A naturalistic path to a language without nominals

[dɮ the phoneme]

I've mused a number of times before that, upon a little thought, I can see a relatively naturalistic and straightforward path by which a language could eliminate its class of nominals entirely, and end up with only verbs. As far as I know, this is not (unambiguously, uncontroversially) attested. The last time I posed my thoughts on this, and mused on how it could arise, dhok said to me that he would be tempted to simply re-analyze various bits of the "verbal" morphology as case marking, and thus claim the language does have nominals after all. What I have set out to do here is therefore as follows: I want to sketch in more detail this verb-only language and the diachronic path by which it could arise, and I want to put to the board (including dhok, if he is so inclined) the challenge of producing a synchronic analysis of the resulting language in which, indeed, there are nominals after all and the various bits of morphology I am positing as verbal are in fact merely case marking. I am curious if such an analysis can be made, or if the resulting language is genuinely structurally verb-only in a way that cannot be reduced or explained away. If so, I think the question "why has this never arisen naturally?" is a somewhat interesting one.

Diachronic Development

Right, we start with a language that has polypersonal agreement and converbs. This is probably means the language is head final. Let's say word order is SOV, and the basic verb template is SUBJ.AGR—OBJ.AGR—ROOT—CONVERB. Obviously there could and probably would be more complexity here, perhaps TAM marking or so on, but it's not needed to illustrate the point. The language also makes a proximate/obviate distinction, where the obviate or "4th person" is used to mark a backgrounded or less discourse-central 3rd person argument. Let's say the agreement affixes are as follows:

Code: Select all

	Subject		Object
1	na-	 	nu-
2	ta-		tu
3	∅-		∅-
4	ka-		ku-

This is excessively regular, and you'd probably expect to see some kind of number marking too, but again those things aren't necessary to illustrate the point. You can image these are the singular forms, and I've just left out the plurals, if you want. Third person arguments being zero-marked is not essential here, but I believe it makes the diachronic shifts that I'm going to posit more believable by making certain nominal and verbal forms look more similar to each other. One way or another, zero-marking of 3rd persons is quite common crosslinguistically.

Alignment here is nominative-accusative, and intransitive verbs simply take the subject agreement prefixes.

Finally, let's say that the converbial or conjunctive suffix is -i. Main verbs simply leave the converb slot in the verb template empty.

So you might have sentences like

Code: Select all

tap   tu  na-∅-kip-i   sak  na-∅-fan
store LOC 1A-3P-go-CNV fish 1A-3P-buy
"I go to the store and buy the fish"

All pretty natural so far. This is basically just Japanese if it had prefixed argument indexing.

Right, now suppose that there is no copula, and that nouns can simply be conjugated like verbs for an equative sense. This is also variously attested in the languages of the world; Nahuatl comes to mind.

Code: Select all

sak  na-∅-fan
fish 1-3-buy
"I buy a fish"

na-sak
1-fish
"I am a fish"

Now here is where something interesting can happen. Using the converbial suffix, you can get various types of appositional and related senses out of nouns:

Code: Select all

na-sak-i   wat   na-tun
1-fish-CNV water 1-like
"As a fish, I like water" (lit. "Me being a fish, I like water")

rap   ∅-nang-i   bam ∅-tob
raven 3-bird-CNV away 3-fly
"the raven, being a bird, flies away"

Again, I am told that Nahuatl does this.

Among languages with noun incorporation, there is a tendency to use free nominals to introduce new information, and to use incorporated nominals to refer back to topical or old information. In a similar vein, I can imagine a language like this introducing new information with a free nominal, and then referring back to it with a verbalized nominal in the converbial form. This is a key step in the diachronics here, but it seems to me an fairly natural one.

Code: Select all

(1) rap,  wol sa   ∅-man-i,   kang tu  ∅-tob
    raven fox from 3-flee-CNV sky  LOC 3-fly
"Raven, fleeing from Fox, flew into the sky"

(2) ya  ∅-rap-i     samfam     ∅-dup
    and 3-raven-CNV north.star 3-become
"and Raven became the north star" (lit. and he, Raven, became the north star)

Ok, everything is now set up to fully eliminate subject nominals. Topic constructions often seem to be generalized into default or unmarked constructions, since we so often find ourselves talking about things that are topical. The quantum leap here is to reinterpret the converbial constructions as in (2) as the least marked form of clause, with free nominals as rap in (1) reinterpreted as 3rd person verbs merely adjunct to the clause. Free nominals as adjuncts is already a preferred analysis of many polysynthetic languages, so maybe you already have that for free. One way or another, in doing this, you eliminate nominals as subjects. This is the crucial step. This is also where I believe 3rd person arguments being zero-marked on the verb makes things "easier", since the resulting construction looks almost like a nominal here anyway. But you could imagine the same system, maybe, with explicitly cross-referenced third person arguments. In that scenario, I believe it becomes even harder to synchronically analyze the final product as in fact having nominals-in-disguise, but we'll get there when we get there.

Right, so, we have successfully eliminated subject nominals. But object nominals are still a necessity. A sentence like

Code: Select all

∅-log-i   ∅-kip-i  ∅-geg-i   ∅-∅-fan
∅-man-CNV 1-go-CNV 3-egg-CNV 3A-3P-buy

does not mean "a man goes and buys eggs", but rather "a man goes and is an egg and buys (something)". You still need independent nominals in object position to get the desired meaning:

Code: Select all

∅-log-i   ∅-kip-i  geg ∅-∅-fan
∅-man-CNV 1-go-CNV egg 3A-3P-buy
"A man goes and buys eggs"

But! This is where the proximate/obviate distinction comes in! You can say

Code: Select all

∅-log-i   ∅-kip-i  ka-geg-i  ∅-ku-fan
∅-man-CNV 1-go-CNV 4-egg-CNV 3A-4P-buy
"A man goes and buys eggs" (lit. He, a man, goes and they.OBV are eggs and he buys them.OBV)

And perhaps, although this is less precedented in natlangs I believe, we can imagine that this construction would be preferred when the eggs are old information. And, if this gets used enough, the same reanalysis could take place, with this construction interpreted as the default and free nominals falling out of use.

What we end up with, then, is a language where

1. there is no morphologically distinct class of nominals,

and

2. argument cross referencing and clausal conjunction interact in complex ways that I believe cannot easily be analyzed simply as a system of case marking.

The Final Result

The final result is a language with the following features:

We have the same set of verbal agreement markers as before:

Code: Select all

	Subject		Object
1	na-	 	nu-
2	ta-		tu
3	∅-		∅-
4	ka-		ku-

And the same converbial suffix, -i. This suffix can be used to conjoin clauses, and to indicate sequential actions. It also obligatorily appears on (transitive or intransitive) subjects and objects. 3rd person objects take the prefix ka-, which is the same as the verbal obviate subject agreement prefix. Transitive verbs with explicit third person objects correspondingly take the obviate object prefix ku-. Transitive verbs without an independent "nominal" object argument take instead the 3rd or 4th person object prefix on the basis of semantics. In equative clauses, the first element takes the CNV suffix and the second element is unmarked.

More: show

Code: Select all

Vocab:

gan  "to see"	
mel  "to hear"	
hit  "to hit"	
tang "fruit"	
kaw  "to eat"	
rap  "raven"			
pap  "to cry"				
law  "to laugh"
nang "bird"
		
Data:						

na-tu-gan
"I see you"

na-gan
"I see it"

gan
"it.PROX sees itself"

ku-gan
"it sees it.OBV"

ka-gan
"it.OBV sees it.PROX"

na-tu-gan-i na-tu-mel
"I see you and hear you"

na-tu-hit-i ta-pap
"I hit you and you cry"

na-gan-i na-kaw
"I see it and I eat it"

na-gan-i na-ku-kaw
"I see it.PROX and I eat it.OBV"

ka-gan-i ka-kaw
"it.OBV sees it.PROX and eats it.PROX"

ka-tang-i na-ku-gan
"I see the fruit"

ka-tang-i na-ku-gan-i na-ku-kaw
"I see the fruit and it eat"

rap-i gan
"the raven sees it"

rap-i nu-gan
"the raven sees me"

rap-i ka-tang-i ku-kaw
"the raven eats the fruit"

rap-i nu-gan-i law
"the raven sees me and laughs"

rap-i ka-tang-i ku-gan-i ku-kaw
"the raven sees the fruit and eats it"

na-rap
"I am a raven"

ta-rap
"you are a raven"

rap
"it is a raven"

rap-i nang
"the raven is a bird"

This is already kind of weird and hard to analyze. Now imagine we had run this whole process the same, but 3rd person arguments were explicitly indexed on the verb from the get-go. We would end with something like:

Code: Select all

	Subject		Object
1	na-	 	nu-
2	ta-		tu
3	sa-		su-
4	ka-		ku-

Where, now: the suffix -i conjoins clauses, and indicates sequential actions. It also obligatorily appears on (transitive or intransitive) subjects and objects. 3rd person subjects take the prefix sa- and 3rd person objects take the prefix ka-, which are the same as the verbal proximate and obviate subject agreement prefixes respectively. Transitive verbs with explicit third person objects take the obviate object prefix ku-. Transitive verbs without an independent "nominal" object argument take instead the 3rd or 4th person object prefix on the basis of semantics. In equative clauses, the first element takes the 3rd person subject prefix and CNV suffix, and takes just the 3rd person subject prefix.

The sentences from above would read:

More: show

Code: Select all

Data:						

na-tu-gan
"I see you"

na-su-gan
"I see it"

na-tu-gan-i na-tu-mel
"I see you and hear you"

na-tu-hit-i ta-pap
"I hit you and you cry"

na-su-gan-i na-su-kaw
"I see it and I eat it"

na-su-gan-i na-ku-kaw
"I see it.PROX and I eat it.OBV"

ka-su-gan-i ka-su-kaw
"it.OBV sees it.PROX and eats it.PROX"

ka-tang-i na-ku-gan
"I see the fruit"

ka-tang-i na-ku-gan-i na-ku-kaw
"I see the fruit and it eat"

sa-rap-i sa-su-gan
"the raven sees"

sa-rap-i sa-nu-gan
"the raven sees me"

sa-rap-i ka-tang-i sa-ku-kaw
"the raven eats the fruit"

sa-rap-i sa-nu-gan-i sa-law
"the raven sees me and laughs"

sa-rap-i ka-tang-i sa-ku-gan-i sa-ku-kaw
"the raven sees the fruit and eats it"

na-rap
"I am a raven"

ta-rap
"you are a raven"

sa-rap
"it is a raven"

sa-rap-i sa-nang
"the raven is a bird"

This version of the language seems less plausible to me, but even more difficult to analyze as anything other than a purely verbal language.

The Challenge

Ok, now comes the reason I am posting this. I want to challenge people to see if they actually can make a reasonable analysis of this language, compatible with the data above (and any more example sentences you ask me to supply) that revives the presence of nominals and challenges my assessment that this is a necessarily all-verb language by its structure. This goes for both the marked-3rd-person and unmarked-3rd-person versions, which may permit different analyses. In particular, since this was all sparked by a comment of dhok's, I am curious if he as such an analysis. In any case this has been an interesting exercise, and I'm glad I went through it, so hopefully it furnishes some conlanging inspiration for others as well.

[zompist]

This is a very offhand reaction... to me you still have two classes of words: those which require arguments and those which don't. That is, "buy" seems like a clear verb, and "raven, man, egg" sure don't look like prototypical verbs.

It's pretty common for nouns to be able to be used with a copular meaning, so that's certainly no problem. On the other hand the copula is often a weird misbehaving sort of verb, so your nominals are not that verby.

If I didn't have your theory on this language I'd be very tempted to consider -i as a marker of argument status, though this may be because I don't understand what's happening when you apply it to "buy". It doesn't seem to be the same thing that's happening with "man" and "egg".

It's also typical for nouns to get more nouny in derivational morphology. Are there words for "men" or "manhood" or "thing like an egg"?

[/nɒtɛndəduːd/]

unfortunately, I don't believe it's possible to create a language without nominals, as without proper names for things you can't exactly talk about them. however, an alternative path is possible, as is inversely shown in the conlang Kēlen, in which there are only four very basic verbs which, if looked at from the right angle, can be discounted as verbs and instead considered nouns. similarly, you could theoretically come up with a conlang in which there are only a small handful of 'nouns' that can be used in the place of a large quantity of them, and could instead be considered less of a noun and more of a verb that can be used as a noun, if that makes any sense.

[dɮ the phoneme]

This is a very offhand reaction... to me you still have two classes of words: those which require arguments and those which don't. That is, "buy" seems like a clear verb, and "raven, man, egg" sure don't look like prototypical verbs.

In both versions of the language (the one with unmarked 3rd person and the one with the marked 3rd person; resp. Version 1 and Version 2), "raven" behaves exactly like any intransitive verb, and "buy" exactly like any transitive; there is no difference in morphology or marking-obligatoriness anywhere.

Version 1:

Code: Select all

rap
"it is a raven"

na-rap
"I am a raven"

ta-rap
"you are a raven"

rap-i nu-gan
"it is a raven and sees me"

ta-rap-i nu-gan
"you are a raven and see me"

etc.

law
"it laughs"

na-law
"I laugh"

ta-law
"you laugh"

law-i nu-gan
"it laughs and sees me"

ta-law-i nu-gan
"you laugh and see me"

etc.

fan
"it buys it(self)"

na-fan
"I buy it"

nu-fan
"it buys me"

ta-nu-fan
"you buy me"

ta-nu-fan-i ta-law
"you buy me and laugh"

etc.

No difference.

Respectively in Version 2:

Code: Select all

sa-rap
"it is a raven"

na-rap
"I am a raven"

ta-rap
"you are a raven"

sa-rap-i nu-gan
"it is a raven and sees me"

ta-rap-i nu-gan
"you are a raven and see me"

etc.

sa-law
"it laughs"

na-law
"I laugh"

ta-law
"you laugh"

sa-law-i nu-gan
"it laughs and sees me"

ta-law-i nu-gan
"you laugh and see me"

etc.

sa-su-fan
"it buys it(self)"

na-su-fan
"I buy it"

sa-nu-fan
"it buys me"

ta-nu-fan
"you buy me"

ta-nu-fan-i ta-law
"you buy me and laugh"

etc.

Again, the same.

Version 2 illustrates the underlying structure (at least in my analysis) of Version 1. Version 1 is precisely Version 2 with the sa- and su- 3A and 3P prefixes realized as phonologically zero. There are no other differences. But I believe Version 2 is both less naturalistic and less interesting; Version 1 may permit some unexpected alternate analyses. In Version 2, all words are obligatorily marked, with the same set of person prefixes. The only distinct word classes are transitive and intransitive. In Version 1, no words are obligatorily marked (alternate formulation: all words are obligatorily marked, but the 3A and 3P prefixes are phonologically zero), and all words have the option to take the same subject agreement prefixes.

[zompist]

This is a very offhand reaction... to me you still have two classes of words: those which require arguments and those which don't. That is, "buy" seems like a clear verb, and "raven, man, egg" sure don't look like prototypical verbs.

In both versions of the language (the one with unmarked 3rd person and the one with the marked 3rd person; resp. Version 1 and Version 2), "raven" behaves exactly like any intransitive verb, and "buy" exactly like any transitive; there is no difference in morphology or marking-obligatoriness anywhere.

You're missing the point, which is based on prototype theory. A prototypical verb is transitive. It has an extra semantic slot, more morphology, and often different syntax (e.g. sentence order). An intransitive verb is less prototypical, and a copula is even less so. (Note that if you have aspect, the copula is may be deficient.)

If you really want to muddy the waters, you probably need to think of ways the non-nouns have more than a copular meaning. You could for instance form words like many Native American languages do, as sentences: 'water' < 'it flows'. Though that might not really do it: after all, that's just etymology.

[Ares Land]

That's, indeed, reminiscent of Nahuatl -- Michel Launey calls it omnipredicative.

One interesting test is, are there parts of morphology that don't apply to (semantic) nouns? A good test case is tense.
How do you say 'it was a bird'? Can you say

Code: Select all

bird-PAST

You can't do that in Nahuatl, btw, you have to say:

Code: Select all

catca 		tōtōtl
be.there.PAST 	bird
"It was/there was a bird."

In fact, I don't know of any natlang that allows tense marking on nouns in that context. (Though I use the idea in some conlangs.)

[dɮ the phoneme]

That's, indeed, reminiscent of Nahuatl -- Michel Launey calls it omnipredicative.

One interesting test is, are there parts of morphology that don't apply to (semantic) nouns? A good test case is tense.
How do you say 'it was a bird'? Can you say

Code: Select all

bird-PAST

You can't do that in Nahuatl, btw, you have to say:

Code: Select all

catca 		tōtōtl
be.there.PAST 	bird
"It was/there was a bird."

In fact, I don't know of any natlang that allows tense marking on nouns in that context. (Though I use the idea in some conlangs.)

By fiat, in this lang, all morphology is identical between verbs and nouns. But examples like yours from Nahuatl probably answer the question of "why don't natlangs do this?", in the sense that even in the nearest-to-omnipredicative languages, there is some resistance to allowing all verbal morphology on verbalizes nominals, making some of the diachronic leaps I posited impossible or unlikely.

[Ares Land]

But examples like yours from Nahuatl probably answer the question of "why don't natlangs do this?", in the sense that even in the nearest-to-omnipredicative languages, there is some resistance to allowing all verbal morphology on verbalizes nominals, making some of the diachronic leaps I posited impossible or unlikely.

Or it might be that it's possible, but just didn't happen or we have no trace of the languages were it did happen. (Then again, I use it in a conlang family, precisely because it's unusual... for esthetic reasons, I feel a language spoken on a different planet should be a little unearthly.)

[bradrn]

It’s worth noting here that there are many natlangs with a highly reduced set of true verbs, such that words which are verbs in other languages behave more similarly to nominals. But there seem to be none with a reduced set of nouns — there are some where nouns and verbs behave very similarly, but that’s not the same thing. I’m sure this says something or other about human cognition.

[Tsimaah]

From a computer science perspective, a theoretical language has been created which has only functions,https://en.wikipedia.org/wiki/Lambda_calculus, which is at least vaguely analogous to a natlang having only verbs. That said, I don't think lambda calculus could be hacked into a denotational human language. All it is when you boil it down is a series of string-rewriting operations that a machine could use to express and calculate the answers to certain kinds of math problems, I don't think you meaningfully shove things like proper nouns or types of event like "walk" or "love" into lambda calculus because these things cannot be boiled down to calculations and string-rewriting rules.


When it come to denotational, human-usable languages, there are two levels on which a language could be said to lack a noun/verb distinction. There is the, IMO, uninteresting approach of having words with what is clearly nominal meaning zero-derive into intransitive verbs which are used to refer to membership in a set (with proper nouns, like "the sun", referring to sets with only one member). Languages constructed in this way still have semantics which distinguish between objects and relationships between them. In your example, your language still has objects ("I", "you", "3rd person", and "4th person") embedded inside it and verbs which describe the properties of and relationships between these objects.

What it would take, in my opinion, to completely obliterate nouns would be to refer to everything as an event or process and not an object. What would it take to do this with something so very noun-like as a specific human individual? Take Abraham Lincoln. On the one hand, he was a distinct object in history, on the other hand, he was a process, something which started approximately in 1809, ended in 1865, and was located in various well-defined places at each point in time in between. He could be described as a given long and skinny 4-dimensional tube-of-flesh shaped patch of spacetime opening sometime in 1808 as an tiny ovum-sized point at the moment of his conception and eventually tapering off sometime after 1865 in a kind-of-hard to define way as his corpse decayed. This definition is particularly useful as a way to get around the ship of Theseus paradox, as it is likely that there weren't any atoms that were resident in his body during that entire time period, so you'd have to define "him", roughly, as whatever atoms were located beneath the boundary of his skin at a given moment. Fuzzier objects like hurricanes are a bit easier to conceive of in this way, we might speak of there being a given, well-defined Hurricane Irma, but what truly separated it from other storms going on in the world at the time, and when did it truly start, and when did it truly end?

In this framework, words like "human" would refer simultaneously to the spacetime coordinates of every human being to exist, "Abraham Lincoln" as some small subset of these, and a word like "walk" would refer to every time and place where walking had ever took place. To describe relationships between them, the only operation would be coordinate intersection. To say "Abraham Lincoln is human." would be to say that his coordinates lie within the coordinates of all human bodies to ever exist. To say "the sun shone as it rained" would be to say that, at some point in space and time, there was a patch in which both raindrops were falling and sunlight was penetrating through. Of course, this approach has some serious flaws. What is it that distinguishes a patch of spacetime containing a human body from some random region of interstellar space? How is it that speakers of this language are defining which patches of reality they are giving names to and coming to agreement about these?

Back to the less esoteric topic of natlangs. I am an advocate of Martin Haspelmath's frameworkless linguistics, https://dlc.hypotheses.org/966, and I don't think a human language can be made nounless in such a deep, semantic way such as what I just mentioned. The notion of "no noun/verb distinction" only appears in a superficial, syntactic way when using a Chomskian approach of attempting to define different word categories within a language as "noun" or "verb" using purely distributional properties.

The grade-school definitions of nouns as "people, places, or things" and verbs as "things that happen or things you do (or also, relationships, though that one tends to be skipped)" is, I think, a perfectly valid way of thinking about things. I have yet to encounter a natlang that truly managed to collapse the semantic distinction between these two things. When I hear about some nounless or verbless language, the first thing I look into is "Well then how do you say something like 'My friend brought me some food.'? And the answer to this question has always been something completely ordinary and understandable, never some radical acid-trip mind expanding thing.


(I see after writing this that someone else already brought this up about Nahuatl)

Also, in the case of Nahuatl, there are two big (morphological) things that distinguish nouns from verbs.

Nouns can take possessive prefixes in addition to predicates: nopil: "My child(s)" vs "nipilli": "I am a child".

Only verbs can take tense: "nichocâ": "I will cry", vs "nipilz*": "I will be a child". (I think you have to conjugate a copula in this situation and use "nipilli nicatê" to speak about nouns in other tenses).




In the Algonquian languages, you can, in fact, get almost any morpheme which appears on a verb onto a noun and vice-versa. In Ojibwe and in most Algonquian languages, you can say things like "nimishoomis-iban" to mean "My grandfather, deceased"), related to "nibosigo-ban", "had he not died". You can mark pejoratives on nouns, "od-animoshi-sh-an: his stupid dog", and on verbs, (with a slightly different, pitiative meaning), "niwaabamaa-sh-inaan aw gwiiwizens mawid": "We see the poor little boy crying".

I have seen analyses which propose that the "independent order" verb inflections in these languages originally arose from possessed verbal abstract nouns, thus completely obliterating the verb/noun distinction for a subset of forms. https://www.jstor.org/stable/1264842?seq=1. What was proposed to have happened is that forms ancestral to the modern "ni-miiji-min", "we eat it", used to just mean "our food" (which modernly is the distinct ni-miijim-inaan, though the two forms become identical if you want to say "our former food/ we had eaten, ni-miiji-minaa-ban) (something still visible in the word "miijim", "food", itself, which can also mean "unspecified people eat" in some dialects. However, this development failed to eliminate the verb/noun distinction entirely, subjunctive mood verbs continued to have normal verb-only person markings, and subsequent semantic drift and analogical leveling caused the independent order verbal and possessed nominal inflectional paradigms to drift apart over time.

[zompist]

Just responding to one point in your excellent post.

In this framework, words like "human" would refer simultaneously to the spacetime coordinates of every human being to exist, "Abraham Lincoln" as some small subset of these, and a word like "walk" would refer to every time and place where walking had ever took place. To describe relationships between them, the only operation would be coordinate intersection. To say "Abraham Lincoln is human." would be to say that his coordinates lie within the coordinates of all human bodies to ever exist. To say "the sun shone as it rained" would be to say that, at some point in space and time, there was a patch in which both raindrops were falling and sunlight was penetrating through. Of course, this approach has some serious flaws. What is it that distinguishes a patch of spacetime containing a human body from some random region of interstellar space? How is it that speakers of this language are defining which patches of reality they are giving names to and coming to agreement about these?

I do enjoy some speculative semantics. This sort of scheme could certainly be used for a conlang— semantic theories don't have to be watertight. But I don't think it really solves many semantic problems. The first problem is with edge cases.

Is Sherlock Holmes human? Not in the above framework, since he is not part of the physical universe. Yet it seems absurd to claim that he wasn't human. And such things aren't limited to fiction; they come up every time we have an irrealis condition. ("If I hadn't stopped for that chai latte, I'd wouldn't have missed that flight.")

Is a Neanderthal human? An Australopithecine? You can't answer that without getting into the sort of features and judgments that your spacetime coordinates approach seems designed to avoid.

On the other hand, I think your final question is far easier than it looks, because evolution has answered it for us. Organisms down to the bacterium level have, for 500 million years, worked out heuristics for deciding what is ourselves, what is a threat, what is an ally, what is food, what can be ignored. It's a difficult problem (imagine programming it procedurally) but our brains solved it for us, and language just takes advantage of this ability.

[sasasha]

The Challenge

Ok, now comes the reason I am posting this. I want to challenge people to see if they actually can make a reasonable analysis of this language, compatible with the data above (and any more example sentences you ask me to supply) that revives the presence of nominals and challenges my assessment that this is a necessarily all-verb language by its structure. This goes for both the marked-3rd-person and unmarked-3rd-person versions, which may permit different analyses. In particular, since this was all sparked by a comment of dhok's, I am curious if he as such an analysis. In any case this has been an interesting exercise, and I'm glad I went through it, so hopefully it furnishes some conlanging inspiration for others as well.

I like what you've done. I don't think it's nounless.

Your nouns are still "raven", "sun", etc.

The affixes that express their case relationships (-i, ka-, ∅- etc.) are identical to and derive from verbal morphology.

So what? Affixes with distinct jobs don't have to be phonologically distinct or have distinct histories.

I might well be missing something here. I have a 'verbless' conlang, Peoppaeq, but I'm not adamant that it is actually verbless; only that diachronically it has developed its functionality exclusively out of what was previously nominal morphology.

(By the way, Peoppaeq does make use of fairly 'speculative semantics'. I try not to have dummy 'non-verbs', like you're claiming your rap is a 'non-noun'. Every action/process must be presented as an interaction of substantives e.g. 'a roof is leaking' might be expressed 'water to house-floor from sky'. So I think it goes further in trying to eliminate a whole semantic class than your example. But I still wouldn't claim it has actually done it.)

I reiterate that what you've laid out is interesting and I enjoy the way it works; I particularly like option 1 with the zero 3rd person marking.

[bradrn]

So what? Affixes with distinct jobs don't have to be phonologically distinct or have distinct histories.

The difficulty here is in defining ‘distinct jobs’. One of my favourite papers is on this point: namely David Gil’s Escaping Eurocentrism: fieldwork as a process of unlearning. It’s well worth reading — I think his first example (the one on Hokkien) is a particularly good case study of how ‘distinct jobs’ can turn out to be not so distinct depending on who you ask.

But perhaps more relevant here is his second example, on word classes in Tagalog. His conclusion that ‘Tagalog simply does without the traditional parts of speech’ is probably too broad: I’ve heard that there are syntactic tests which can distinguish nouns and verbs in Tagalog. But it’s true that the distinction seems much less relevant in Tagalog than it is in some other languages.

This leads me to mention something else: if a language is ‘nounless’, presumably it has a single part of speech, which we’re calling ‘verbs’. But there’s no reason that we can’t call that single part of speech ‘nouns’! After all, it’s not like there are any other parts of speech to do a language-internal comparison with… and trying to compare parts of speech across languages is rarely a productive process. So in this case I think it’s better to use wording similar to that of Gil — ‘part of speech is not a relevant concept in this language’.

[sasasha]

So what? Affixes with distinct jobs don't have to be phonologically distinct or have distinct histories.

The difficulty here is in defining ‘distinct jobs’. One of my favourite papers is on this point: namely David Gil’s Escaping Eurocentrism: fieldwork as a process of unlearning. It’s well worth reading — I think his first example (the one on Hokkien) is a particularly good case study of how ‘distinct jobs’ can turn out to not be so distinct depending on who you ask.

[...]

So in this case I think it’s better to use wording similar to that of Gil — ‘part of speech is not a relevant concept in this language’.

Useful points, and I like the paper, and broadly agree. I should have said, "I don't think it's necessarily nounless."

As I tried to imply in my further paragraphs (though I could have done a better job), I think it's folly to make definite claims that a language absolutely lacks classes of verbs or nouns.

The original question was 'can someone analyse this as still having nominals?' to which my response was directed.

In this case, within the class of 'verbs' there is a strongly represented subset with substantive+copula meanings. Whilst typologically they may or may not prove distinct, semantically, they are distinct, as Tsimaah nicely illustrated.

[bradrn]

The original question was 'can someone analyse this as still having nominals?' to which my response was directed.

Well, in that case, my answer is: yes they can! If a language has only a single word class, I could label this class ‘nominal’ as easily as ‘verbal’, and I don’t see how there’s any grounds for calling one label better than the other. (Of course, I’d argue that both are worse than ‘no word classes at all’.)

[sasasha]

The original question was 'can someone analyse this as still having nominals?' to which my response was directed.

Well, in that case, my answer is: yes they can! If a language has only a single word class, I could label this class ‘nominal’ as easily as ‘verbal’, and I don’t see how there’s any grounds for calling one label better than the other. (Of course, I’d argue that both are worse than ‘no word classes at all’.)

Sure! That too.

Another way into this: the ‘pseudo-nominals’ here are clearly derived from a class of nominals in the parent language. Even if it is possible to analyze the result as not a distinct class of nominals, it’s hardly Eurocentric to judge them as basically the same thing as nouns. It the-parent-language-centric. It’s appropriate, in other words.

[Tsimaah]

I wonder how you'd get nouns to syntactically merge not just with intransitive verbs, but also with transitive, or even ditransitive, verbs?

To get it done intransitively is easy (here I give examples using English words)

Dog bark: The dog barks
Dog Golden Retriever: The dog is a golden retriever.

For the purpose of this exercise I think you should stick with working with countable common nouns that don't refer to relationships.

These, IMO, are the nouniest nouns. If you do things like

"Girl student teacher" to mean something like "The girl is a student of the teacher", you are exploiting the verb-like relational
semantics of the noun "student". In a language like Mohawk, a lot of familial relationships are expressed with transitive verbs for precisely this reason. Even in English, you can say things like "He sired me", rather than "He is my biological father" (even if such verbified family language is usually semantically narrowed to pedigreed dogs or racehorses).

I think you've got to stick with nouns whose core meaning is tied strictly to shape, species, and other primarily physical characteristics.

There's a lot of ways to do it but not any I see at the moment that could consistently apply to most nouns.

You could do by usage (only works for tools):
"Girl frisbee dog" (girl throws a frisbee to the dog, i.e., uses the frisbee),
"Car ice-scraper windshield (ice-scrape the car's windshield)"

Chained copula (works for a lot of things, but is a pretty marginal construction, these aren't the sorts of things people have reason to say very often)

"Buster dog animal": Buster is a dog, which is a kind of animal

Possession plus copula, but that, to me, seems to clearly parse as a subset of the intransitive copula, with (possessor possessum) forming a smaller noun phrase

"Jackass car BMW" (That jackass's car is a BMW)

Anyone have any other ways they can think of doing this?

[dɮ the phoneme]

This leads me to mention something else: if a language is ‘nounless’, presumably it has a single part of speech, which we’re calling ‘verbs’. But there’s no reason that we can’t call that single part of speech ‘nouns’! After all, it’s not like there are any other parts of speech to do a language-internal comparison with… and trying to compare parts of speech across languages is rarely a productive process. So in this case I think it’s better to use wording similar to that of Gil — ‘part of speech is not a relevant concept in this language’.

Agreed. I was being a bit sloppy with terminology, since the single word class in the daughterlang has morphology derived exclusively from the verbs of the protolang.

The original question was 'can someone analyse this as still having nominals?' to which my response was directed.

Well, in that case, my answer is: yes they can! If a language has only a single word class, I could label this class ‘nominal’ as easily as ‘verbal’, and I don’t see how there’s any grounds for calling one label better than the other. (Of course, I’d argue that both are worse than ‘no word classes at all’.)

What I really meant to ask "can anyone produce an analysis in which there are two distinct word classes, such that one has morphology structurally similar to what one sees on nouns in other languages, and one has morphology structurally similar to what one sees on verbs in other languages; i.e. can you make the marking of the object-referring intransitive verbs (per my analysis) look like case marking or other typical nominal marking by relabeling it in the right way?". Hopefully this question leaves less ambiguity.

[sasasha]

Something like this:

Language’s nouns are substantives, and follow a nominative/accusative alignment. The nominative suffix is -i. The accusative affixes are ka-...-i (triggering agreement on transitive verbs in ku-).

Nouns exhibit predicative behaviour when unmarked or in equative clauses. They can be modified by the personal prefixes when one of the elements of an equative statement is a pronoun... [etc]

It may be noted that in form and behaviour the case marking affixes are identical to the converb suffix and the 3rd person subject prefix, and this is indeed their derivation. In some analyses there are no distinct classes of nouns and verbs in Language. However, in this analysis, the large set of words which can take no arguments (except in the case of equative statements) and have substantive meanings rather than referring to actions or procedures are designated nouns.