Periods mark the end of declarative sentences and usually imperatives and optatives. They are sometimes used with fragments of the conventional sort (e.g., 'kama pona.')
Exclamation marks are also used sometimes at the end of imperatives and optatives and regularly with interjections and vocatives standing alone.
Question marks are used at the end of questions of all sorts ('seme'. 'x ala x', 'anu', those done with intonation only).
A colon is used between sentences -- or a sentence and an expression -- when the first sentence contains 'ni' referring cataphorically to the following expression either as the event or the message or the words intended by 'ni standing alone or expand the description of of the object modified by 'ni'. This calls for some example:
'mi toki e ni: sina wile kama' "I said that you should come", the message intended.
'mi kama tan ni: ona li wile.' "I came because she wanted me to." event intended
'nimi mi li ni: Kipo' "My name is 'Kipo'. The word intended
'jan ni li awen: ona li kama tawa moku' The man who came to dinner stayed. expanding the description (Note that this can also be done the other way 'round: 'jan li kama tawa moku, jan ni [or just 'ona'] li awen.', without a colon.)
The colon does not have to come immediately after the 'ni': 'mi toki e ni tawa ona: jan suli li kama'
Quotation marks are used around direct quotations, either after 'nimi' within a sentences or after 'ni:' in a separate sentence, 'mi toki e nimi 'jan suli li kama' tawa ona' or 'mi toki e ni tawa ona 'jan suli li kama'' I use (when my typing cooperates) single quotes (raised and inverted commas in some faces) for tp expressions. Others use doubles. And some use one for basic quotes and the other for quotes within quotes.
There is a longstanding suggestion that foreign names in their native spelling be enclosed in "foreign quotes" (double or guillemets), si 'ilo "Skype"', 'jan "Clifford"' and so on (these words are proper names, so need supporting adjectives). This is not done consistently, if at all.
If you want to talk about a linguistic expression (as we seem often to do), the easiest way is to enclose it in quotes (I use single for tp, double for English, etc.). Such expressions, complete with quotes, are proper names of the expression and so, in tp, needs a noun, usually 'nimi. No one does this consistently (Sonja included) with the resulting literal nonsense ('nimi mi li Kipo' being the most common type, even in pu, and 'nimi kepeken li namako' being a particularly blatant case).
The other punctuation marks don't seem to yet have even as loosely fixed a place in tp. Mato uses dashes analytically to mark idioms in his parser, so 'jan-pona' for "friend", but not in regular use. No one has suggested dashes to interruptions in the usual flow of a sentence. But then, no on has suggested interrupting the flow of a sentence. Similarly for parentheses in both senses. And so on.
Except that commas have had a few suggestions and seen some actual use, more or less regularly. And the point of all this is to look at a few possible uses for comas (and semicolons if the need arises).
The chief actual use of commas has been to set off 'la' phrases. While these are already set off pretty clearly by 'la', something more is felt necessary, probably following L1 habits. pu usually does this, occasionally omitting the comma with no pattern I can discern. pu -- and others probably following it -- also put the comma before the 'la', even though the 'la' is an essential part of the phrase (the alternate view, that every tp sentence has an initial 'la' which is dropped when nothing -- except 'taso', 'anu', and 'o' and whatever else we may find -- precedes it, is discredited both by repeated 'la' phrases and the general complexity it introduces).
Given that commas are unnecessarily on either side of 'la' in general, they might still have a role in some cases. The general idea for using commas is to break up certain sorts of ambiguities, mainly in grouping. The most common suggestion is for use between a DO and a final PP, to resolve the 'ona li pana e sitelen tawa mi' ambiguity, leaving "my motion picture" intact, but doing "give the picture to me" as 'ona li pana e sitelen, tawa mi'. This has been often proposed but rarely used.
Other possible places for use are in modifier strings, particularly around 'pi' and its right-hand end. 'pi' means that at least the next two words are to be taken together, but places no outer limit on how many more may be involved, although sometimes more are automatically included because of the 'pi' rule. For example, in 'tomo pi jan pi sona pi unpa pipi', "house of an entomological sexologist" , each successive 'pi' is the second word of the 'pi' before, so it and all after it must be included in the scope of the previous 'pi'. But matters are rarely so clear: does 'tomo pi jan lawa mute' mean "house(s) of many rulers" or "many houses of the ruler(s)"? (In this case, if we think ahead, we may decide to break modifier-order patterns and write 'tomo mute pi jan lawa' for the second case. But this option is not always available, though we can always go back to the basic "use more simple sentences" advice.) A comma. presumably before the 'mute' in the second case, helps a lot, But is not a generally successful strategy, since complexities multiply: 'kulupu pi jan lawa pi ma tomo suli', where the 'suli' can be intended for 'kulupu' and 'jan lawa' as well as 'ma tomo' and a comma does not decide between the first two (but does reduce the possibilities but one and -- it consistently used -- its absence decides definitively for 'ma tomo'). Nor does context help much in such cases, since all may seem plausible, including that the 'suli' quite properly goes in all the possible slots (which may, of course, be part of the point). Further, several of these problems can arise in a single string, suggesting the need to call upon semicolons to deal with some cases, when commas no longer clarify much (maybe even in the case above: getting back to 'kulupu' by 'kulupu pi jan lawa pi ma tomo ; suli'. I have little hope of any of this being used regularly and consistently (and some hope that we will come to find it generally unnecessary).
Going back for a moment to 'la', there are clearly some cases, especially when sentences are involved, where the grouping of 'la' phrases makes a difference. In 'p la q la s', if the grouping is to the right '(p(qr)' this amounts to saying (as we cannot in tp) 'if p and q, then r', but the left grouping (native to tp?) '(pq)r' is the more obscure "If if p then q, then r" or "if p implies q, then r" , which turns out to be useful in math, but less so in life. We don't have an official rule about grouping, but I would suggest the right grouping as standard and then allow the other with a comma. (Whatever the grouping, the order of 'la' phrases is often important 'tenpo pini la ken la' cannot generally be exchanged with 'ken la tenpo pini la': It may have been possible once that I run a mile, but, since I didn't, it is not now possible that I did do it.)
Semicolons are also available for another task, that of connecting separate sentences which are more closely joined that a period suggests. The colon is already use for these in the special cases involving 'ni': indirect and direct discourse and restrictive relative clauses, as we would say for English. But there are other sentence pairs (and more?) that cling together more tightly in our minds, cases where our English sense demands an "and" but tp doesn't offer one. For example, some may happen as the result of a confluence of events: 'utal li open tan ni: jan Papi li nasa; jan Sali li wile unpa' "The fight started because Bobby was drunk and Sally was horny", where a period would leave Bobby's state as the whole explanation and Sally merely an incidental fact. there several idiomatic acases where this cvonnection is also useful, e.g., comparative, 'jan Papi li pona mute; jan Sali li pona lili' "Bobby is better than Sally". There are probably many other such.
All of these remarks are about writing conventions. It is assumed that there are differences in speech that convey the same information (and maybe much more). However, no one has suggested what these differences might be in detail, other than to say that they are probably suprasegmental, matter of pauses, pitch and stress changes, and the like. Exactly what may not be easily answered, since different people from different L1s may have different ideas of what would be appropriate (even the rising final inflection for bare question may be mainly an English habit). For now, the writing is used on its own (and is still the principle form of tp communication) and the spoken version remains to be seen (or, rather, heard).