or aspect.
- A verb must be nominalised to take nominal suffixes like declension
- A noun can never be nominalised
- A verb can never be verbalised
- An adjective must be verbalised to take verbal suffixes like tense mood and aspect.
- An adjective can never be nominalised.
Therefore, adjectives pattern with nouns. The only reason to postulate a category of adjectives is if we are influenced by indoeuro languages.
Alright. Sounds fine to me. That's exactly what "distributional criteria" means. But you're only looking at morphology there-- also look at syntax-- do adjectives and nouns function the same way in syntax? Are there some positions that only permit one or the other?

As MalFet said above, is it possible to (with the right context) have both of the following?
Blue book [the sort of book that is blue]
Book blue [the sort of blue that is book-like]
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: jkpate on March 25, 2014, 05:47:22 PM
Briefly, Combinatory Categorial Grammar is an alternative to phrase structure grammar. In the simplest form, there are two syntactic categories, N(oun) and S(entence), and these will correspond to the semantic types Entity and Proposition. Complex categories are built by concatenating categories with a directed slash. S\N is a complex category that combines with a noun to the left, since the slash points to the left, and returns a sentence. In more common phrasing, this is (one possible) syntactic category for intransitive verbs. Each syntactic category is associated with a semantic category, in the lambda calculus, whose type is defined in terms of entities and propositions. The type of the semantic category for this proposition would from entitities to propositions, or <e,t> in Montague's notation. Modifiers typically have a syntactic category of the form X/X or X\X: they combine with the thing they modify but do not change the syntax.

There are other rules for combinations as well; for example, S/(S\N), an S looking for (S\N) on its right, can compose with (S\N)/N, (S\N) looking for N on its right, to yield the category S/N. This is useful for analyzing things like gapping. Here is a good short introduction (http://www.inf.ed.ac.uk/teaching/courses/nlg/readings/ccgintro.pdf), and this one also looks good (http://web.uvic.ca/~ling48x/ling484/notes/ccg_intro.pdf). For a much (!) more comprehensive overview, have a look at this paper (http://homepages.inf.ed.ac.uk/steedman/papers/ccg/SteedmanBaldridgeNTSyntax.pdf).

Here are some things to keep in mind when reading about CCG. First, sometimes people include apparent simple categories in their derivations besides S and N. Sometimes they are just abbreviating long complex categories (writing TV for transitive verb instead of the much longer category (S\N)/N), while other times they have really added new simple categories. Second, different applications will consider different combinatory rules. In particular, theoretical papers often include lots of type raising and compositional rules because these expand the expressive power beyond a context free grammar, but papers with computational implementations often omit those rules so the grammar can be implemented and evaluated as a context free grammar.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on March 25, 2014, 08:40:25 PM
to be honest, i'm kind of troubled that no one on this forum can give me necessary and sufficient conditions to determine if there are a category of adjectives in a language. Nothing personal guys, but I didn't think it was a tough question.  You can ask anything about the language and I can provide working translations... think of it as a fieldwork exercise!

Or is the training these days so much more theoretically-orientated?

There is no analysis without theory. That's not a pro-theory sentiment; it's a pro-fieldwork sentiment. Your questions have very different answers depending on the framework you presume. It's not that people are unable to answer. It's that any answer will carry baggage from theoretical commitments that aren't necessarily yours.

Instead, the responses you've been getting have been the ones that should help you to answer the question in your own terms.

Sure, but what I keep saying in nearly every post is that there are verbalizers and nominalizers.
Here's my logic in saying that there are only two categories:
- A noun must be verbalised to take verbal suffixes like tense, mood, or aspect.
- A verb must be nominalised to take nominal suffixes like declension
- A noun can never be nominalised
- A verb can never be verbalised
- An adjective must be verbalised to take verbal suffixes like tense mood and aspect.
- An adjective can never be nominalised.
Therefore, adjectives pattern with nouns. The only reason to postulate a category of adjectives is if we are influenced by indoeuro languages.

Typologically, something akin to a noun/adjective distinction is *extremely* common, so this isn't really an Indo-European bias. If anything, English (with its rather inexplicit morphology) should condition us for the opposite expectation.

The question here is a very simple one: does this language treat entity-like things differently from property-like things grammatically? If talking about "a blue that is book-like" feels strained, that's a good clue that your speakers intuit a difference. Some languages don't.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on March 26, 2014, 01:12:41 PM
Quote
Typologically, something akin to a noun/adjective distinction is *extremely* common, so this isn't really an Indo-European bias.
Just to add to that: it's somewhat common for languages to lack a distinction between verb and adjective. This appears to be the case (in some ways at least) for Japanese, and more clearly for other languages. So finding a N+A vs. V distinction might be unexpected for that reason as well.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on March 27, 2014, 12:19:13 AM
I guess I naively assumed that terms like "noun" and "adjective" would have a consensual definition within linguistics. After all, phoneme and phone (see my previous discussions on this topic) were definable across language without resort to theoretical analyses (though surely there is much debate on e.g. distinctive features, feature geometry and the like).  The point is, we have a clear method to determine if a sound is a phoneme in a language (just like for minimal pairs and test) but not one to determine what a category of a word is. That surprises me.

Anyway, as for book blue and blue book story, perhaps it's interesting to note that a modifier like (in English) "very/much" can mean "good" when the head of a NP (though there are other words for "good" too). This sort of evidence further strengthens my claim.

1. very/much + big = 'very big' & 'many big.ones'
2. want.1sg + very/much-ACC = 'I want many'
3. very/much-verbalizer-1sg = 'I am good/fine'

Anyway, thanks for your help, all y'all.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on March 27, 2014, 02:39:30 AM
I guess I naively assumed that terms like "noun" and "adjective" would have a consensual definition within linguistics. After all, phoneme and phone (see my previous discussions on this topic) were definable across language without resort to theoretical analyses (though surely there is much debate on e.g. distinctive features, feature geometry and the like).  The point is, we have a clear method to determine if a sound is a phoneme in a language (just like for minimal pairs and test) but not one to determine what a category of a word is. That surprises me.

You're making a category error here, though. The morphosyntactic equivalent of phoneme isn't "noun" or "verb" but rather "lexical class". A particular lexical class (like "noun") would be more analogous to a particular phoneme (like /p/ or /b/). Particular phonemes are not definable across languages, at least not pre-theoretically.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 02, 2014, 09:33:44 PM
Quote
The morphosyntactic equivalent of phoneme isn't "noun" or "verb" but rather "lexical class". A particular lexical class (like "noun") would be more analogous to a particular phoneme (like /p/ or /b/). Particular phonemes are not definable across languages, at least not pre-theoretically.

Thanks for that distinction. It seems reasonable. But the point remains. If I want to know if a language has /p/ /b/ distinction, there's a good way to check (mininmal pairs, e.g.) -- but how can I check if a language has a NOUN ADJECTIVE distinction?(according to some consensusal definition of nouns and adjs; akin to the consensual definition of /p/ and /b/)
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on April 02, 2014, 10:36:53 PM
Quote
(according to some consensusal definition of nouns and adjs; akin to the consensual definition of /p/ and /b/)
That's inaccurate. P and B are given as labels for distinctive sounds because they seem generally like the sound in English (etc.) with those labels. They are not defined beyond that. There are two sounds (phonemes), and we know that through minimal pairs. Then we give them names out of convenience, but not because they "are" those labels. As an example of this, just consider the use of "R" across various languages, or the /a/ phoneme as well-- lots of variation. To say there "are nouns in languages" or there "are Bs in languages" is imprecise.

Quote
Thanks for that distinction. It seems reasonable. But the point remains. If I want to know if a language has /p/ /b/ distinction, there's a good way to check (mininmal pairs, e.g.) -- but how can I check if a language has a NOUN ADJECTIVE distinction?
You've already stated the answer: you need to search for a minimal pair. If you cannot find a minimal pair, then they are not contrastive.

After this, you can arbitrarily apply labels "noun" and/or "adjective" as needed based on criteria like whether the class typically refers to properties or things, but that's just out of convenience.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: jkpate on April 02, 2014, 10:41:04 PM
Thanks for that distinction. It seems reasonable. But the point remains. If I want to know if a language has /p/ /b/ distinction, there's a good way to check (mininmal pairs, e.g.) -- but how can I check if a language has a NOUN ADJECTIVE distinction?(according to some consensusal definition of nouns and adjs; akin to the consensual definition of /p/ and /b/)

Again, we're getting into a question of definitions. When you say "consensual definition" what you really mean is "theoretic definition" of /p/ and /b/. How do you know what /p/ and /b/ are if they are not shorthands for stuff that you saw in some language data? If /p/ and /b/ are universal feature specifications, then you are assuming a theory of segments that ignores gradation. If /p/ and /b/ are shorthands for articulatory sequences, then you are assuming a motor theory of segments. If, on the other hand, you have been writing "p" and "b" as a convenient shorthand for yourself, then the use of two symbols is a fact about you and your perception, not the language; there's no pre-theoretic reason to draw a correspondence between sounds that you have labeled "p" when listening to one language and sounds that you have labeled "p" when listening to another language.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 03, 2014, 04:57:24 AM
Quote
You've already stated the answer: you need to search for a minimal pair. If you cannot find a minimal pair, then they are not contrastive.

So what is the minimal pair in e.g English that shows adjectives exist as a separate category from nouns?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 03, 2014, 04:58:34 AM
Quote
When you say "consensual definition" what you really mean is "theoretic definition" of
Well, I'm not sure about that. We agree that the prototypical /p/ is not voiced, for example.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on April 03, 2014, 05:19:27 AM
Quote
When you say "consensual definition" what you really mean is "theoretic definition" of
Well, I'm not sure about that. We agree that the prototypical /p/ is not voiced, for example.

That's a handy simplification, but it's rarely actually true if you get down and dirty in the phonetics. There's a reason, for example, that English doesn't have a word "sbin".
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 03, 2014, 05:38:28 AM
Quote
That's a handy simplification, but it's rarely actually true if you get down and dirty in the phonetics. There's a reason, for example, that English doesn't have a word "sbin".
OK, we're getting of track here -- simplification or not, the fact is phoneticians agree that the prototype of the phoneme /p/ is not voiced. It does not have the distinctive feature for voice.

Anyway, while we're on the topic for distinctive features, what are the "distinctive features" for adjectives in e.g. English that make them distinct from nouns in the same language?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on April 03, 2014, 07:12:42 AM
Quote
That's a handy simplification, but it's rarely actually true if you get down and dirty in the phonetics. There's a reason, for example, that English doesn't have a word "sbin".
OK, we're getting of track here -- simplification or not, the fact is phoneticians agree that the prototype of the phoneme /p/ is not voiced. It does not have the distinctive feature for voice.

Anyway, while we're on the topic for distinctive features, what are the "distinctive features" for adjectives in e.g. English that make them distinct from nouns in the same language?

It's not off track, though. That's the very important thing. You keep talking about "the prototype of the phoneme /p/", but no such thing exists without a great big pile of theory. Seriously. Phoneticians do not agree that the phoneme /p/ is not voiced, because that's not how phonemes work. Phonemes -- fundamentally -- do not possess *any* extent of cross-linguistic identity. There is no prototype from which individual languages are deviations. The only thing that exists is contrast, and we ascribe these contrasts with conventional labels to make it easier to talk to each other. The fact that we call all these different things /p/ is a descriptive convenience.

The same is true for functional categories like adjective and noun. What makes an adjective an adjective is that it's not a noun. What makes a noun a noun is that it's not an adjective. If you want to postulate (as many do) a more universal schema, you're doing so in terms of a particular theoretical frame.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on April 03, 2014, 08:15:46 AM
The prototype for noun is "thing" and the prototype for adjective is "property", but that won't get you far in the analysis of course.

There is no universal test, in the same way that the "bat vs pat" test is valid only within English. Based on the grammatical rules you have, are there constructions where nouns and adjectives are distinct?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 03, 2014, 11:30:28 AM
Quote
The only thing that exists is contrast, and we ascribe these contrasts with conventional labels to make it easier to talk to each other. The fact that we call all these different things /p/ is a descriptive convenience.

I was under the impression that all phonemes could be described with reference to distinctive features, depending on what language they were a part of.

Anyway...

Quote
There is no universal test, in the same way that the "bat vs pat" test is valid only within English.
I disagree. Minimal pairs are a kind of test... no?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on April 03, 2014, 11:51:23 AM
Quote
I was under the impression that all phonemes could be described with reference to distinctive features, depending on what language they were a part of.
In a specific theory that defines particular distinctive features, yes.
Without that theory, certainly not.

In that regard, features for phonemes are much more researched than whatever the equivalent would be for word classes, but either way both topics are up for debate.

Quote
I disagree. Minimal pairs are a kind of test... no?
In the same sense that "multiple choice" might be a good kind of test for a math class.
What I meant is that there is no ready-to-go-just-fill-in-some-sentences test for nouns vs. adjectives. It depends on the language in question.

In short: the answer is checking for contrastiveness (using minimal pairs).
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 03, 2014, 11:53:09 AM
 :o
Yes!
What are some good minimal pairs to start checking with?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on April 03, 2014, 12:12:15 PM
Quote
What are some good minimal pairs to start checking with?
That's exactly what I said isn't universal! You can't check "bat" and "pat", nor can you check based on English nouns/adjectives.

But you should be able to fairly easily think of a handful of uses of each, and then see whether they're interchangable. We've said that already-- check MalFet's earlier post for at least one example.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 03, 2014, 12:49:19 PM
ok, let's stick with English nouns vs adjectives. On what basis would you claim they are separate categories? Adjectives, in Eng, after all, are rather noun-like.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on April 03, 2014, 02:58:33 PM
very happy
*very book

This dog is big.
*This big is dog.

dog > dogs
big > *bigs

dog > *dogger
big > bigger

etc.

These exact tests will not apply to other languages, but some may and the general idea will apply.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on April 03, 2014, 07:48:45 PM
Quote
The only thing that exists is contrast, and we ascribe these contrasts with conventional labels to make it easier to talk to each other. The fact that we call all these different things /p/ is a descriptive convenience.

I was under the impression that all phonemes could be described with reference to distinctive features, depending on what language they were a part of.

Anyway...

Everything I'm saying about phonemes applies to features as well. If your theoretical model presumes the existence of features, then that's actually your basis of contrast. Phonemes then become composites and nothing more.

Quote
There is no universal test, in the same way that the "bat vs pat" test is valid only within English.
I disagree. Minimal pairs are a kind of test... no?

Minimal pairs are a demonstration of contrast. That's it. They do not, at face value, tell you anything about the nature of the contrast. There is nothing inherent to the universe that mandates certain kinds of contrasts to exist in language. These contrasts are language specific, and it is methodologically dangerous to go in assuming we should be able to find minimal pairs of a particular type in a given language.

In other words, you just need to look for distributional criteria: syntactic and morphological contexts in which one class of word can appear but another can't. That's it. We can't tell you ahead of time what those contexts will be. Daniel has illustrated some of those contexts for English, but as he says there is no reason those same distributional criteria will work in the languages you are working on.

Typologically speaking, most languages make a distinction in basic predication. If you really can say "This big is dog" as easily as "This dog is big", it is entirely possible that the language does not distribute nouns and adjectives contrastively. More tests are needed, but that's a good start.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 03, 2014, 11:10:29 PM
Thanks, guys.

Quote
very happy
*very book
Both would be ok (though 'very book' would mean 'many books', although it's the same 'very' as in 'very happy')

Quote
This dog is big.
*This big is dog.
hard to say because the copula is different... but sorta possible to have both orders

Quote
dog > dogs
big > *bigs
'bigs' is ok. (meaning in Eng 'the big ones')

Quote
dog > *dogger
big > bigger
augmentive doesn't exist in this language.

Quote
Typologically speaking, most languages make a distinction in basic predication. If you really can say "This big is dog" as easily as "This dog is big", it is entirely possible that the language does not distribute nouns and adjectives contrastively. More tests are needed, but that's a good start.
I really appreciate this test -- but it's a tough one because the copula doesn't exist in this language, at least in a comparable way. Can you or anyone think of a similarly simple "first pass" test?
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: Daniel on April 03, 2014, 11:48:09 PM
Quote
hard to say because the copula is different... but sorta possible to have both orders
It's not a question of whether it can be said/translated but whether it is the same construction for both. If not, that's clear evidence that they behave differently.



Also:
The big dog
*The dog big
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: MalFet on April 04, 2014, 02:38:56 AM
I really appreciate this test -- but it's a tough one because the copula doesn't exist in this language, at least in a comparable way. Can you or anyone think of a similarly simple "first pass" test?

As I keep trying to say, I can't, no, because I don't know anything about this language and there's no test that can be reliably applied cross linguistically.

Typologically speaking, it would be very unusual (though not impossible) for a language to treat subjects and predicates identically in all instances. That said, I can't guess how or if this language happens to treat them differently. If the language you're looking at does indeed decline to make a subject/predicate distinction, you're probably looking at a language in which word class doesn't play a particularly important part. If it does make a subject/predicate distinction in some cases, the question then becomes whether all non-verb words can appear in all positions within these structures.
Title: Re: What are nec and sufficient conditions for you to agree that a lang lacked adjs?
Post by: zaba on April 04, 2014, 09:00:27 AM
Thanks everyone!