I've been playing with some clong ideas in my head for a while now and before they get all mushy and soggy and disintegrate through the great sieve of the mind, maybe it would be nice to put them here and get some feedback.
The main inspiration was a screenshot from Xu Bing's "Book From the Ground". It's a book that tells a story via emoji only, in a language everyone is supposed to understand on first read. (I will try uploading the screenshot, but currently it doesn't seem to work.)
In the screenshot i saw that the glyph for movement is used in two ways: <agent><moves><place> ~ The agent goes (in-)to the place. <place><moves><agent> ~ The agent moves from the place.
[Edit: My original attempt of marking things conflicts with HTML, so trying to resolve that...]
So the idea then is this: Predicates describe different but related properties/relationships, depending on what type of object is put into their slots, and which slots are used.
Types would probably be somewhat animacy-related, such that for example for a verb of actions X, the default assumption is that if you have an animate object A and an inanimate object B filling the slots, the animate does the action to or towards the inanimate for A X B, and some different but related action for B X A.
There probably needs to be type conversion to be able to speak about asymmetric predicates between objects of the same type. (would "Alice walk Bob" mean Alice goes to Bob, or Bob goes away from Alice (or both?)?)
Saying something like "Alice Bob walk" could mean that Alice and Bob walk somewhere together.
On the other hand, "Alice Bob walk park home" could mean "Alice walks to the park, Boh walks home."
(I might add more later, feel free to voice ideas on what's already here!)
Another idea is the feature I call âanti garden-pathâ, which is not a super accurate name, but Iâll explain in more detail:
Assume a well-formed sentence S1 made up of morphemes [x1, x2, âŠ, xn] (This is an ordered list).
Now every well-formed sub-sentence S2 [x1, x2, âŠ, xm] with m<n must be implied by S1.
This means that adding morphemes or words to a sentence may only make it more specific, and never change its meaning in a way that contradicts a shorter version of the sentence.
One thing Iâm wondering (and expecting not to be the case) is whether any transitive predicate has a somewhat reasonable alternative reading. (We can probably dismiss the idea of uniqueness of the alternative reading).
Everything that could be interpreted as some kind of (literal or figurative) motion is pretty straightforward:
X <move> Y: X moves to(wards) Y // Y moves (on) from X X <learn> Y: X learns Y // Y forgets X X <gather> Y: X gathers Y // Y scatters X
âŠ
(Apparently, the alternative reading here is the inverse âmotionâ.)
Other things are a bit less clear:
X <see> Y: X sees Y // Y looks somewhere from X? X <break> Y: X breaks Y // Y mends X?? X <write> Y: X writes Y // Y reads X??
This is a very intriguing feature to me! The well-formed in âevery well-formed sub-sentenceâ will need to pull a lot of weight, Iâd expect. Interestingly, I think this is pretty distinct from the usual parsing meaning of âgarden-pathâ. Well, it depends when you consider a sub-sentence well-formed, I suppose. As an example, is âthe complex housesâ a well-formed sub-sentence in itself? If not, âthe complex houses married soldiersâ satisfies your property. I canât help but wonder how garden-path-y you could make the syntax by really pushing the well-formedness constraint to its limit.
Generally youâll need a lot of forethought to make this work. Your disjunctions at least are going to need to go up front, since they broaden the meaning of a sentence rather than narrowing down. And youâll need to be very careful about implication. For example, does âthe complex houses married soldiersâ mean that the complex houses only married soldiers? If so, you canât have afterthought conjunction either, since âthe complex houses married soldiers and their familiesâ would have more people living in the complex than âthe complex houses married soldiersâ.
The well-formed in âevery well-formed sub-sentenceâ will need to pull a lot of weight
Haha, yep! I intentionally left âwell-definedâ not well-defined, because the implications of choosing a definition range from âevery sentence ever fulfills thisâ to âyou can never negate or disjunct thingsâ.
Generally youâll need a lot of forethought to make this work.
Also spot on. Every potentially problematic feature (negation, disjunction, exclusiveness, âŠ) would need to be wrapped in parentheticals or somesuch.
New thought: What if a clong was engineered for Nonviolent Communication? What features would help with structuring expressions and dialogue in a way that would follow NVC principles?
LĂĄadan tried to introduce some features like that, I feel, e.g. speech acts and evidentials.
A preliminary idea is to have distinct structures (lexical, maybe even syntactical), to express the 4 components Observation, Feelings, Needs, Requests.
An underlying base language for general discussion, such that it is very clear what speakers are currently trying to achieve.
On top, a general expectation to express all 4 of these NVC components in order, whenever talking about interpersonal relations.
Maybe there are 4 different registers in a loglang?
Observation would try to limit the ways in which non-facts can be presented. No counterfactuals, no modals, no optatives/imperatives âŠ
Feelings and Needs would both be very basic, very reduced to those things deemed valid feelings/needs by NVC (cf. for example Wayback Machine )
Requests would once again be a fuller register, basically for optatives only.
Iâm apparently gathering vague ideas for one or multiple future conlangs here. Maybe itâll all make sense in the endâŠ
So, reading up on Latejami (thanks to @zearenâs post about it), I liked the derivational system for different verb forms.
In short, a basic concept like e.g. knowledge, is âconjugatedâ (for lack of a better word) for:
how it applies to one or more entities, which may occur as agents, patients, or âfocusâ (which is to say some oblique, non-agentive/-patientive relation).
whether it is stative (maintaining a steady state) or dynamic (changing state)
Examples:
know, patient (plus oblique), stative: âpatient X experiences a steady state of knowledge (about oblique Y)â
know, patient (plus oblique), dynamic: âpatient X learns (oblique Y)â
know, agent plus patient (plus oblique), stative âagent X maintains patient Yâs knowledge (of oblique Z)â
âŠ
Now Latejami diffetentiates these conjugations by suffixing different morphemes to (and choosing a default unmarked version of) verbs.
But what if we had (please tell me a better term for this) pseudo-verbs, that begin a sentence and only carry the information about which types of entities we want to describe, leaving the actual verbs unchanged?
Then, using <XYZ> to refer to one word, the languages structure could be like: <Agent, Patient, dynamic> <Alice> <Bob> <know>
âAlice teaches Bob.â
One positive aspect is that we get sentence fences for free, because a sentence always starts with the pseudo-verb, and ends with the actual verb.
I wanna share a Thought i've been having concerning the above-mentioned splitting of verbs into the syntactical marker of agent-patient-oblique-static/dynamic and the lexical-meaning carrying part.
First let's introduce some preliminary morphemes so this becomes a bit more fun.
some (not all) slot-structure markers: na /na/ - Ps - patient experiences steady state cu /Êu/ - Pd - patient experiences change of state gınna /gÉnËa/ - PAs - agent maintains patient's steady state gıccu /gÉÊËu/ - PAd - agent causes patient's change of state
an example of the "meaning-carriers": vois /'vÉÉȘÌŻÊ/ - knowledge-related things leu /leuÌŻ/ - life-related things
some pronouns: zim /dzim/ - 1s ba /ba/ - 2s Ćżul /zul/ - 3s
What if we analyze only this marker as a verb? Consider e.g. the sentence
gıccu zim ba vois ~ "You teach me." (You cause a change of my knowledge-state.)
We could assume that vois is an argument to the very general verb gıccu, same as zim and ba.
That would mean that arguably this language would only have very few actual verbs.
More examples: na Ćżul leu ~ "They are alive." (They experience a state of being alive.) cu ba vois ~ "You learn." (You experience a change of knowledge-state.)
What do y'all think? Is this a reasonable approach? Or would it make more sense to say the verb is vois or leu and the markers are just that, case markers?
I believe KÄlen does something very similar! It is said to have no verbs, instead having a small closed class of four "relationals" that relate nouns to each other:
So, what would a verb-less language look like? Possibly the language would have a small number of words that do the functions of verbs without any of the semantic content. In other words, words that would tell how many arguments to expect and what the relationship is between these various arguments. This is what I have done with KÄlenâ given it a closed class of "relationals" that perform the syntactic function of verbs.
There are only four relationals: LA, which asserts that an argument exists in a location or a state; NI, which asserts that an argument is or has relocated or changed its state; SE, which asserts that an argument is related to a source and/or a goal; and PA, which asserts that one argument contains another. Combine these with case markers, mood markers, and various modifiers, and we have even more ways to express the relationships between the various arguments in the sentence.
An example (remember that NI is the relational for a change in state):
ñi jacÄla jahĆ«wa;
NI N.sg(bowl) N.sg(broken)
The bowl's state has changed to "broken".
"The bowl is broken" or "The bowl broke."
I think it depends on how you end up being able to combine them, and how many there are of each. If your system is truly orthogonal and you can combine any marker with any meaning-carrier, then I don't see a reason to prefer one analysis over another. But if, say, you have 10s of markers and 100s or 1000s of meaning-carriers, and not all meaning-carriers can take all markers, then I think it will be simpler to consider, for each meaning-carrier, which markers it can take, rather than for each marker what meaning-carriers can take it. And in that case, with a small and potentially closed class of markers interacting with a large class of meaning-carriers in a way that is analyzed as mostly depending lexically on the meaning-carriers, I think it would make the most sense to consider the meaning-carriers "verbs", and the markers as some sort of grammatical particle. But I don't have much experience with the sort of linguistics wherein one thinks about what a "noun phrase" or "verb" is.
Also, I am highly intrigued by this sort of system, and I wonder if you could take it even further into the realm of grammatical cases, having meaning-carriers be merely very general types of events, and then having particles that let you elaborate on parts of their structure. So you'd have "an even of rapid movement + (subject: me) + (cause of event: loud noise) + (event is a movement away from: risk) + (event is a movement into: cave) + (event takes place within: recent past)" for something akin to "hearing thunder, I quickly escaped into a cave to avoid dying from a lightning strike". I don't think I am reinventing Ithkuil, but I do have a weak suspicion
Oh! I always wanted to look more closely at KÄlen, but until now only checked out the ceremonial writing system...
Now that you mention it I dimly remember reading that it's a verbless language in the past.
The KÄlen system does indeed look very similar to whatbI'm imagining, although the Latejami-one uses more different markers.
I see these markers more just as super general verbs, because i can very much imagine a null meaning-carrier, and then e.g. the PAd marker woukd just translate to something like "P's state is changed by A".
I imagine that all meaning-carriers do work with all markers. Effectively they just tell you whether a state is maintained or changed, and by whom.
I think your "event of rapid movement" is basically a verb to me
In the end, maybe the most reasonable explanation really is that the verb splits.
The original Latejami system also differentiates between state verbs and action verbs. I haven't written about the latter yet because I think the handling would be very parallel. They take the same combinations of agent, patient, oblique, and stative/dynamic. The difference is more in the lack of description of the resulting state after the action. II wonder currently whether the distinction is maybe unnecessary.
a! that changes the analysis completely and pushes it very much towards "the markers are the verbs"!
That was my intention
Ah, now we're again in the sphere of linguistics which is more towards "what is a good way to explain natural languages" than "how does grammar formally work": I can see what you mean by verbs splitting and how to model it formally, but I have no idea what one "ought to" call the two lexical categories you end up with for each part of the split, nor even what I would search for to find out. Like, I see "syntax" as a field as having two parts, one which is the mechanistic "formal syntax", "how are parse trees formed?" field, and one which is the "[???]" field concerned with what sort of nodes a parse tree should be able to have and what to call them, answering questions like "what is a 'determiner'? do all languages have noun phrases?", and I have no idea how to refer to that second part of the field or how to find information about it.
Like, lojban can be analyzed in a way that doesn't involve verbs or nouns, just predicates of different arity, but the people who say "lojban has verbs and nouns, actually" must have a theory of what it means for something to be a verb or a noun and a reason to prefer theories that concepts like verbs and nouns, and I would genuinely love to know where they get that from
I think some things can simply be modeled different ways without leading to different predictions by the model. Whether you call something a noun or a verb or a groopyfloop is only relevant insofar as the rules governing the language apply differently to them.
E.g. in Lojban, sumti and selbri are very similar to English's nouns and verbs, and require different syntax. But whether the resulting logical form (assuming there is one for Lojban) is different is an entirely different question I feel like.
No disagreement from me there. But some subfield of linguistics has identified modular/reusable features that seem to work across many languages, and have named them things like "noun phrases", "complementizers", and "verbs". That's what I would like to know more about.
