Adding onto this, this fact leads to a common construction in toki pona speech: pona lukin for “good-looking, beautiful” which takes advantage of the “verb > direct object of that verb” process
1 Like
I’m kind of toaq-analysis-ing this, you could in theory look at words as containing “to be X”, similar to various unary verbs in toaq
e.g. kıa kúa “the room is red” ≈ tomo li loje
I’m mainly doing this to describe some transitive constructions which do cause Y to be X e.g. mi pakala e nasin toki la nasin toki li pakala
1 Like
it could also just be "good of the eyes" and "of" can be "to" because "of" is the least specific preposition in English or something idk
1 Like
yeah that’s true too! i just seem to intuit it as satisfying that pattern ahah
This post is about @lipamanka's analysis of toki pona. I'll address the broader points about linguistics separately 
Thank you for the derivational patterns! If you have any data from your listing of which sequences work and which don't, I would love to see it, because I am interested in developing an analysis that doesn't overgenerate as much or at least prevents infinite loops.
I'm probably also gonna go back and look at all the judgments people have made about various sentences in this thread and the other one from the point of view of this analysis at some point.
Let me for the purposes of the next section split this in two, as V →p N for the patient and V →e N for the event
Let me split this as V →e A for relating to the event and V →a A for being the agent.
Then, having both V →e N and V →e A seems redundant to me, you can make one from the other with A→N and N→A. I think this hints that a simpler analysis should be possible.
I have become convinced of this analysis!
This all works in accordance with that analysis too, very nice.
Likewise this. As well as the responses from @cosmicPython's response and @selxehe's response.
These are derived as V→N, and the two different V→A, right?
I think you can analyse "mi pakala e nasin toki" as a transitive verb, and "nasin toki li pakala" either as lifting the adjective "pakala" into a verb A→V "making something be broken" or derive it from the transitive verb "pakala" as V →p N → A → V "making something be having the qualities of something someone has broken".
Again two different derivations with practically identical meaning, indicating the possibility for simplification; in this case we might be able to say that "pakala" is only a transitive verb? More generally, we can probably also dissolve the distinction between nouns and adjectives, and end up with two lexical categories of "single place predicates" and "two-place predicates" to use some loglangy terms for them. Along the lines of this:
2 Likes
yeah so here it's really easy to collapse the agentive into the general relational one, because this derivation also works with patient, locative, instrumental, dative, ablative, comparative, literally anything except for negative. like sike utala could be the ball that attacks, the ball that is attacked, the ball that is used to attack, etc.
yep! "jan wile" can also mean "wanted person" or "person used for wanting" or "person from a desire" or "person approaching a desire" etc.
First of all, I apologize for how harshly I worded things.
Let me try to explain myself: when my interest in loglangs and generally conlangs returned recently, I initially sought out the only loglang I knew of at the time (other than Loglan), lojban. Now, lojban is very very unspecified, but it has as a goal to be pretty fully specified, which leads to a lot of situations where you'll ask a question and be given several answers, sometimes mutually compatible, sometimes not, each based on a different theory of how the language "should" work, and having various degrees of compatibility with the language as observed. For this reason I, well, started learning Toaq instead, but also developed somewhat of an "allergy" to statements about how something works in a conlang that aren't clearly marked as either prescriptivist or descriptivist, and, if prescriptivist, marked for source ("this is what the refgram says", "the refgram is unclear here but this is community consensus", "this is my personal analysis", things like that). When you made your initial claims about your analysis, including the rather strong claims about intransitive verbs that both for me and others seemed (at the time; incorrectly) like they were contradicted by the observed data, it reminded me of the lojban situation.
But toki pona is not in that situation. Even pu behaves and is even to some extent treated more like just one grammar for a language that exists independently of descriptions of it. As such, freaking out over your statements of the shape "toki pona doesn't work like that" was unfounded. Apologies.
I want to be bit clearer about my point here. If L'académie Française publish a dictionary of French, they intend to stipulate that no words outside it exist. However, a descriptivist dictionary makes no such claim. Specifically, the fully general argument that "this word has not been attested, therefore it is not in the language" is not valid, as opposed to a claim like "this word has not been attested, therefore it is not in this theory of the language".
The same applies at all levels, also to, for example, the phonotactics. If a specific word form has not been attested, it could mean that it's phonotactically prohibited (even that is too strong a word, theories of linguistics really ought to be probabilistic, but I'll leave that aside), or that it is perfectly valid but simply hasn't been used yet. The argument of "this has not been attested" works to some extent, but can only go so far: absence of evidence is evidence of absence, but only weakly so (bolding this because I think it's a pretty good summary of one part of my argument).
For grammar, we're in an even stickier situation, because (my impression is that) a lot of linguists are ideologically committed to grammatical formalisms that allow infinitely many sentences, even though that would of course be impossible to observe. All such theories must thus generalize the observed data in some way. And even if you work within some bounded framework where the set of grammatical sentences really is finite, you cannot expect to have observed them all. The only alternative to generalization would be to collect all the sentences you can find and declare that your theory of the language is that it consists of exactly those sentences and no others. Any prediction that a word/sentence which has not been attested is valid, I call speculation (and this sums up the other half).
The same goes for semantics, but even worse, because what a sentence "means" becomes an almost infinitely complex maze of nuance if you attempt to pin it down fully. And you have context dependent situations like with the various poems and stories people have written with the effect that once you get to the sentence "colorless green ideas sleep furiously", it makes sense to you.
I hope this illustrates my point, to an extent that helps illuminate whether we actually agree, or, if not, where specifically we disagree?
1 Like
my problem here is primarily how these rules can be applied in pedagogy. i think it's annoying when a learner is told that they can derive usage from a set of rules they're given. like if i was going off of the basic rules of tok pisin grammar, i may assume that i can use a specific word in a specific way, but when i say it to a speaker they may misunderstand me.
in toki pona, people often take "rules" and try to derive usage from them using logic. this entire situation is my go to example for why this is bad: people will assume without checking (you checked, good on you!) that "mi moku e pan" can mean "i butter the bread." and that is incorrect BECAUSE it will be misunderstood to mean something different.
1 Like
I don't disagree with you on that being a problem at all. Where I do disagree is on where the problem lies. I can't express it much less opaquely than this, but I consider "this model is sometimes wrong" to be an issue with the model, not with the idea of trying to build models, whether or not that model is written out explicitly or only exists in the mind of the person learning. With the caveat of course that for a living language any model is going to be wrong sometimes, if for no other reason than due to language evolution. But, I believe, if a model can exist in someone's head, then it could theoretically be formalized even if it would be extremely arduous. After all, you could have the model consist of simulating their brain. But we want simpler and cheaper models than that, even at the cost of inaccuracy. For example, we may choose to disregard gradience in order to binarize binarize syntax into grammatical or ungrammatical.
1 Like
oh I agree with the part about models. I just think more about the practical applications of formal grammars more than most linguists
3 Likes
are there other derivational patterns for adverbs? like “mi pali pona e tomo” vs “mi pali pi ilo ni e tomo” or something
1 Like
In my nasin, the first one means "I build houses well", and for the second, I would say "mi pali e tomo kepeken ilo ni".
2 Likes
those are the same meanings as adjectives. toki pona doesn't have adverbs; usually, in languages, adverbs don't modify the verb directly as they do in toki pona.
However, there is the concept of an applicative, which I've thought about using instead for some modifications of predicates in toki pona, especially when they're nouns. "mi X e Z kepeken Y" can always become "mi X Y e Z" right? that's pretty cool.
1 Like
oh hey! I’ve said something like this for kepeken usage before ahah
mi sitelen kepeken nasin ni ≈ mi nasin ni e sitelen mi
1 Like
or! mi sitelen pi nasin ni
2 Likes