What is pu

[jan Ape]

This is the only confusing word to me. "Sina pu" means "you have the official toki pona book?"

[janMato]

(Edits in bold)

We're in the world of pre-constructing languages here. This is even harder than re-constructing languages like proto-indoeuropean.

Safest Possibilities:
pu - word reserved for future use, similar to how computer languages reserve the word "goto", with the plan to use it in the future or to never use it.
Pros: It is on the list, it is in fact reserved.
Cons: Doesn't help us use it, in fact, a reserved word discourages use, both as root and as proper modifier,
e.g.
?/* mi kute e nimi 'pu'. I hear the pu word. (illegal, not yet defined, if indeed reserved words are illegal to use)
?/* mi kute e nimi "Pu". Sounds just just like the potentially illegal sentence.

Loan word based theories
pu - uncarved block, taoist jargon. Assumes this is a loan word that has been promoted to root word.
pu - loan word promoted to root word, neo-eponym means "teddy bear", or proper name meaning one particular Pu. But if it is a proper name, it can't be a root word, so if it does mean teddy bears, it must be closer to the class. All someday will be moot at any rate.

Pros: Sonja's words fall not far from their trees. If it is an obvious loan word, usually the word has a meaning similar to the
Cons: Can't tell if this is from Chinese or British English.

Daoist Theories
pu - something that is hard to put into words. A big theme in Taoism is that if you can put it into words, it is false. So it could be something like "unspeakable truth" and acts as sort of a pro-form (deixis?) to stand for some unspeakable truth (or just something that is hard to express)
pu - a verbal place holder, like blah-blah-blah. It is almost as if jan Sonja didn't feel like writing "tempo ni la sina jo e lipu mute lawa pi mama pi toki pona" today and plans to fill it in later.

Fabulous Theories
pu - it means what you or the community wants it to mean (theory proposed elsewhere by someone else...), this is somewhat plausible as the community is developing idioms and assigning them meaning, why not root words?

Theories that from a long way off look like flies
...

[janKipo]

'kute' You hear it, not freeze it. 'pu' "Pu"
Actually, it is never clear what "Pooh" means other than this particular teddy bear (which is called Edward bear, not Theodore bear, in England, which opens a whole nother line of question).
pu, uncarved block, is less a truth than a person or thing that has de, a kind of pure potential to be formed by Dao. But we can't say that, of course.

[janMato]

I've realized I've left off a theory-- that "pu" is a possible clitic, maybe like a verbal comma or semi-colon.

If it is a semi-colon like English, then it would join related sentences that seem like they're missing the co-ordinating conjunction, such as For, Nor, And, But, Or, Yet, So, Like. (seems unlike, see below)
OR it joins related sentences where there is a conjuctive adverb, like furthermore, elsewhere, equally, hence, henceforth, however, just as, etc. (seems unlike, see below)
OR it splits lists of lists, e.g. apple, oranges; booze, beer, vodkas (unlikely, this would be mostly useless)

I'm not sure why it would be a co-ordinated conjuntion, because we already have "taso". And if we want to allow more, it would make more sense to overload, "en", "anu", "sama", "ala", "tawa" and allow them as sentence intials, the way "taso" is legal at the start of a sentence now.

If it is a semicolon like programming, then "pu" is a verbal period (end of sentence marker!) Unlikely, but potentially useful. Sentence splits can cause extra ambiguity.
If it is a semicolon like in Church Slavonic, then "pu" is a verbal question mark! Again, unlikely. If we get a end of sentence marker, it will cover all types of sentence endings (!, ., ..., and ?)

If it is a comma, as in in a separation of lists, that is unlikely, we have "li-chains" "e-chains" and prep-chains, already. So this is unlikely.

If it is a comma, as in the puctionation that joins the dependent and indepedent and dependent clauses, i.e. pu would be like the word "that".

Personally, I think it would be awesome to get relative clauses.

Here is why:

mi lukin e soweli pu moku e kasi pi ma supa. I saw the cow that eats the grass of the plains. I saw the grass-eating plains animal. An action is the salient feature of this animal.
This also might be usable when the salient feature is a prep-phrase (i.e. the prep phrase modifies something other than the verb)
vs
mi lukin e soweli ni: soweli li moku e kasi pi ma supa. (using e-ni-clauses) I see this animal, the animal is eating the plants of the plains. This kind of missing the point, that the salient feature of the cow is that it eats grass on the plains.
vs
mi lukin e soweli. soweli li moku e kasi pi ma supa. I see an animal. The animal is eating plans of the plains. Same problem as above and worse, it makes it sound like the topic has decidenly switched to the animals eating-- where as "pu" as relative clause marker lets the reader know that we are still talking about a cow, which may or may not be eating right now.
vs
mi lukin e soweli pi kasi moku pi ma supa. I see the animal of the edible plants of the plains. "pi" is so semantically bleached, it doesn't signal that exactly what the relationship between the soweli and the kasi is (namely that one eats the other... a cow isn't a type of plant, it doesn't have a plant, etc)

UPDATE. This also seems to be leading towards the Japanese attributive verb! Exciting stuff for the easily excitable, like myself

[janKipo]

Ends of sentences shouldn't be a problem, since we have a diostinctive juncture (assumed, of course -- I don't know just what it is; probably like English wind-down).
The various uses of 'pu' seem unlikely to be what Sonja has in mind (if anything -- the existence of 'pu' is stunningly evanescent).
The problem with relative clauses (you seem to be interested mainly in restrictive ones -- rightly, I suppose) is that it opens up intrasentential recursion and the possibility of sentences of Joycean or Proustian or Kantian complexity (and the few things we have now are bad enough though only additive). We have some devices for doing the work of relative clauses, using 'ni' in various ways, as you note. They are subject to the problems you note as well. For restrictive clauses, the rule seems to be to get the restriction in first and then go on with it. But that is probably not always possible -- or easy to do in speech, or free from the suggested problems: 'soweli li moku e kasi lon ma supa. mi lukin e soweli ni.' Not a big problem, but a slightly annoying one.

[janMato]

Ends of sentences shouldn't be a problem, since we have a diostinctive juncture (assumed, of course -- I don't know just what it is; probably like English wind-down).

Interesting implication, because this langauge was designed (like many easy languages) to preclude needing inflections, tones, irregularities, etc. So it's surprising that each of these has creaped into toki pona. I presume the verbal TP question is a tonal phenomena, the declarative sentence tone and accent. a a a must be tonal because if you say a a a with flat tone and accent (or worse, without the glottal stops) it sounds creepy.

The various uses of 'pu' seem unlikely to be what Sonja has in mind (if anything -- the existence of 'pu' is stunningly evanescent).

Most existing clitics are short CV (li, la, pi, ... e breaks the pattern), so it's very possibly some sort of clitic.

The problem with relative clauses (you seem to be interested mainly in restrictive ones -- rightly, I suppose) is that it opens up intrasentential recursion and the possibility of sentences of Joycean or Proustian or Kantian complexity (and the few things we have now are bad enough though only additive)..

With great power comes great responsiblity-- I think we can recurse without blowing our stacks.

I'm pretty sure we can recurse already in la phrases-- the syntax section on the wikipedia article a sentence can contain sub-clauses and the subclauses can contain sentences. Are "la" phrases necessarily additive? It looks like they can be intrepreted as recursive i.e. sentence can be embedded into a sentence.

soweli pi moku ala li lukin e soweli lili la lili li moli. If the hungry animal sees the small animal, the small one will die.
soweli pi moku ala li lukin e soweli lili la lili li moli la meli lili li pilin ike. If it is the case that if the hungry animal sees the small animal, the small one will die, then the little girl will be unhappy.

If prep phrases ever modify nouns should be embeddable into prep phrases (or some version of the "pi pi" rule) that would be recursion again.

mi lukin e pipi lili kepeken ilo pi lukin lili. I look at bugs with a microscope.
* mi lukin e pipi lili kepeken (ilo (lon supa pali mi)) pi lukin lili. I look at bugs with a microscope on my desk.
* mi lukin e pipi lili kepeken ((ilo pi lukin lili) (lon supa pali mi)).I look at bugs with a microscope on my desk.

I'm not smart enough to figure out this one for sure, but it feels like the result of recursion when we get a lonely noun in the middle of a pi-chain-triple

jan => person
pali ala => unemployed
jan pi pali ala => unemployed person
telo=> water
telo nasa=> peculiar water

This recursive rule, a noun phrase can be made of np+pi+np, seem obvious and necessary

? telo nasa pi jan pi pali ala. Booze of the working man.

If we could not use such a rule, then we'd get

telo nasa jan pi pali ala.... which is nonsense, both jan and nasa are describing the telo as qualities instead of some other relationship (usually a HAS-A releationship).
telo nasa pi jan wan pi pali ala.... which has an unnecessary "wan". So either tp has the same sort of peculiarity that english does where we plug in place holders (like "it's raining), or we're doing recursion. And obviously in this case the pali is modifying jan. It's a working sort of man, not a working sort of water.

'soweli li moku e kasi lon ma supa. mi lukin e soweli ni.' Not a big problem, but a slightly annoying one.

To get the above sentence to work as "I see the animal that eats grass on the plains", the first sentence would have to be interpreted as "There exists such an animal that eats grass on the plains. I see it." It's a difference that context wouldn't be able to resolve, especially if one was in a field or other places where you might find cows.

[janKipo]

I'm not quite sure why the first sentence needs this rather abstract interpretation nor why you think it lacks it. It seems to function smoothly at the concrete level, but has the more abstract as a possibility. In any case, it is the standard technique for the problem at hand and beats the alternatives for theoretical simplicity.

Working backward. Yes there is already recursion in the 'la' slot and that is a source of some concern. But it is inherently restricted, generating more and more (ambiguous) antecedents, not free-form sentences. The same is true with PP on NP constructions and, indeed, with 'pi' constructions generally. The rules are meant to provide some sort of optimal balance between clarity and economy: no clusters means a long string of simple sentences, in which one is likely to lose the thread; unrestricted clustering leads to clumps difficult to pull apart, So we are left with speaker and hearer struggling to a reasonable place. Guided by the rules of Conversation, of course. The whole can't be set in stone, but has to be worked out in each situation. I suppose that the same would apply to more free form sentential recursion, but that lure is more easily rejected (as PP on NP may even be) because the dangers are so apparent and the work-round so easy and natural (depth grammar and all).

Well, I am not sure that the structure words of tp are strictly clitics, but 'pu' would fit with the -- as would several other words already established.

The fact is that descriptions of languages (natural and constructed alike) rather spend much time on suprasegmental phenomena and tp is no exception. The only note I can find on it at all is that questions don't have to have the rising tone at the end, even though many tp questions are not the sort that have that tone in English anyhow. So I assume that tp, when spoken naturally, rather than read or squeezed out painfully in practice, has a sentence-long tune with sectional motifs of various sorts and some standard coda. I assume that that is usually that of English for the present group, although the somewhat (sometimes markedly) different contours of various Frenches and some Slavic and Germanic languages will also be heard -- with unknown effects on comprehension. We need some face-to-face somehow. In any case, every sentence has such a contour, even if it is a flat line (as it almost never is, even with remarkably unexpressive sorts).

[janMato]

I'm not quite sure why the first sentence needs this rather abstract interpretation nor why you think it lacks it. It seems to function smoothly at the concrete level, but has the more abstract as a possibility. In any case, it is the standard technique for the problem at hand and beats the alternatives for theoretical simplicity.

You're referring to this set? Enlighten me, what did I say in tp, and how was one supposed to say what I said on the english side?

soweli pi moku ala li lukin e soweli lili la lili li moli. If the hungry animal sees the small animal, the small one will die.
soweli pi moku ala li lukin e soweli lili la lili li moli la meli lili li pilin ike. If it is the case that if the hungry animal sees the small animal, the small one will die, then the little girl will be unhappy. (I.e. this dog eat dog world is distressing to watch)

The rules are meant to provide some sort of optimal balance between clarity and economy: no clusters means a long string of simple sentences, in which one is likely to lose the thread;

Exactly. A deep sentence with clauses is complex because it is a deep tree (when graphed)... [actually google wasn't much help saying what "deep grammar" means]. A long chain of sentences connected by anaphora is complex because you have to resolve more and more distant anaphora. The simpler solution is to let speakers choose between the two, sometimes a deep tree will be simpler than a long chain of independent sentences connected by anaphora.

Sometimes I feel that tp has thrown out the baby (simplicity) with the bathwater (unnecessary complexity). Too many obligatory grammatical rules, irregularities, etc. is obviously "bathwater". Relative clauses are "baby" (And anaphora that agrees with the antecedent is "baby" as well, eg. "mi lukin e jan e soweli e meli e meji. ona soweli li pona tawa mi." Agreement isn't always baroque complexity.)

Anyhow, if Japanese verbal adjectives or IE style relative clauses are obvious and necessary, then either Sonja or the community will start using them by accident or design. We'll have to wait and see.

The fact is that descriptions of languages (natural and constructed alike) rather spend much time on suprasegmental phenomena and tp is no exception. The only note I can find on it at all is that questions don't have to have the rising tone at the end, even though many tp questions are not the sort that have that tone in English anyhow.

Where'd you find the reference to that? I'll have to look up and see if suprasegmental tone, volume, stress are linguistic universals or just the way IE languages do things. If the various question patterns for IE languages are not universals, than we can't really say that toki pona necessarily uses rising final tone to indicate questions, or falling final tone to indicate indicative statements. I suspect we're in the same state as we are with respect to reduplication...potentially useful technique, but no clue what it means except in some narrow cases (accidental reduplication).

[janKipo]

Gee, I thought I did this already. well, some of my bonier mots down the hole in the center of the disc (but there aren't discs anymore, nor holes -- is a puzzlement)

I was working backwards so the first remark was to the last scrap of your piece, saying the standard restricted relative clause had to have some "There is a ..." interpretation and asking why this was necessary and also why you didn't think it was present. 'soweli li moku e kasi lon ma supa. mi lukin e soweli ni'

As for the iterated 'la's, it is not clear whether they are (p>q)>r, as you have them, or P>(q>r) or even (p>r)&(q>r) or something not describable in a linear array.

For now, it is enough that in the deep structure, every clause -- and, indeed, every modified noun (and similarly, though slightly differently for verbs, etc,) -- is represented by a fully developed S node (a sub-tree headed by S and ending in ultimate elements). At that point, there are a number of possible developments, depending on how the S nodes are connects and on the ultimate language, to the surface presentation. One extreme possibility is a string of simple sentences held together by a complex of cross-references, another -- at the opposite end -- is a single complex sentence with massive modifier clusters, relative clauses, and what all else. Somewhere between is a fairly clear yet reasonably compact utterance, maybe several sentences, but relatively transparently connected, toward which the conventions of conversation guides the speaker in the usual contexts. Not very precise but about as good as it gets. tp cuts off a part of this possibility, relative clauses, but they are ones which already lie very close to the surface and so detach easily. The point is, of course, that all these come ultimately from the same tree; it is only the path chosen to the surface that differs. Complexity, in the practical sense, is a surface phenomenon. But even as a surface phenomenon, relative clauses are scarcely less complex than connected sentences; they appear to be slightly more so in terms of what is required to get them.

Why "IE relative clauses" btw -- the IE languages display a variety of styles and the SAE ones (not even this quite accurate) are found in many other families (similarly, I think for "Japanese verbal adjectives" but I am less ure about that). Each language has its own devices for surfacing underlying structures (usually several).

The only note on tp inflection is in an old Pije lesson (I haven't checked the revisions) on tp questions -- I don't remember which kind ('seme'' or 'x ala x'). What the contours of tp sentences are/will be/ought to be I don't know. My best guess is that now thye correspond pretty much to the contours of the speaker's native language, with what result for intelligibility I don't know. What the official line (should tp take one -- a somewhat unprecedented move) will be, I wouldn't care to guess (but expect an English or one of the Frenches). But, in any case, every spoken sentence has a contour and the interesting question are now whether those in tp fall into a small number of patterns, corresponding to intentional features of the utterances involved, or are just all over the place (for each speaker and then collectively). (Incidentally, again, there aren't IE contours; each IE language (indeed, dialect) has its own and they are often unrelated as are those from other language families. Listen to a Midwestern American (I think Wisconsin, to go to an extreme, but Ohio will do) and a Parisian Frenchman ask a question, or that American and a cockney, for that matter.

[janMato]

Gee, I thought I did this already. well, some of my bonier mots down the hole in the center of the disc (but there aren't discs anymore, nor holes -- is a puzzlement)

linja pi nimi mute li jo e monsuto nimi. soweli mi li tawa lon ilo pi kalama lili. soweli li weka e nimi mi. ni li pali tu.
This thread is cursed... my cat stepped on the keyboard and wiped my response. So this is my 2nd response.

I was working backwards so the first remark was to the last scrap of your piece, saying the standard restricted relative clause had to have some "There is a ..." interpretation and asking why this was necessary and also why you didn't think it was present. 'soweli li moku e kasi lon ma supa. mi lukin e soweli ni'

Not sure anymore. I was just trying to convey that if you put "I see that animal that eats grass" into two sentences, it feels like the two sentence version will layout a story where there is a cow that is eating grass and we're looking at it. Whereas "I see that animal that eats grass" means I'm looking at a cow (which may or may not be eating grass at the moment. Maybe it is watching TV) mi lukin e soweli pu moku e kasi pi ma supa. The cow isn't eating grass, but that action is the salient attribute of a cow.

As for the iterated 'la's, it is not clear whether they are (p>q)>r, as you have them, or P>(q>r) or even (p>r)&(q>r) or something not describable in a linear array.

(p>r)&(q>r) easier to parse, e.g. tenpo sin la mi jo ala lon insa la mi moku e moku. (If it's morning and I'm empty,) I'll eat.
P>(q>r) looks like recursion, and frankly hurts my head to imagine what it means. Maybe that if P is true, then (q>r) is how the world works-- some sort of meta-conditions that must be established before we can decide what the pre-conditions are.

For now, it is enough that in the deep structure, every clause -- and, indeed, every modified noun (and similarly, though slightly differently for verbs, etc,) -- is represented by a fully developed S node (a sub-tree headed by S and ending in ultimate elements).

I don't follow. If I'm reading the wikipedias phrase grammar right, sentences can only be interjections, vocative sentences or sentences with subject predicates. Bare noun phrases can't be sentences, etc. What is an S-node?

Complexity, in the practical sense, is a surface phenomenon. But even as a surface phenomenon, relative clauses are scarcely less complex than connected sentences; they appear to be slightly more so in terms of what is required to get them.

I just did a test translation using "pu" as a sort of "verbal adjective" in the foreign languages section. I ended up using some "pu" phrases and "e ni" phrases and neither seemed to be dominately more useful than the other. "pu" phrases would be most useful for modifying the pre-"li"-phrase, because that is the one that would be the furthest from a subsequent sentence with anaphora.

Why "IE relative clauses" btw -- the IE languages display a variety of styles and the SAE ones (not even this quite accurate) are found in many other families (similarly, I think for "Japanese verbal adjectives" but I am less ure about that). Each language has its own devices for surfacing underlying structures (usually several).

Japanese Style:
Asoko ni suwatte iru wakai josei wa musume desu.
over there LOCATIVE sitting young woman TOPIC daughter is.
That young lady sitting over there is my daughter. The descriptive verbal stuff is placed where an adjective would go. Maybe this is just participles, I'm not sure.

Stúlkan sem starir á hafið
The woman who stares at the sea. The descriptive verbal stuff is in a phrase that starts with "sem", roughly who/what/which/that. An Icelandic adjective would have gone first.

My best guess is that now they correspond pretty much to the contours of the speaker's native language,...

Hmm. Well, if we don't have a con-culture and if two people wanted to talk toki pona to each other, then we'd expect them to lapse into calqes, metaphors from the host culture, and why not the intonation, too? I guess a language designer has to stop designing somewhere and it's anyone's call what users should use to fill in the gaps.