Typological correlates of verby and nouny adjectives

[chris_notts]

I've posted on this subject before, but I was reading a book by Dixon and I was reminded of potentially contradictory claims about the behaviour of adjectives between him and Stassen. I should say that by "adjectives" I mean the set morphemes expressing key properties such as size, colour, etc. Both authors discuss typological implications involving whether adjectives behave more like nouns or verbs, from a morphological and syntactic point of view. This involves factors such as:

1. Do predicate adjectives take verbal morphology, or pattern with nouns (e.g. in taking a copula)?
2. Do attributive adjectives occur in relative clauses like verbs? Do they agree in nominal categories such as case?

Dixon makes the claim that adjectives in dependent marking languages, where roles in the clause are marked by cases, tend to be more noun-y, whereas adjectives in languages which are not dependent marking tend to be verb-y. He does, however, say that "It should be emphasized that this is very much a first run-through of the data." So if Dixon is right, then roughly:

dependent-marking <-> nouny adjectives

although of course he is aware of exceptions to the proposed relationship. You can read about his views in at least the following books:

Adjective Classes: A Cross-Linguistic Typology (edited by R. M. W. Dixon and Alexandra Y. Aikhenvald)
Basic Linguistic Theory Volume 2: Grammatical Topics (by R. M. W. Dixon)

Stassen, on the other hand, wrote a big book called "Intransitive Predication" which established a typology of predication for locative, nominal, adjectival and verbal predication. He is only concerned with predication, so he does not look at the the behaviour of adjectives in attributive function, only their behaviour as predicates. He claims on the basis of a large sample (+ discussion of each language family / area in detail) that:

verbal tense marking <-> nouny / non-verby adjectives

That is, adjectives behave similarly to verbs in predicates if and only if verbs do not take obligatory tense marking. He defines tense marking for the purpose of his study to be a verbal morphologically bound distinction between past and non-past, and does not treat languages with a future vs non-future distinction as exceptions. He says that this is because tense is less relevant to adjectives, since they tend to describe more time-stable states of affairs than the most stereotypical verbs do.

Incidentally, Stassen says that a claim almost opposite to Dixon's was made in 1951 by Ernst Locker. Locker claimed that a language had nouny adjectives if and only if it had PNG (person/number/gender) marking on verbs, whereas Dixon suggests that head-marking languages are more likely to have verby (non-nouny) adjectives. Stassen finds no evidence for the two-way implication, but does find a (far from universal) tendency in his sample for languages without PNG marking on verbs to have verby adjectives.

It seems to me that this could be explained by his tense claim, since isolating languages tend not to have bound PNG marking. That is, if knowing that a language lacks PNG marking raises the probability of it also lacking tense (or aspect, or mood, or ...) marking, i.e. of being isolating, then you would expect a weak tendency for languages without PNG marking to have verb-y adjectives. However, it obviously couldn't be explained by Dixon's proposal, since it directly contradicts it.

Anyway, we have two proposals. Keeping in mind that both have known exceptions, so are at best statistical universals (= strong tendencies), we have:

dependent-marking <-> nouny adjectives
verbal tense marking <-> nouny / non-verby adjectives

If both are true, then we would have as a consequence that dependent-marking <-> verbal tense marking, which would be interesting. However, I doubt that this is the case because I can think of a lot of counter-examples, including almost the entire Bantu language family, which tend to have both multiple past and future tenses and be predominantly head-marking. If Bantu languages we so typologically unusual, you would expect at least some of them to have drifted back towards typological normality.

I was actually relatively convinced by Stassen's book, since he goes to great lengths to go through his sample and discuss each language / family in his sample, so I'd be more inclined to reject or weaken Dixon's proposed typological implications. The only major issue Stassen has is his tendency to try to reinterpret any ambiguity (which always exists in typology given the fact that languages are messy and sources are of varying quality) in favour of his own theory and try to explain away borderline counter-examples in less than convincing ways. But I don't think it was that big an issue in "Intransitive Predication" - I had far more objections to some of his handwaving in "Predicative Possession", another huge work on typology of his.

[chris_notts]

I've not actually read this book, but just from looking at the contents it looks like the author has arrived at the conclusion that adjectival predication has something to do with tense marking:

http://www.amazon.co.uk/Typology-Adject ... 476&sr=8-5
The Typology of Adjectival Prediction by Harrie Wetzer

It would be interesting to compare it with Stassen's work.

EDIT: Looks like Stassen was one of the supervisors of his dissertation, so I would assume that the conclusions are similar.

[merijn]

I've not actually read this book, but just from looking at the contents it looks like the author has arrived at the conclusion that adjectival predication has something to do with tense marking:

http://www.amazon.co.uk/Typology-Adject ... 476&sr=8-5
The Typology of Adjectival Prediction by Harrie Wetzer

It would be interesting to compare it with Stassen's work.

EDIT: Looks like Stassen was one of the supervisors of his dissertation, so I would assume that the conclusions are similar.

I read it a couple of years ago. And iirc he does indeed say that being nouny or verby is dependent on tense marking. I have to say that I didn't really bought his reasoning for why this is the case, and I also felt that some of his criteria for deciding whether a adjective is nouny or verby were a bit arbitrary, but it is a very extensive study.
I also want to add that I feel that people are missing something when they are comparing the categories of words instead of the categories of phrases, because one class of adjectives in Zulu (which form AP's) behaves in many ways exactly like PP's in its conjugation.

[chris_notts]

I also want to add that I feel that people are missing something when they are comparing the categories of words instead of the categories of phrases, because one class of adjectives in Zulu (which form AP's) behaves in many ways exactly like PP's in its conjugation.

Do you have any examples?

[merijn]

I also want to add that I feel that people are missing something when they are comparing the categories of words instead of the categories of phrases, because one class of adjectives in Zulu (which form AP's) behaves in many ways exactly like PP's in its conjugation.

Do you have any examples?

There is basically a two split way in predication between verbs and non-verbs in Zulu in tenses other than the present tense, and that is that all non-verbs need an auxiliary, but verbs don't need an auxiliary. Within the non-verbs there is a second split in the 3rd person indicative present tense where one class of adjectives and nominal predication don't need subject agreement, but another class of adjectives and PP's do have subject agreement, so you say on the one hand "ngumfundisi" "he is a teacher" "muhle", he is beautiful", but on the other hand "useLondon" "he is in London" and "umatasa" "he is busy. On top of that, one class of PP's, the so-called locatives, must be doubled in the negative with -kho there, so you say "angikho eLondon" lit: I am not there in London". In the class that behaves like PPs in the present tense I have seen this construction as well, although the non-doubled construction is also possible, so I am not busy, can be both "angimatasa" and "angikho matasa". Note that this is my impression, and not checked with native speakers.

[marconatrix]

Many languages (Russian, Arabic ...) suppress the copula, at least when tense is left unmarked. Can one really draw a line between this situation and languages like Nahuatl where nouns/adjectives are described as having implicit verbal qualities. E.g. the words 'man', 'big' with appropriate affixes can mean 'he is a man', 'it is big' etc. That is, once an explicit copula drops out of use, nouns/adjectives are reinterpreted as having verbal force? So this is a special case of grammaticalization maybe, of a null marker?? Is that possible? In Iroquoian langs IIRC 'nouns' include a formative which may once have been an auxillary verb, but is now 'semantically bleached', which would mean that syntactically everything is technically a verb.

The copula seems to be a very unstable 'isotope'. When it doesn't vanish as above there seems to be a strong tendency in many languages for it to be combined with or replaced by locative or existential ('there is ...') constructions.

In other words, would it be helpful to categorise langs on the basis of the nature of the copula?

[chris_notts]

Many languages (Russian, Arabic ...) suppress the copula, at least when tense is left unmarked. Can one really draw a line between this situation and languages like Nahuatl where nouns/adjectives are described as having implicit verbal qualities. E.g. the words 'man', 'big' with appropriate affixes can mean 'he is a man', 'it is big' etc.

I think Stassen does draw a line between use of a zero copula and verbal use of a root, for a certain definition of "verbal", in his study of intransitive predication. I can't remember off the top of my head what it is though. I guess it's mainly hard to differentiate if the verb is morphologically inert and does not have agreement / TAM marking / ..., since obviously with a zero copula such morphology is missing.

[Radagast revived]

Many languages (Russian, Arabic ...) suppress the copula, at least when tense is left unmarked. Can one really draw a line between this situation and languages like Nahuatl where nouns/adjectives are described as having implicit verbal qualities. E.g. the words 'man', 'big' with appropriate affixes can mean 'he is a man', 'it is big' etc. That is, once an explicit copula drops out of use, nouns/adjectives are reinterpreted as having verbal force? So this is a special case of grammaticalization maybe, of a null marker?? Is that possible? In Iroquoian langs IIRC 'nouns' include a formative which may once have been an auxillary verb, but is now 'semantically bleached', which would mean that syntactically everything is technically a verb.

The copula seems to be a very unstable 'isotope'. When it doesn't vanish as above there seems to be a strong tendency in many languages for it to be combined with or replaced by locative or existential ('there is ...') constructions.

In other words, would it be helpful to categorise langs on the basis of the nature of the copula?

I sense another chance to plug some work of mine: Adjectives in Hueyapan Nahuatl: Do they exist? And if they do what kind of adjectives are they?.

I argue that the only basic wordclass split in Nahuatl is between words that can form predicates and words that can't - this basically reanalyzes "adjectives" as stative verbs. I believe this analysis also makes sense in some (most?) other languages with "verby"-adjectives - at least the ones with obligatory headmarking of grammatical relations like Nahuatl and Mohawk.

[Astraios]

Many languages (Russian, Arabic ...) suppress the copula, at least when tense is left unmarked. Can one really draw a line between this situation and languages like Nahuatl where nouns/adjectives are described as having implicit verbal qualities. E.g. the words 'man', 'big' with appropriate affixes can mean 'he is a man', 'it is big' etc.

Yes, you can. Lakota doesn't suppress copulas, and only VPs can contain the irrealis enclitic ktA, yet a 'noun' can appear in a 'verb' place:

Hé wičháša kte.
he [be.a.]man=IRR
He will be a man.

Hebrew does suppress copulas, and only verbs can be conjugated for tense, yet a noun can't appear in a verb place:

Hu yihye gever.
he be.FUT man
He will be a man.

It just doesn't make sense to say Lakota and Hebrew can both use nouns as verbs, or that Hebrew and Lakota are both merely suppressing their copulas, even though it's true that in sentences unmarked for TAM they appear identical:

Hé wičháša. ~ Hu gever.
he [be.a.]man ~ he [be] man
He is a man.

[merijn]

Many languages (Russian, Arabic ...) suppress the copula, at least when tense is left unmarked. Can one really draw a line between this situation and languages like Nahuatl where nouns/adjectives are described as having implicit verbal qualities. E.g. the words 'man', 'big' with appropriate affixes can mean 'he is a man', 'it is big' etc. That is, once an explicit copula drops out of use, nouns/adjectives are reinterpreted as having verbal force? So this is a special case of grammaticalization maybe, of a null marker?? Is that possible? In Iroquoian langs IIRC 'nouns' include a formative which may once have been an auxillary verb, but is now 'semantically bleached', which would mean that syntactically everything is technically a verb.

The copula seems to be a very unstable 'isotope'. When it doesn't vanish as above there seems to be a strong tendency in many languages for it to be combined with or replaced by locative or existential ('there is ...') constructions.

In other words, would it be helpful to categorise langs on the basis of the nature of the copula?

I sense another chance to plug some work of mine: Adjectives in Hueyapan Nahuatl: Do they exist? And if they do what kind of adjectives are they?.

I argue that the only basic wordclass split in Nahuatl is between words that can form predicates and words that can't - this basically reanalyzes "adjectives" as stative verbs. I believe this analysis also makes sense in some (most?) other languages with "verby"-adjectives - at least the ones with obligatory headmarking of grammatical relations like Nahuatl and Mohawk.

If I were to write a book about lexical categories or adjectives I would say that one important feature that adjectives have, sometimes opposing other lexical categories in the language, is that the meaning of the adjective is almost universally predictable from the core meaning "having property X" in almost any language in the same way. When used attributively it always means a noun that has property X, when used predicatively it means the subject has property X and when used as an argument (which is not possible in many languages) it refers to a someone or something with property X. So to use examples from English, Dutch and Zulu, when the word for green modifies a noun "a green man" "een groene man" "indoda eluhlaza" it always means that the noun (man/man/indoda) has the property of being green, when it is used predicatively "the man is green" "de man is groen" "indoda iluhlaza", again means that the subject has the property of being green, and when used as an argument (which is restricted in English, unless you count the constructions with "one") "ik zie een groene" "ngibona eluhlaza" it refers to something or someone that is green.

How would your "adjectives" behave in this respect, and do they differ from "nouns" and "verb"? Going by examples 1-4, and you know the language much, much better than I do, so you can also tell to what extent these examples are representative, it seems that there is a split between nouns and verbs, in how the predicative meaning and the argument reading relate to each other. With nouns the meaning of the predicative use is "subject has the property of being X" iow "subject is X", and the meaning of the argument use is "the argument has the property of being X iow referent is an X". With verbs if the meaning of the predicative use is "subject has the property of being the agent of X", the argument use is different compared to nouns, it is not "the argument has the property of being the agent of X", but in (6) the argument of "like" is the whole clause "that I eat". It seems from the rest of your paper that "adjectives" behave like nouns in this respect, (as in any language I know).
WRT attributive use, you give an example of an ordinary noun used attributively "a pig woman" meaning "a dirty woman, example (11). This example is not conclusive to me to see what the meaning is of nouns used attributively. Could it also mean "the woman who has a lot of pigs"? If someone is wearing a wig, can you say "the wig woman" without meaning "the woman who is a wig"? In that case it seems to me that nouns and adjectives behave differently this way.

Anyway, I hope you don't mind these questions. It was a very interesting read, and I want to thank you very much for letting us read it. It is just something that always irks me whenever I read the claim in an article that language X doesn't have lexical category Y.

[Whimemsz]

I assume Stassen includes aspect along with tense in his calculations? In any case the Algonquian languages seem like an obvious counter-example to his trend: they have a large assortment of aspectual and tense preverbs, but "adjectives" are only distinguishable from other verbs on semantic grounds (e.g., Ojibwe gii-ozaawaa, "it was brown/yellow/orange", wayiiba da-ginoozi, "soon s/he'll be tall") (although no Algonquian language that I know of can go as far as languages like Lakota and Nahuatl and treat noun stems as attributive verbs; in such cases you either use a copula, use a copular derivational suffix on the noun, or use a demonstrative or predicative adverbial particle). A number of Algonquian languages also have some form of "preterit" inflection (as with Ojibwe -ban, e.g., ozaawaaban "it had been/was brown/yellow/orange").

Interestingly, though, Ojibwe has also borrowed the preterit suffix to use in noun inflection as well, where it marks deceased people, former possession, or destroyed/ruined objects, e.g. nookomisiban "my late grandmother", gijiimaaninaanibaniin "our (incl.) former canoes", iw waakaa'iganiban "the former (now-destroyed) house".

[Radagast revived]

If I were to write a book about lexical categories or adjectives I would say that one important feature that adjectives have, sometimes opposing other lexical categories in the language, is that the meaning of the adjective is almost universally predictable from the core meaning "having property X" in almost any language in the same way. When used attributively it always means a noun that has property X, when used predicatively it means the subject has property X and when used as an argument (which is not possible in many languages) it refers to a someone or something with property X. So to use examples from English, Dutch and Zulu, when the word for green modifies a noun "a green man" "een groene man" "indoda eluhlaza" it always means that the noun (man/man/indoda) has the property of being green, when it is used predicatively "the man is green" "de man is groen" "indoda iluhlaza", again means that the subject has the property of being green, and when used as an argument (which is restricted in English, unless you count the constructions with "one") "ik zie een groene" "ngibona eluhlaza" it refers to something or someone that is green.

How would your "adjectives" behave in this respect, and do they differ from "nouns" and "verb"? Going by examples 1-4, and you know the language much, much better than I do, so you can also tell to what extent these examples are representative, it seems that there is a split between nouns and verbs, in how the predicative meaning and the argument reading relate to each other. With nouns the meaning of the predicative use is "subject has the property of being X" iow "subject is X", and the meaning of the argument use is "the argument has the property of being X iow referent is an X". With verbs if the meaning of the predicative use is "subject has the property of being the agent of X", the argument use is different compared to nouns, it is not "the argument has the property of being the agent of X", but in (6) the argument of "like" is the whole clause "that I eat". It seems from the rest of your paper that "adjectives" behave like nouns in this respect, (as in any language I know).
WRT attributive use, you give an example of an ordinary noun used attributively "a pig woman" meaning "a dirty woman, example (11). This example is not conclusive to me to see what the meaning is of nouns used attributively. Could it also mean "the woman who has a lot of pigs"? If someone is wearing a wig, can you say "the wig woman" without meaning "the woman who is a wig"? In that case it seems to me that nouns and adjectives behave differently this way.

Anyway, I hope you don't mind these questions. It was a very interesting read, and I want to thank you very much for letting us read it. It is just something that always irks me whenever I read the claim in an article that language X doesn't have lexical category Y.

It seems that you are trying to define adjective as a semantic category rather than as a syntactic one. That is basically Dixon's definition - "any discrete lexical class that contains property concept terms in a given language". For most people this definition isn't attractive because it basically doesn't justify separating adjectives out as a word class which is normally defined as a class with distinct syntactic properties - not just semantic ones or morphological ones.

In Hueyapan Nahuatl the issue is that the word class that property words do not behave in the same way and that none of them seem to have syntactic properties setting them apart from other predicate forming words. They can be used attributively, but this is a very rare usage and it is not exclusive to property words. The only criteria by which it seems some property words can be separated syntactically from other word classes is that they seem to occur more frequently in position before the noun they modify (but this may be because they are "light").

As for the example of the "dirty/pig woman" I am not going to venture a guess about other possible meanings - that would have to rely on a thorough semantic study, this is a single occurrence in a text with a meaning that is clear in the context. I don't think that "wig" woman would work - exactly because "pig" is read metaphorically as a property word i.e. "having the property of piggishness" - I don't think that wig would do that.

Interestingly it seems that in Nahuatl the word for good was derived in this metaphorical way because it seems to be etymologically derived as being a noun meaning "edible thing". This is actually a really interesting process that I hadn't thought of that the creation of adjectives can be a metaphoric process - extracting the property from a word and grammaticalizing it.

[merijn]

It seems that you are trying to define adjective as a semantic category rather than as a syntactic one. That is basically Dixon's definition - "any discrete lexical class that contains property concept terms in a given language". For most people this definition isn't attractive because it basically doesn't justify separating adjectives out as a word class which is normally defined as a class with distinct syntactic properties - not just semantic ones or morphological ones.

In Hueyapan Nahuatl the issue is that the word class that property words do not behave in the same way and that none of them seem to have syntactic properties setting them apart from other predicate forming words. They can be used attributively, but this is a very rare usage and it is not exclusive to property words. The only criteria by which it seems some property words can be separated syntactically from other word classes is that they seem to occur more frequently in position before the noun they modify (but this may be because they are "light").

As for the example of the "dirty/pig woman" I am not going to venture a guess about other possible meanings - that would have to rely on a thorough semantic study, this is a single occurrence in a text with a meaning that is clear in the context. I don't think that "wig" woman would work - exactly because "pig" is read metaphorically as a property word i.e. "having the property of piggishness" - I don't think that wig would do that.

Interestingly it seems that in Nahuatl the word for good was derived in this metaphorical way because it seems to be etymologically derived as being a noun meaning "edible thing". This is actually a really interesting process that I hadn't thought of that the creation of adjectives can be a metaphoric process - extracting the property from a word and grammaticalizing it.

I'd say my criterion is about argument structure, and from my perspective argument structure is as much syntax as it is semantics. It is certainly not lexical semantics, as Dixon's definition is. The bottom line is that in most languages, in the different uses of words (as an argument, as a predicate or as a modifier) you have to know whether a word is a noun, adjective or verb to know what is the argument of which predicate. It seems to me based on your examples that there is a split between nouns and verbs in this respect. Based on your hunch expressed in the post above, there doesn't seem to be a difference between nouns and adjectives in this respect.
Personally I would definitively argue for a difference between nouns and verbs in Hueyapan Nahuatl, not just based on the criterion above, but mostly based on the fact that you need a copula in the non-present tense for nouns. I worked on a project on non-verbal predication through different languages and the presence of a copula, even if it was not present in the present tense, was probably the most reliable indicator that a word wasn't a verb. Also the distribution of the case marker is from my perspective a big indicator of different lexical categories, but here this is probably influenced by me coming from a generative background. In classical generative grammar, and the basic idea has survived into modern day minimalism, even in a mutated form, an NP must be marked for case either overtly or covertly. I would argue that case-marking is overt in Hueyapan Nahuatl, and the lack of case-marking on adjectives and verbs means that they are not heading NP's and that they are not of the same category as nouns. Again, this is coming from someone with a different background than you, and working in a different framework, so the question, "what does it mean to be a noun and to be a verb?" has a slightly different meaning.

One comment and one question I forgot in my last post. 1) I recall that whereas nouns are used overwhelmingly as arguments, and verbs overwhelmingly as predicates, adjectives are not used overwhelmingly as noun modifiers in most languages that have them, but rather as much as they are used as predicates. I don't know where I read that and I may have remembered it incorrectly.
2) If you have time, would you mind giving us a quick outline how PP's behave in your language, and if they differ from adjectives?

[Radagast revived]

I must admit I don't fully understand your point about a difference in argument structure in nouns and adjectives in Hueyapan Nahuatl. I don't see the possibility of making any such distinction in HN - which HN words would you classify as adjectives according to those criteria?

About the Noun verb distinction I agree completely: There is definitely a syntactic and morphological difference between verbs and nouns in Hueyapan Nahuatl. But the difference is not in their ability to form predicates - just in the way they do it. And yes I agree that for those who would argue that there is no difference between the two the fact that the past requires copula is basic evidence against their position that they would somehow have to deal with. (they do)

I don't know what you mean when you say that HN nouns could be interpreted as being marked overtly for case - the absolutive case only marks that something is an unpossessed noun. Some "adjectives" (e.g. "good") are marked with the absolutive suffix - and many nouns aren't.

I admit that it is likely my lack of familiarity with generative grammar that makes me not get this right away. But I am actually really interested in understanding that point of view which is also why I've tried to engage with Baker even though it is a little alien to me (and sometimes I feel it violates the data). If you have time to explain I'd be happy to listen.

[Radagast revived]

1) I recall that whereas nouns are used overwhelmingly as arguments, and verbs overwhelmingly as predicates, adjectives are not used overwhelmingly as noun modifiers in most languages that have them, but rather as much as they are used as predicates. I don't know where I read that and I may have remembered it incorrectly.
2) If you have time, would you mind giving us a quick outline how PP's behave in your language, and if they differ from adjectives?

You are right that modifying nouns isn't necessarily the most basic function for adjectives even in the languages where they do so frequently. I also don't take this as a necessary evidence that they aren't adjectives - but if they in fact do not modify nouns at all which is a distinct possibility depending on whether they are better interpreted as relative clauses then I think I also don't think it makes sense to consider them a separate adjective class.

2. PPs are made either with possessed relational nouns or with locative suffixes on nouns. So basically they are NPS that function as modifiers of a phrasal head and are marked as not being its direct arguments.

[merijn]

I am sorry that I only answer now, but life happened (or rather I had to do something that was more pressing than writing an answer on a messageboard, and I kept postponing it, and I promised myself to do that before answering posts on a message board).

I must admit I don't fully understand your point about a difference in argument structure in nouns and adjectives in Hueyapan Nahuatl. I don't see the possibility of making any such distinction in HN - which HN words would you classify as adjectives according to those criteria?

Yes, based on your description, I don't think there is a distinction between adjectives and nouns, but there is a distinction between nouns and verbs in this respect. I will explain what I mean using more formal semantics, if that is clearer to you,although I am not good in the formalizations of formal semantics. What I think is almost universally true about adjectives is that if we represent the meaning of the adjective with p the meaning of the adjective
1) used predicatively is p(subject) for instance "John is green" green(John)
2) when used attributively it is &p(noun) for instance "I see a green man" see(I, man)&green(man)
3) when used as an argument (which not all languages allow and some language (such as English) have restrictions on it), it is best described as x&p(x) where x is a variable (here I am less sure if I used the correct formalizations for what I am trying to say but what the hell) so Dutch "ik zie een groene" is see (I, x)&green(x) where x is a variable.
With nouns
1) when used predicatively it is the same as an adjective p(subject) "John is a teacher" teacher (John)
2) when used as an argument again the same as an adjective "I see a teacher" see(I,x)&teacher(x)
3) when used attributively however, if possible languages vary what the meaning is. In English, for instance, when names of cities are used attributively it means "coming from" so "I see a Kansas City girl" means see(I, girl)&come.from (girl,x)&Kansas City(x), in some other languages when used attributively it means "made of ", or "having a connection with", and in some languages it may be exactly the same as adjectives "being X". It seems to me that HN is like the last group and doesn't make a distinction between adjectives and nouns.
With verbs (I will be using intransitive verbs but the same is true for transitive verbs)
1) when used predicatively, again the same as adjectives p(subject) "John eats" eat(John)
2) when used as an argument, usually it is the whole proposition (which is p(subject)) that serves as an argument "John wants to eat" want (John, (eat(John))) (this sentence has the same meaning as your example sentence for the use of a verb as an argument), or "I see John eat" see (I, (eat(John), or "eating is easy" easy ( (eat (x))) where x refers to let's say the general public. This is clearly different from nouns and adjectives
3) when used attributively, for instance in a relative clause, it is again the same as adjectives "I see a man who eats" see (I, man)&eat(man)
I hope this has cleared it up for you

About the Noun verb distinction I agree completely: There is definitely a syntactic and morphological difference between verbs and nouns in Hueyapan Nahuatl. But the difference is not in their ability to form predicates - just in the way they do it. And yes I agree that for those who would argue that there is no difference between the two the fact that the past requires copula is basic evidence against their position that they would somehow have to deal with. (they do)

How do they do that? I am a little skeptical because it happened more than once that I researched a language that supposedly didn't make the difference between nouns and verbs only to find out that there were plenty of differences between nouns and verbs but that nouns could be used predicatively without a copula and that was supposedly why they were exactly like verbs. Not needing a copula is something that is very, very common. In English even you don't need a copula in secondary predication "I made my horse a consul" vs "I made my consul sign this treaty" and the standard analysis for "be" in Generative Grammar is that it is just the bearer of tense and subject agreement

I don't know what you mean when you say that HN nouns could be interpreted as being marked overtly for case - the absolutive case only marks that something is an unpossessed noun. Some "adjectives" (e.g. "good") are marked with the absolutive suffix - and many nouns aren't.

OK, that was something that I hadn't noticed when reading your article, but I should have. Still, in Generative grammar that something is marked for case is an important argument for it being an NP (and since "adjectives" may have overt case and some "nouns" may not have overt case an important argument that there is no adjective-noun distinction in HN)

I admit that it is likely my lack of familiarity with generative grammar that makes me not get this right away. But I am actually really interested in understanding that point of view which is also why I've tried to engage with Baker even though it is a little alien to me (and sometimes I feel it violates the data). If you have time to explain I'd be happy to listen.

1)Let me try to explain the case story one more time, and I hope you will get it this time. I'll tell you how it worked in G&B, and with more details. The first idea is that NP's need to have case, which is overt in some languages and covert in other. Then there are two types of case, inherent and structural. NP's that have inherent case have case by virtue of their Theta role, for instance Ablative in Latin or the locative cases in Uralic languages. NP's that have structural case, such as accusative or nominative, have to receive that case from a case assigner, in English that is the inflection for nominative case, and transitive verbs and prepositions for accusative case. This basic idea is still more or less accepted by people in mainstream generative grammar, although there are always people arguing against it, and there are lots of little problems with it. But I think that the gist of it is correct.
2)If I were to advise someone who wants to study linguistics but can't choose between a university with a functionalist background and one with a generative background, I would advise this person to go to the generative university all thing equal. I come from a generative background and still it is more common for a generative article to go over my head than for a functionalist article to go over my head.
3) Baker is certainly a very original thinker within linguistics whose ideas are worth taking in and reflecting upon, (my idea of the meaning of adjectives is based on an idea of him), but he tends to be a bit old-fashioned in that when a principle was popular in the 1980's or 90's, but the generative linguistic community has changed its mind about it, and with good empirical reasons, Baker still uses it and most of his ideas are not accepted by most generative linguists. But what is more problematic (because what would you care about generative linguistics infighting?) is that many of his ideas, when you think them through, are just not consistent with the data. But, as I said, he is certainly worth reading.

You are right that modifying nouns isn't necessarily the most basic function for adjectives even in the languages where they do so frequently. I also don't take this as a necessary evidence that they aren't adjectives - but if they in fact do not modify nouns at all which is a distinct possibility depending on whether they are better interpreted as relative clauses then I think I also don't think it makes sense to consider them a separate adjective class.

In Zulu adjectives that modify a noun have the morphology of a relative clause, and in my MA thesis I described them as being relative clauses (I only dedicated a small paragraph to them though) and I recall reading (I believe in a book by Baker coincidently) that in a lot of languages adjectives can only modify nouns if they are in a relative clause, and despite that, they are definitely a separate category. (In Zulu there is one guy, whose name escapes me, but who certainly isn't dumb, who argues that Zulu adjectives are really nouns, but that is completely bovine feces. There are lots and lots of morphological and syntactical differences, and the fact that in one category of adjectives in some noun classes, but not all, the agreement morphemes are the same as the noun class prefixes of nouns, doesn't nullify that, nor does the fact that some adjectives in the other category of adjectives can be shown to be nouns historically)

[Radagast revived]

I hope this has cleared it up for you

Yes thank you - that was a very clear summary of what you meant.

2)If I were to advise someone who wants to study linguistics but can't choose between a university with a functionalist background and one with a generative background, I would advise this person to go to the generative university all thing equal. I come from a generative background and still it is more common for a generative article to go over my head than for a functionalist article to go over my head.
3) Baker is certainly a very original thinker within linguistics whose ideas are worth taking in and reflecting upon, (my idea of the meaning of adjectives is based on an idea of him), but he tends to be a bit old-fashioned in that when a principle was popular in the 1980's or 90's, but the generative linguistic community has changed its mind about it, and with good empirical reasons, Baker still uses it and most of his ideas are not accepted by most generative linguists. But what is more problematic (because what would you care about generative linguistics infighting?) is that many of his ideas, when you think them through, are just not consistent with the data. But, as I said, he is certainly worth reading.

2. I would very much advise in the opposite direction. I think time has run from the generative approach and it just doesn't know yet. I very much doubt that generativism will have a presence in mainstream linguistics 50 years from now.
3. I agree that Baker does violence to his data at times - i think this is a problem with a lot of generativist grammar - which I often find to be wilfully ignoring/rejecting the possibility of diversity/variability in order to fit data into predetermined formal descriptions. And the reason I like Baker is because I think he at least acknowledges the existence of languages that are not well described by the same tools devised basically for describing English - and he has the highest respect for and undertsnading of the quality of data and the limitations it sets on interpretation. In that I think he is similar to someone like Ken Hale who was also interested in finding novel ways of applying generative descriptions to languages that are just different from the ones studied by most generativists and see where that would take the generative theory. In my other article I defend Baker's description of Nahuatl against another generativist who uses the most naive interpretation of really crappy data to reject the possibility of the polysynthesis parameter. In that case he does more violence to the data than Baker does - and then extends data from one dialect (2 highly bilingual speakers) not just to the entire community (including monoglots) but to the enitre langauge group of 1,5 million people and then on to the entire possibility of having a polysynthesis parameter (e.g. he also uses it to reject Baker's Mohawk data). To be fair it was a youthful article by a recent PhD who didn't work with Nahuatl per se but with bilingualism - but I think the nonchalant relation to data is characteristic of a lot of formalist linguistics (not just generativism but also typological work and perhaps most work in phonology).

[merijn]

2. I would very much advise in the opposite direction. I think time has run from the generative approach and it just doesn't know yet. I very much doubt that generativism will have a presence in mainstream linguistics 50 years from now.

Well, first of all, when you are studying linguistics now, you're not learning to be a linguist in 50 years time, but rather in 10 years time. Second of all, I agree that generative linguistics in its current form will not be a presence in mainstream linguistics 50 years from now, but neither will most forms of functionalist linguistics. However, I think there will be still linguists who in essence ask the same question generative linguists ask; what model will explain what expressions are grammatical and what expressions are ungrammatical. And those linguists will build on work done by amongst others generative linguists. In my experience it is all too common to ignore ideas of another framework, (I have been told once by a teacher that a certain linguist wasn't worth reading because she was a functionalist) which I think is a very silly thing to do and IMO it is much easier to start from generative linguistics and then read functionalist literature and then the other way around.

3. I agree that Baker does violence to his data at times - i think this is a problem with a lot of generativist grammar - which I often find to be wilfully ignoring/rejecting the possibility of diversity/variability in order to fit data into predetermined formal descriptions. And the reason I like Baker is because I think he at least acknowledges the existence of languages that are not well described by the same tools devised basically for describing English - and he has the highest respect for and undertsnading of the quality of data and the limitations it sets on interpretation. In that I think he is similar to someone like Ken Hale who was also interested in finding novel ways of applying generative descriptions to languages that are just different from the ones studied by most generativists and see where that would take the generative theory. In my other article I defend Baker's description of Nahuatl against another generativist who uses the most naive interpretation of really crappy data to reject the possibility of the polysynthesis parameter. In that case he does more violence to the data than Baker does - and then extends data from one dialect (2 highly bilingual speakers) not just to the entire community (including monoglots) but to the enitre langauge group of 1,5 million people and then on to the entire possibility of having a polysynthesis parameter (e.g. he also uses it to reject Baker's Mohawk data). To be fair it was a youthful article by a recent PhD who didn't work with Nahuatl per se but with bilingualism - but I think the nonchalant relation to data is characteristic of a lot of formalist linguistics (not just generativism but also typological work and perhaps most work in phonology).

It is not my experience that it is rare for a generative linguist to be interested in non-western languages. Most linguists I know who work on the syntax of Zulu do so from a generative point of view. What in my experience is more common is that generative linguists focus on one language, and quite often on one aspect of that language whereas functionalists are more likely to focus on a group of languages, and show little specialization in what subject they cover. Generative linguistics is a framework that encourages to look at every little detail of a certain construction of certain language, whereas functionalist linguistics is better to explain the bigger picture. Of course it depends, but generally there is very little research a generative linguist can do that s/he can do by consulting grammar books, the data that are needed are too fine-grained. That is because they ask different questions, the question a generative linguist asks is "what model generates the grammatical expressions of a language and only those?" whereas a functionalist asks (although it is much harder to generalize for functionalists, because as I heard it say, "they are like protestants, they only agree that they disagree with the pope (Chomsky)") "what evolutionary forces can explain that the language ended up this way?"

I think your perceived sloppiness for the data comes from two sources. One is that generative analyses are more easily refuted than functionalist analyses. If you say that it is encoded in the UG that all languages are SVO a language with another word order completely refutes your thesis, whereas if a you say that it a tendency for languages that they are SVO one SOV language is no problem. On top of that, in Generative linguists it is often the case that if you make one change in the model to explain how subject agreement with quirky subjects works in Icelandic, it also affects the way you have to analyze the licensing of NP's that are negative polarity items in Zulu (this is a real-life example). So it is much harder to make an analysis that does not violate the data at some point.
The second source is that I do think that Generative Linguists tend to publish before they have looked through if it accounts for all the possible data. I think that you can express the attitude behind it best by saying that Generative Linguistics is work in progress. If a new idea doesn't explain all the data, it is still better than no idea at all, and even if it fares worse than the previous idea for most data, but better for one particular set of data, it may still be considered a good article, because it may lead to analysis that does explain all the data better. A lot of important concepts are introduced with attitude "let's suppose this is the case and see how far this brings us" rather than "this is the case".

[Radagast revived]

Well I suspect you're right that it is easier for a generativist to read functional literature than the other way round. The reason I think functional frameworks will be around in 50 years but generative one's wont' is that currently 90% of language documentation is carried out in a functional [ostensibly theory free] framework.

[chris_notts]

I assume Stassen includes aspect along with tense in his calculations?

He only claims that an (obligatory?) past vs non-past distinction in verbs means that adjectives will be non-verbal. Obviously this is a statistical universal so there will be some exceptions, including Japanese and Korean. He explicitly doesn't include languages where verbs inflect for aspect in the set of languages which should have nouny adjectives.

In any case, your examples suggest that Algonquian languages may be an exception. If I look up what Stassen says in his book about them, then I find that the examples from this family in his sample are claimed to mark aspectual and modal notions rather than tense. His sample contains the following languages from Algic: Yurok, Blackfoot, Cree, Kickapoo, Menomini, and Ojibwa. He does discuss some 'Problematic Cases', but I don't think any of those languages are discussed in that section.

[chris_notts]

Well I suspect you're right that it is easier for a generativist to read functional literature than the other way round. The reason I think functional frameworks will be around in 50 years but generative one's wont' is that currently 90% of language documentation is carried out in a functional [ostensibly theory free] framework.

And the ones that aren't tend to obscure the way the language actually works and, as you say, bend it into bizarre shapes to make it fit.

[Whimemsz]

I assume Stassen includes aspect along with tense in his calculations?

He only claims that an (obligatory?) past vs non-past distinction in verbs means that adjectives will be non-verbal. Obviously this is a statistical universal so there will be some exceptions, including Japanese and Korean. He explicitly doesn't include languages where verbs inflect for aspect in the set of languages which should have nouny adjectives.

In any case, your examples suggest that Algonquian languages may be an exception. If I look up what Stassen says in his book about them, then I find that the examples from this family in his sample are claimed to mark aspectual and modal notions rather than tense. His sample contains the following languages from Algic: Yurok, Blackfoot, Cree, Kickapoo, Menomini, and Ojibwa. He does discuss some 'Problematic Cases', but I don't think any of those languages are discussed in that section.

Unfortunately not a lot has been written on tense/aspect in Algonquian (at least not that I've found). But I'm confident that Ojibwe gii- is primarily a tense marker; Cree kî- seems to be entirely parallel, as does Potawatomi ki- (e.g., Pot. ki-pyé "s/he came"). Some other Algonquian languages are also described as having past (and other) tense markers. For example, Cheyenne h- (návóómo, "I see/saw him [unmarked for tense]"; náhvóómo, "I saw him"), Blackfoot ii- (one of several complex ways of marking past tense, e.g. nitsííkska'si, "I ran"), Fox kiih-, etc. All of these seem to reflect Proto-Algonquian *ki:h-. The preterit *-(e)pan is also widespread and apparently was originally "some sort of" past-tense marker.