Pretty Little Girls School (Teachers (Union ....))

James Cooke Brown, the creator of Loglan, confronted the ambiguity of "pretty little girls school" and various extensions to find all the possible readings and to build into his language a way to disambiguate them.  In the right grouping fashion of English and Loglan, he found the following possibilities:
P(L(GS)) a pretty example of a small school for girls.
P((LG)S) a beautiful example of a school for little girls
(PL)(GS) a beautifully small example of a school for girls
(P(LG))S a school for little girls who are beautiful
((PL)G)S a school for girls who are beautifully small
In each case, any binary pair could conceivably be not a subordination relation but a parallelism:
pretty examples of small girls and small schools, for example.

These patterns can be extended indefinitely in the grammar and quite a bit further in even in practice: pretty little girls school teachers union official, say.  Loglan's requirement of complete freedom from syntactic ambiguity naturally required a general scheme which would teat each such pattern differently yet still fall under a general rule.  The resulting system, while not terribly complex, is somewhat difficult to produce and interpret on the fly, but it does work.

Happily, tp does not require complete freedom from ambiguity but merely strives to bring out major differences in structure and ease the disambiguation task to (usually) manageable proportions.  It seems likely that, in the spoken language, further disambiguation comes in changes in voice and perhaps the written language should reflect that, but now we have only the abstract grammatical structures to suggest thaty these added features may occur and need to be recognized.

So, for tp, we might take something like "good little fish bowl", which, since tp groups left rather than right, comes out as 'tomo kala lili pona'.  As in the case of Loglan, the uniform default grouping is unmarked, though the grouping is mirror-imaged. The remaining cases are (with hypothetical commas to mark possible signs of the differences involved)
tomo pi kala lili, pona
tomo kala pi lili pona
tomo pi kala lili pona
tomo pi kala pi lili pona
Hopefully, the pattern is clear enough to extend to cases of good little fish bowl shelves (supo ...) and good little fish bowl shelf brackets (palisa ...) and beyond (but remember that tp style favors many short sentences over any kind of long structure).

When the items within a grouping are conjoined rather than subordinate, the rule seems to be the same: a pair in a modifier position is set off by 'pi'.  New issues arise about conjoined heads: does the 'lili' in 'tomo en kala lili pona' modify both 'kala' and 'tomo' or only 'kala' and the same applies to 'pona' and 'kala pili'?  that is, is this ((t+k)l)p, (t+(kl))p, t+((kl)p)?  And, it follows, 'tomo en kala pi lili pona; might be either t+(k(lp)) or (t+k)(lp).  This suggests that the rules for conjoined terms needs some further study, recommending commas at least and maybe even different markers.  Or, of course, prohibitions against certain kinds of conjunctions.  But this latter seems to prohibitive, since, even with many sentences rather than longer strings, some conjunctions seem inevitable (the red-and-blue ball on the beach, for example).  But at least more study and thought is needed here.

'la' phrases

What all can go in front of 'la' in a sentence and to what end?  This is going to be an ongoing listing, adding new points as they turn up but striving to be, eventually, complete and informative.

1. Conditional sentences, the "if" part of "if,,, then..."
In S1 la S2, the overall claim is that S2 is true at least in a situation where S1 is true, should such a situation arise in the relevant time.  It can also be said as "When S1, then S1" and maybe even "S1 only if S2" and "Whenever S1, S2: and so on, though each of these adds (in English) other information or various implications that are not in the tp form: that S1 is likely to occur, say, or definitely did not occur but might have.  And so on.  Whether or not these added implications can be built into some other structure in tp, they do not occur in the rawest form here,

S1 la S2 is a relatively unsophisticated conditional, and possibly an ambiguous one.  The grammar, in a rare display of recursion,  seems to allow either S1 or  S2 to already include a 'la' phrase, so that both '(S1 la S2) la S3' and 'S1 la (S2 la S3)' appear to be grammatical.  The complexity of interpreting the first, however, makes the second the far more likely reading, so conditions may generally be taken as accumulating from the left (against the usual tp pattern).  This is useful, since tp does not permit sentential use of 'en' (or some other "and" word), though conjoined conditions are quite common and this patterns provides a way,  (The ambiguity may turn up again with non-sentential phrases and may be resolved differently there.)

For the Lojbanists among us, sentential 'la' does not seem to be a mere logical connective, but "modal".  That is, the  mere fact that S1 is not true at any point in the proceedings or that S2 is true, does not make the whole conditional true, though a relevant occasion (and that takes some unpacking) of S1 true and S2 false does falsify the whole.  Some connection, however worked out, is implied by the conditional claim.


2.  Tense. A. Vector

The most common use to 'la' in running text is probably to specify tense, which is not a mandatory category in tp but a felt need for many of us with tensed L1s.  The stock expressions here  are
'tenpo pini'  past (at some time now finished)
'tenpo kama' future (at some time yet to come)
'tenpo ni' present (at this time)
They place the whole event of the sentence after 'la' at a particular (though generally unspecified) time in relation to the present (or, as we see, the time of the events being narrated).  This temporal location may bleed over to other sentences in a narrative sequence, where, without further explicit markers, events in later sentences may be attached to the earlier established time (or a little later to keep up with the flow of events).  In this context, then, one of these markers may indicate a time remote from an already remote point and not explicitly attached to the time of uttering the sentence.  Even 'tenpo ni' in a narrative, may be just "at that time", not the "now" of the speaker, but of the event.  (How to break out of this context to return connection to the speaker's present is not an issue that has been dealt with - 'tenpo ni kin' or 'tenpo ni sin' seem likely candidates.)

'tenpo X la' has another version, a terminal PP 'lon tenpo X'.  The relation between the two is not worked out, but I assume that some differences will emerge, even if only rhetorical ones.  Other prepositions than 'lon' may also have a role (more later).

Whatever else may be the case with PPs here, they pretty clearly have restricted scope relative 'la' expressions. In S1 la S2 lon tenpo X, 'lon tenpo X' applies only to S2 directly (whatever may be the informal implications) .  In 'tenpo X la S1 la S2'  'tenpo X' may apply only to S1, but more likely applies to the whole conditional (we expect to use these initial 'la's to distinguish some of the various kinds of conditionals).  If 'tenpo x' is only for S1, it would be more natural (to a good Lojbanist, certainly) to say 'S1 lon tenpo X la S2.'

Taking these phrases as being, in effect, about some unspecified point in time rather than simply as directions without points at the end, opens a variety of possibilities.  The following sections look at some of these.(The theoretical purely vector tenses might use just the modifiers as pre-la or as prepositional complements.  But at least 'pini' and 'ni' have other uses both with 'la' and prepositions, so the present plan works better.)

B. Spcified times

One immediate response to saying that tense is about some unspecified time, is to ask about specified times.  Why not allow pinpointing a time ?  So we get dates as 'la' phrases.  Or we would if we had a settled system for dates, probably 'tenpo sunpo nanpa tu tu pi tenpo mun nanpa luka tu' or, colloquially, 'tu tu pi luka tu' (4/7) ?  but at least two such notions are well established in tp usage: going to the extremes and moving stepwise through time.

At the extreme we have (Gen 1:1) 'tenpo open la' (or just 'open la;, but this has a rhetorical use as well "to begin with")"At the beginning time"  (and, one expects in these apocalyptic times, 'tenpo pini la' now in a very different sense).  Presumably -- though not in the corpus yet -- other times that are pinned down to events will serve as well: 'tenpo pi utala suli'  "during the big war" and 'tenpo pi utala pini' "during the last war" (clear from context, we hope, though not in abstraction).

Stepwise we have 'tenpo suno pini', "yesterday" and similarly with 'pimeja' for "last night" and 'esun' for "week" and 'mun' for "month" and the same pattern with 'kama' and 'ni'  'sike suno' meaning "year" can also fit into this pattern, without needing 'tenpo'.   Strictly, all theses 'pini's and 'kama's are just any past/future one, but the use for the immediately past/future one is generally the understood convention (how to counter this convention is not clear, as is the reason for wanting to)

C. Quantifiers over moments of time

Another obvious question arising from a form that says "at some time in the past", say, is "What about 'at all times in the past'?"  and "What about getting rid of 'in the past'?".  So, 'tenpo pini' leads immediately to 'tenpo pini ali' (or is it 'tenpo ali pini'?) and just 'tenpo ali', "always".  And that leads naturally to 'tenpo ala pini' (or 'tenpo pini ala') "never".  But, if something can happen at no time and at every time, it can happen at one time or two or (in tp) at many,  And so we get 'tenpo wan pini' (or 'tenpo pini wan') and so on through whatever numbers we have, also with directional restrictions -- just 'tenpo mute la' "often"

At this point we  notice that 'tenpo seme' is also a possible 'la' phrase, here being taken as "How often"?  It could however equally be just "When?", as king for a specification of the sort just discussed or soon to be.  There is no obvious way to specify beforehand what kind of answer is sought (cf.  'sina seme?') , though I suppose modifiers could be used in some way: 'tenpo seme pini' vs. 'tenpo pini seme',  say.
 
The fact that something can occur on several occasions allows another way to specify a time: ordinally: 'tenpo nanpa tu la'  "on the second occasion".

Two side points here.  The same event occurring on several occasions is different generally from a single event that involves repetitive action, though they are clearly related.  The locution we are discussing here is about something that happens on separate occasions, not about repetitive actions.  So "He has attacked him three times" is clearly ' tenpo tu wan la jan mije nanpa wan li utala e jan mije nanpa tu' (to be terribly fussy),  But how do we say that on one occasion he hit him three times: 'tenpo wan la ona li utala e ona ... tu wan' but we do not know what goes in the blank -- nor exactly where the blank is.  It is assumed that 'tenpo' plays a role here, even though time is not particularly significant, compared to the notion of instances (but L1 is a powerful force here and most of ours use a time word here), so 'utala pi tenpo tu wan' is the leading contender.

Secondly, a number of "tenses" which seem to fit with this "always", "often: "once," "never" sequence do not fit naturally here: "usually, regularly, generally, more often than not" and so on.  They will turn up later in this discussion of 'la'.

D. Tensors

Since tenses place an event in relation to the present as being displaced from it forward or backward in time, it is natural to ask "Displaced how much?"  If we have a vector from one point to another, we can ask about its tensor, how long is the arrow?  tp provides days, weeks months and years as units (see above) and a few numerical units.  The move one of these units fore and aft is already covered, so the rest follow naturally 'tenpo suno pini tu' (or is it 'tenpo suno tu pini'?) is "two days ago" and so on in obvious patterns, including 'mute' for "a long time", measured in whatever units.  (It might be that, at least for 'mute', a useful distinction could be made between ' tenpo mute pini' and 'tenpo pini mute', one being "a long time ago"(unit unspecified) and the other "on many occasions in the past".  It is not clear which should be which, though I will -- I think at this moment -- argue for 'tenpo mute pini' for "a long time ago" and then extend that to answer other questions about the various placements of quantifiers in these expressions.  Stay tuned)

E.  Stretches of time.

In addition to the question of how much time has passed since an event, one can ask how long  the event took.  What little usage there is seems to point to a tensor without a vector for this purpose:  ' tenpo suno tu la jan li tawa tomo'  "For two days the person went toward home.'  and especially 'sike suno mute la jan li lon' "The person is old" (so maybe 'jan li lon sike suno mute'?).

3, Continuity

One common use of 'la' phrases is to attach the current sentence to the previous one,  The flip side of 'tan ni:' "because of the following" is 'tan ni la' "because of the preceding, therefore". Although far less common, this naturally suggests 'tawa ni:" and 'tawa ni la', "in order to, for".  

'taso' is another continuity word, but it does not require 'la'.  I suppose that 'anu' can be used in this way as well, again without 'la'

The most common continuity expressions, yet the ones with the fewest uses in tp, are temporal relations: "before, after, since, until, during/while" (others?).  I am inclined to think that "after that" is just 'ni pini la', "that being past" with some device for tensors again (presumably just the ternsor with 'pi' and the connector).  Of course, the case could be made that 'pini ni', "in this's past" means "before this", but I find that confusing (as others do my version).  The same contrast can be made for 'kama',  of course.  Perhaps we have to allow both and let folks sort out their own usage here (to everyone's confusion).  

"Since, until, during" delimit stretches of time again, not by a metric but by reference to another point or stretch,  Further, within that stretch we can manage the array of tense relation "Since I saw him, I have never/often/always/twice thought of him" So, the expression seems to be a first 'la,' after which others are possible but not required,  The temptation here is just to use "before" and "after" again, with the other terminus, now" understood, even though that leaves the stretch indistinguishable from a mere sequence if no subordinate tenses are used: 'ona li kama weka.  ni pini la mi li tawa noka'  "After he left, I walked" or "Since he left, I've been walking"  Perhaps  using modal 'awen' in the second case would do it,  Or putting 'awen' in the 'la' phrase: 'ni pini awen'  Or something else.  Such as insisting on secondary tenses to avoid ambiguity.  

But 'awen' does seem to fit naturally into "during, while." If "after this" is "this being done" (or "in this's future"), then "during this" seems to be "this going on" (or "in this's going on", the two seem to collapse), so "meanwhile: is 'ni awen' (or 'awen ni'),  This fits in pretty much with the modal uses of the three verbs for aspects, insofar as tp uses these at all ('pini' does both terminative and perfect, 'kama' inchoative, 'awen' progressive and superfective.  'open', with no role here, does initiative.) 

4.  Sources

This covers a variety of cases.

We can talk about where something is true as well as when.  Thus, we can use place names in 'la' phrases to discuss customs, habits, and the like of a particular place or even a group: 'ma Mewika la, sike suno tu tu ali la, jan ali li wile e jan lawa ma.' [Just noticed another temporal expression not discussed above.]   Like similar temporal expressions, these have a corresponding PP with 'lon'.

We can talk about where we get our information: "according to Wikipedia, " 'ilo Wikipesija la'.  Or, whether we take it as information or not, the source of a claim "In John's opinion," 'jan San la' ('mi la' is probably the one most usually accurate, but it is assumed throughout tp.)  All of these have the PP version with 'tawa' "according to" (though 'tan' would seem to make more sense).

The usual reasons for mentioning sources, aside from accuracy and avoiding responsibility, is to give some indication of what reliance the hearer should put on the claim (and maybe convey some indication of the speaker's reliance),  Another way to do this is to indicate how you came by the claim.  Of course, citing a source says you came by it second hand, but you can do that without citing a source: "I hear that...",  "science tells us" ('nasin la' or 'sona la'?), even "it follows that" (though this may overlap with 'tan ni').  So 'kute mi' and 'pilin mi' (stronger than just 'mi') and 'lukin mi' (eye witness) and the  like may be used (though rarely are -- L1 again).
:
:

5. Reaction (emotion, verification)

Aside from our confidence in a claim and its source, we may comment on a claim in terms of how it affects us or how we expect it to affect our hearers.  The most common o these are 'pona la' "Fortunately" and 'ike la' the opposite.  We also have 'wile la' "Hopefully" .  More generally, emotions show up as separate interjections, just 'a!' (for just about any\thing) or adjective interjected (with or without an 'a') to give a more specific response.
The most interesting reactions are actually ot attached but responses to someone else's claim 'lon' and 'ala'  "That right" and "Lies". They can be used in reporting claims, however.

6. Organization

When what one says is organized, one tries to project that organization to one's listeners and so gives guidelines:
open la.  "to begin with"
(nanpa) wan la "First"
wan la "On the one hand"
ante la "on the other hand"  But also "On the contrary" and "Otherwise" (after a list of objections or compliments)
kin la "Moreover, in addition"
pini la "in conclusion"
poka la  "by the way, digressing a bit"  (probably 'awen la' "to get back on track"
ali la  "in summary"
lawa la  "the main point, most importantly"
There ought to be things for "for example" and "expanding on that" but I don't know them

7. Topic?

Some folk say that the 'la' phrase can be used to front the topic, the focus of a point when it might be buried syntactically because of the fixed word order (or just for emphasis even if it normally came first, or just to satisfy some L1 pattern): 'kama kon ante la jan nasin li sona e lon ni'  "Climate change, scientists know it exists."

8.  Modals

The common use of modals in 'la' phrases is 'ken la'. "Possibly, maybe" ,  One would expect then to find 'wile la' "Necessarily, surely", but it hasn't occurred (a few logician show pieces aside).  The closest to it is 'lon la' , but that seem to be "In fact, actually" to contrast with speculation and errors.
The other technical modals (lukin, alasa, sona) don't have the lawlike character that this construction seems to require and 'kama' is wrapped up in tense.
The notions like "regularly, in general, usually, habitually" and the like need expression and words like 'nasin' and 'lawa'  and even 'awen' suggest themselves, but there is no usage.  'nasa,' of course, serves as the opposite point for at least some of this.  'kama la' alone might have the force the force of something occurs without regularity or even against it.

Double objects

A theoretical possibility had occurred to me and then an actual case arose.  The theory is this: 'ken' takes whole VPs as complements, that is verb plus object plus prepositional phrases: "He can hunt birds in the forest": 'ona li ken (alasa e waso lon ma pi kasi suli)'.  'ken' is also a transitive verb meaning "enable" with DO for what is enabled and a complement for what it is enabled to do "Bill allowed Joe to hunt on his land" 'jan Pili li ken alasa lon ma sama e jan So' (or is it 'li ken e jan So alasa lon ma sama'?  It certainly is not 'li ken alasa e jan So lon ma sama', which clearly means that Bill can hunt Joe).  Clearly something has happened to the structure here.  In this case, the PP actually helps matters, since, if we drop it we get 'li ken alasa e jan So' immediately, which tends toward the non-transitive 'ken' reading, though keeps the grammar clear -- for both readings.  On the other hand, 'li ken e jan So alasa'  is clear but more difficult to fit into the grammar, where VP ends with the DO in the absence of a PP.

Now, suppose what is allowed involves an object as well, bird hunting again, say.  We end with
jan Pili li ken alasa e waso e jan So (so Joe is still a possible target on one natural reading) or
jan Pili li ken e jan So alasa e waso.

Once actual cases are laid out, the conclusion seems clear: the second option is to be taken.  Thus, the slot "Modal" becomes more complex than it seemed; it is not just the modal word plus possibly modifiers of manner and negation but also a possible direct object.  After all that the VP comes in.  I think that some examples of this use of 'ken' (and I suppose 'wile' and maybe others) have dealt with the situation otherwise, but I can't find any cases to check.

And then, I note that, if the DO of the modal is not a name, we get another structural ambiguity ' ona li ken e waso laso pona e tomo'  "He allows the bird to paint his nest a nice blue"  or "he allows the blue bird to fix his nest".  So, no solution really works and we have to rely on context (as usual).

'sama' as pronoun

 'sama' is listed as both the reflexive and the reciprocal pronoun for toki pona, for "-self" and "each other".  That is, at some deeper level, the 'sama' as object arises from a more complex structure involving the subject as well.  The simplest cases are:

x li V e x  => x li V e sama

x li V e y, y li V e x => x en y li V e sama
(Some verbs are inherently reciprocal, e.g., 'wan', maybe 'unpa', so only one of the pair need be given.)

Similar rules will work for the case of prepositional objects, including the complements of prepositions as verbs and for modifiers.  But the details seem (to me, now) to get a bit messy, so I'll skip over them.

The case for modifiers and prepositions seems to be just what one expects, the same as above with "M (pi)x"  replacing "V e x" etc. through out, where M might be very complex, involving a verb, perhaps, and objects and other prepositions.  So the whole can be summed up as
x G x => x G sama
xG y, y G x => x en y G sama
There are surely some restrictions on these but just what are not yet clear.

Some complications can arise, ambiguities in fact.
given that is is (almost) always possible to drop the object of transitive verbs, with marginal loss of meaning,
x li V e y, y li V e x => x en y li V e sama => x en y li V

x li V e w => x li V  } => x en y li V [taso e sama ala]

y li V e z => y li V

 x li V e w, y li V e w => x en y li V e sama
(This is a more semantic or pragmatic rule, so probably has a lot of conditions on it.)

Some fine points

This is not strictly about words, but about some constructions which have been discussed on various forums lately.  It is not perfectly clear (to me, at least) whether these are changes from earlier situations, e.g., Pije's lessons, or clarifications or extrapolations from them.  Nor is it clear that these are the final answers, but they are the standard for now (I think).

1.  No 'pi' with 'nanpa' followed by a number.  This does go against at least one example in Pije, but agrees with several early texts.  The pattern is to bring ordinal numbers into line with cardinals: we do not use 'pi' with numbers which take more than one word: 'tu wan', 'luka luka' and the like.  (Note, by the way, that 'luka' as "five," while regularly deprecated, is universally used.).  Thus, just as "three men" is 'jan tu wan' and not 'jan pi tu wan', the third man is designated 'jan nanpa tu wan'.  That is, 'nanpa' initiates a right grouped number.

2.  Still on 'pi', prepositional phrases that are meant to modify nouns (are adjectival) require 'pi' before them. This is just a clarification of a rule in Pije that right grouped strings of two or more terms require 'pi' (though there are some early examples of violations in both directions even without prepositions). This is particularly important toward the end of the sentence, where the adjective PP might be confused with the terminal adverbial PP, the classic "I saw the man with the telescope"  (mi lukin e jan (pi) kepeken ilo lukin).
But the rule applies in all positions: "The man with the telescope saw me too" (jan pi kepeken ilo lukin li lukin e mi kin).

3.  Contrary to what Pije says, but in keeping with some examples, 'en' can occur in verb and DO position.  But only for the mixed-lot case, not for the separable ("logical") "and".  'ni li loje en laso' is for something that is partly red and partly blue, but if there are more than one thing and some are red (totallty. for practical purposes, and others blue (ditto), the form is 'ni li loje li laso' and this is equivalent, as the first is not, to two separate sentences.  Similarly, 'mi lukin e soweli en waso' is legal to describe a chimera of some sort (half beast, half fowl), but not two separate animals. each totally of its proper kind (mi lukin e soweli e waso').

4.  Tying these together, compound modifiers -- which do take 'en' everywhere, whether mixed or logical -- require 'pi'.  Thus, 'mi lukin e waso pi pimeja en walo'; again, a result of the general rule about modifiers of more than one word.  This would apply equally to a particolored bird (or birds) and to several birds of one color each, though some different from others.

5.  The issue of proper of various sorts involving 'toki' has come back into play, after lying dormant for several years with a satisfactory set of solutions -- though ones not natural to English (and, apparently at least Spanish, German, and French) speakers.  The basic principle used earlier was that the DO of a verb had to be something that would also be referred to by that verb used as a noun (the moku e moku principle).  Thus, the DO of 'toki' had to have a message ('toki' again) as its referent.  So, the DO could not be a language or a topic talked about.
a. But the first of these, for which 'kepeken toki ...' was devised, came under attack, because a language is, of course, also a toki, so the standard argument falls through.  Yet the 'toki e toki pona' expressions always feels like a mere carry over of L1 habits and does not have a separate justification within tp.  One can, of course, be found easily by taking the expression as short for (a antidittophatic collapse of)  'toke e toki pi toki pona', which collapse seems likely to have occurred, given the opportunity.  The compromise, which avoids the supposed unnatural 'kepeken' construction and also puts the the language in a modifier role, is to make the verb 'toki pi toki pona', for which there are some early examples.  On the whole, my preference remains with 'kepeken' and the verb modifier as a legitimate alternative (but see below) and shun the DO form.  But the question is again open.
b. The "about" question is not so clearly opened as that of the language used.  It is opened, if at all, by the apparent expansion of the DO of 'toki', even though the cases are not really parallel -- the topic is not a toki, even if the general term for "topic" were 'toki' (which I am not sure it is; it seems more like 'sona' to me -- or 'pilin' or 'nasin').  On the other hand 'e ijo [(pi) topic]' does seem artificial, though literally accurate (it seems a Lojbanic solution, if you will).  The alternatives proposed  to the 'ijo' construction are to use the topic as a modifier to 'toki' or to invent a preposition with the meaning "about".  The first of these introduces another ambiguity: both with the above "in language" construction (for languages might be topics of discussion, as tp often is) and with the usual adverbs of manner: is 'toki wawa' shouting or talking quietly about force?  On the other hand, no preposition leaps to mind as the right one for "about." 'lon' has some history in other languages: "on", for example, but that is clearly just L1 relexing and the same applies to creating new prepositions like 'sike'  (though the process of making a word a preposition is not a problem per se).  On the whole, this issue seems the least in need of revision, but the possibility has arisen,

Some generalities in derivational semantics

Although toki pona has no syntactic parts of speech in the usual way, the vocabulary here is divided into various traditional classes: nouns, (transitive) verbs, modifiers (adjective and adverbs), prepositions and modals.  Except for prepositions and modals, which have some syntactic peculiarities (taking complements in all positions), the motivation for these categories is primarily semantic.  If using a word in the slots suggested by its category is taken to give the basic meaning of the word, then its meaning in other slots can generally be inferred more or less accurately.

Verbs:  The generic type of the DO of a verb has the same name as the verb (or a verb used as a noun is likely the sort of thing that gets verbed by this verb).  The paradigm is moku, v "to eat", n "food", but most other verbs will do as well.  Of course, as a noun, a verb may also just stand for the activity of the verb, "eating" in the paradigm case, a probably other things as the need arises.  As a modifier, the characteristic meaning of a verb x is "suitable for xing", "edible" in the case of moku.   Of course, more complicated notions, derivative from deep sentences, typically, may also apply, the most common being "given to/liable to x" "gluttonous" in the case of moku.  The first guess about a verb without a DO is simply that the DO is omitted for rhetorical reasons.

Nouns.  Using a noun.n as a verb takes on one of two meanings,  1 "cause DO to become an n" or 2 "apply n to DO".  Thus x li telo e y means either "x melts y" or "x waters/washes y".  Clearly, the specifics of these are context dependent: we don't ordinary wash plants for example, nor water a room.  And similar remarks apply to "cause to become water(liquid)", which might be condensation, for example, rather than melting. Not surprisingly, ilo seem to have only the second meaning here, the first being taken by (surprise!) kepeken.  So,  x ilo e y means "x uses tools on y", the specifics coming from context.  As a modifier, a noun n stands for the characteristic of this thing n stands for, though the details may be filled by either context or convention.  Modifier jan means either "human" or "humane" and then others that follow.  By convention, akesi means "ugly" and pipi means "tiny".  Soweli is supposed to mean "cute" but is rarely so used and the other conventions are a not uniformly recognized.

Prepositions.  The object of the preposition (complement) functions like the DO of verbs derivationally, that is their genera are called by the preposition.  The clear case is tan "from", "source".  But the rule applies equally to lon "at", "place, address" and tawa "to", "goal, purpose".  As verbs, prepositions take the causative sense like nouns: x li lon e y is "x places y".  Of course, without the object of the preposition, the meanings of some of these causings is open to either context or convention (placing y might mean either bringing y into existence or to life or ...).  Similar liberty applies to prepositions as modifiers, taking on all the possibilities from their use as nouns and verbs (tawa is particularly productive in this area, getting into motion generally).

Modifiers.  As nouns these fluctuate between the abstraction of their property (most common) and concrete instances (usually contextually suggested).  loje li kule, "red is a color," on the one hand,  and  loje pi ma Pomelan li unpa, "di royte Pomerantsin trent (or yentst, depending on how you were brought up) " on the other.  As verbs, they take the causative sense again, always with the result  DO li kama m, whatever the modifier m was,

Modals are a kind of verb which take verb phrases as complements (much a prepositions take noun phrases).  Some also function as verbs, e.g. wile, "want", probably as special reductions of their modal function (? wile e DO = wile jo e DO ?).  They can be made transitive with a causative sense, but there are problems about where to put the two DOs, one from the causative verb and one from the complementary verb phrase. As nouns, they tend to be the the abstraction of the modality, e.g., ken "possibility, permission", and as modifiers the corresponding property, "possible, permitted"
s
These are the basics; what happens after these is what makes each word unique and special (and harder to get a grip on).

Toward some transformation rules.

Given, for the moment that the basic tp sentence has the form (ignoring conditions)
Subj li Verb (e DO) (PP)
we can reach a few familiar sentences by known transformations:
{o,mi,sina} li A => {o,mi,sina} A  (obligatory)
A li B, A li C => A li B li C
A li B e C, A li B e D => A li B e C e D

(this one is ambiguous, since it might be taken to mean that, if C is a DO followed by a PP, and D is another DO followed by another PP, both of the PP would be copied as well,  We may want this or we may not; at present it seems we do.)
What about PPs?  A li B PP1 A li B PP2 => ?  A li B PP1 PP2?  A li B PP1 en PP2? A li B PP1 li PP2?
A li B, C li B => A en C li B (and similar rules for NP in PP, including complements to Prep verbs)
 A li Mod e ni: A li B => A li Mod B  (???  because of the -- to us -- ambiguity of the modals, this often doesn't seem to work out right and the A li Mod B may just be a construction rule)
Name o o Sent/VP => Name o Sent/VP
A li Prep NP e DO => A li Prep e DO Prep NP  (but 'lon' seems to go over to 'tawa'?)  This may be reversible.


on the noun/verb level
AB, AC => A pi B en C