• Koshon, or in Lojban, .kocon (pronounced as KO-shon).
• Sora’s childhood friend.
• Speaks Lojban quite well.
• He'll be serving as Sora’s teacher throughout this course.

• Sora, or in Lojban, .soran. (pronounced as SOH-rahn).
• A total beginner at Lojban.
• Learning the ropes from Koshon.

• Sevan, or Sevan (pronounced as SEH-vahn).
• Sora’s younger sister.
• Surprisingly has the most Lojban experience in the family.
• A fountain of random Lojban trivia.

Hi everyone, coi rodo! I’m Koshon. Great to meet you all.

Yahoo! I’m Sora! …Wait, Koshon, did you say “shoi rodo”?

That’s a greeting in Lojban, the constructed language. It means “hello, everyone.”

Cool! When I hear “Lojban” I think of logic.

Sharp eye! Lojban belongs to the loglang (logical language) family. Thanks to clever design, the grammar is remarkably simple and it's quite easy to pick up. Want to give it a try, Sora?

If you say it’s easy… okay, I’ll give it a shot!

Something like that, I guess.

And we’ll keep going from there.

First things first: if you can’t read it, you can’t use it! So let’s start with pronunciation and the alphabet.

Lojban is typically written in ASCII, which fits the internet age perfectly. While other scripts exist, ASCII is what most people use. Below are the letters we'll be using (H, Q, and W are not used). Here are approximate English hints — use the play buttons to hear each letter.

Got it. So it's basically the standard Roman-letter values, but without H, Q, or W.

The tricky ones seem to be c (sh), x (that rasping sound), and y (the 'uh' sound), right? What exactly is “x”?

x is a voiceless velar fricative — between “k” and “h”. It’s like ch in Scottish loch, or ch in German Bach, or J in Spanish Jose, or Kh in Modern Arabic Khaled. Try pronouncing ksss while keeping your tongue down and you get this sound. Compare c, k, and x:

• c▶
• k▶
• x▶

(Uhh, velar-what now…?) I’ll replay those a few times before bed. And y is…?

That’s the neutral vowel — English schwa: like the vowel in comma, not the y in misty or cycle.

• y▶

Relax your mouth and voice; you’ll get something like it.

Got it. How do you name the letters? Ay, bee, cee?

Consonants: [consonant] + y + . · Vowels: . + [vowel] + bu

The . represents silence—a brief pause or a glottal stop. Keep in mind it is not a sentence-ending period like in English (we’ll cover those later). Dots are used in letter names to ensure words don’t accidentally blend together.

Also, a crucial rule: the apostrophe only ever appears between two vowels. You'll never see sequences like k'a or d'e. While the apostrophe sounds like a soft h, Lojban treats it as a special glide, not as a standard consonant like b or m.

Makes sense. ts and tc are like 'tz' and 'ch', and dz and dj are just their voiced counterparts. Simple enough!

The chart covers most of it, but you should also know how vowel clusters work. The lerfu audio pack has no separate ia or ua files — only the trigraphs aia and aua, which are exactly i and u as semivowels between vowels:

In these clusters, i and u act as semivowels.

So ies is like English “yes”.

Exactly. Then we have diphthongs:

These are always treated as a single syllable.

What about ao or ea?

Lojban only recognizes those four as diphthongs. If you want to connect a, e, or o directly, you have to separate them with i, u, or an ' (apostrophe).

Fair enough.

Quiz time! Try reading these sentences aloud.

Okay, let's give it a go...

1. shoy... ro... do... mi-heh... so-rahn... sho-ho... ko-shon (something like that?)
2. mi pra-mi do... ee-seh-moo-hee-bo... doon-dah lo mel-bi khru-lah
3. oh-hee... moo... khahg-zhee... so-fuh-bahk-nee... shoo zvah-tee leh poor-dee

How was my accent?

“Soviet cows” — what even is that?

The third one is a famous pangram: it uses every Lojban letter.

Whoa, you’re right!

It's a fun challenge! I even made my own—it includes diphthongs, semivowels, and those tricky affricates (ts, tc, dz, dj). Check this out:

That’s long.

Wait, we never actually talked about stress. I was just guessing that it falls on the second-to-last syllable.

Oh? And what led you to that conclusion?

I told you, it was a guess!

Sure it wasn't a "cheat sheet" guess?

I swear, no cheat sheets! Forgive me! 🙏🥺

Anyway, Sora is right. In Lojban, you generally stress the second-to-last syllable of every word.

(Wow, just ignored me...) See? A woman's intuition is never wrong.

In practice stress rules differ a bit by part of speech — so let’s cover parts of speech while we’re at it.

In Lojban, "part of speech" is tied directly to morphology (the shape of the word). You can usually tell what kind of word it is just by looking at the spelling:

• brivla ("content word" or predicate): Ignoring ' and y, these have a consonant cluster in the first five letters and always end in a vowel.
• cmavo ("structure word" or particle): These have no consonant clusters and no internal consonants (unless it's at the very beginning), and always end in a vowel.
• cmevla (name word): The name-shaped tokens that follow la—they always end in a consonant, and in modern Lojban we usually wrap them in dots (e.g., .soran.). This is a word shape, not the same thing as the predicate cmene (“x₁ is the name of x₂”), which is a separate content word you’ll see in phrases like zo .soran. cmene mi (Lesson 7).

Just remember that the apostrophe (the h sound) doesn't count as a "consonant" for these rules.

Here’s a quick decision tree to keep it straight:

1. Does the word end in a consonant? → treat it as cmevla (a name word; in modern usage usually wrapped in dots).
2. Else (it ends in a vowel): ignoring ' and y, is there a consonant cluster in the first five letters? → brivla (predicate word).
3. Else → cmavo (structure word): no internal consonants, so cmavo can be run together without spaces.

Exactly. The big questions are: “Does it end in a consonant?” and “Is there a cluster in the first five letters?”

One cool thing: because cmavo never have internal clusters, you can actually run them together without spaces. For example, pu zi ze'u ri'a vi ve'a can be written simply as puzize'uri'avive'a.

That saves some space—though you have to be careful not to mistake a long chain of cmavo for a single long word!

Let's revisit stress for a moment. It works like this:

• brivla and cmevla: Stress the second-to-last syllable (y is ignored and does not count as a syllable here).
• cmavo: Stress is free. However, if a cmavo comes right before a brivla, it's usually not stressed.

Stress freedom! Cmavo rules sound a bit fiddly, but "don't stress right before the predicate" seems like a solid default.

True. For cmevla, there's also an exception: you can manually mark stress using CAPITAL LETTERS if you want it somewhere other than the second-to-last syllable. But for now, just sticking to the “second-to-last” rule is plenty.

Let's keep it simple for now! We can worry about the exceptions later.

First the sounds, then the word types... Alright, Koshon, lay that Lojban grammar on me! I'm ready for the "Super Grammar"!

Let's start with basic statements. Think of a sentence as saying:

Who or what / does what / is like what / to whom / where...

Basically, you have "the thing we're talking about" and "what we're saying about it."

Makes sense. Like "I eat this"—I and this are the "things," and eat is the "doing."

Exactly. Borrowing terms from logic, we call the "doing/being" part the predicate (selbri), and the terms that fill its roles arguments. In Lojban those argument terms are sumti; the numbered roles x₁, x₂, … are terbricmi, and their ordered sequence is the terbri (place structure). A full claim is a bridi.

Let me see... A bridi is built from a selbri and the list of terbricmi (slots) that are fulled with sumti. The selbri is the is/does part, and the sumti are the who/what terms going into those slots. Right?

Spot on.

Okay, next.

Every Lojban predicate has a terbri—an ordered list of terbricmi (place slots) labeled x₁, x₂, x₃, ...:

• x₁ is / does ...
• x₁ does x₂ ...
• x₁ is x₂-ish to x₃ ...
• and so on.

Think of x₁, x₂, etc. as numbered "holes" (terbricmi) that you plug sumti into.

To make a sentence, just put the sumti in their numbered order and tuck the selbri in the middle. Simple, right?

Like:

(sumti for x₁) [selbri] (sumti for x₂) (sumti for x₃) …

Wait, where exactly does the selbri go?

Strictly speaking, the selbri can go almost anywhere—but most people put it between the sumti for x₁ and the sumti for x₂, much like English SVO (Subject-Verb-Object).

That feels pretty natural.

Right — a familiar default. Here are some sumti you can use already:

mi

I / me

do

you

zo'e

something unspecified / “whatever”

ti

this

ta

that (near you)

tu

that (yonder)

ra

he / she / it / they (in context)

Some predicates:

ninmu

x₁ is a woman

citka

x₁ eats x₂

zunle

x₁ is to the left of x₂ from viewpoint x₃

sanli

x₁ stands on x₂ using legs/support x₃

nelci

x₁ likes x₂

So, line up the sumti, tuck in the selbri...

1. mi ninmu
2. do citka ti
3. ra zunle tu mi
4. ra sanli tu zo'e
5. mi nelci do

Perfect! We're taking it slow with the vocabulary, but you've already mastered the basic sentence structure.

Honestly, this might even be simpler than English for things like "they're to my left of that."

Ready to put it into practice? Let's take a little field trip—there's a town I've been wanting to show you...

Ooh, an outing! Lead the way!

Welcome to Latcmatcad—the "Cat Town"! In Lojban, we write it as la .latcmatcad. (where la marks it's a name, and the dots are there to separate it from surrounding words).

Ooh! Cats! ...Wait, I don't see any...

That's strange. Maybe they're in a meeting? While we wait for them to show up, let's talk about .i and zo'e.

When you're connecting multiple sentences, you use .i to mark the start of each new sentence. It's kind of like a reversed period (Lojban punctuation often feels a bit "backwards" to English speakers!).

We didn't use .i much in the previous lessons, did we?

If you're only saying one sentence, you don't strictly need it. But when you're speaking in a flow, it's the standard way to separate your thoughts.

Got it: use .i between sentences. ...And we've seen zo'e before too, right?

Exactly. Here's a very handy rule for keeping things concise:

You can almost always drop zo'e if it's at the very end of a sentence, or if it's in the x₁ slot (right before the selbri).

In other words: if a slot for "someone/something" is empty, Lojbanists assume you mean zo'e.

Oh — cats! Look!

Whoa, you’re right — cats!

Say it in Lojban:

Basically, leading zo'e and any zo'e hanging off the end of a sentence are optional. So zo'e mlatu zo'e can be shortened to just mlatu.

Here's the pattern:

• zo'e [selbri] [sumti] … zo'e zo'e … = [selbri] [sumti] …

But be careful! Internal zo'e slots that aren't at the end cannot be dropped, or the other sumti will shift into the wrong places:

• [selbri] [sumti] zo'e [sumti] ≠ [selbri] [sumti] [sumti]

mlatu .i mlatu .i mlatu .i mlatu...

Quite the crowd! Let’s take a walk and practice putting sentences together. I’ll give you the words you need, and you just arrange them. Grab a pen and paper—if you're not sure which sumti to use for a given terbricmi, just use zo'e (and feel free to drop it if it's at the end!).

Phew, I'm beat... a hot Earl Grey is exactly what I need right now.

Nice work! You're really getting the hang of this. Check your answers below (remember, the bracketed parts are optional):

So: pick the selbri, look up its terbri, and put a sumti in each terbricmi. That "fill-in-the-blanks" template is what you called the Place Structure, right?

Exactly! Place Structure (PS) is the usual English name for a selbri's terbri—the ordered terbricmi that define the relationship.

PS—got it. ...All this talk about tea and coffee is making me crave a crêpe!

Mmm, PS and crêpes... a perfect combination!

Wait, how are those related...?

Crêpes are amazing! And wow, I still can't believe how many cats there were!

Yeah... (There goes my diet...)

I'm all powered up now. Basic sentences are starting to feel like second nature!

Great. Once you're comfortable with simple sentences (bridi), we can start fleshing out the words themselves for more detailed expression. First, let's look at how to expand our selbri.

In Lojban, you can chain multiple brivla (predicate words) together. This kind of chain is called a tanru.

So the words on the left modify the ones on the right?

Precisely. And here's the cool part: tanru grouping is always unambiguous. Lojban has strict rules for how these chains are interpreted.

So there's no "attachment ambiguity"?

Exactly. The rules are very logical. Of course, the downside is that sometimes the grammar might force a grouping you didn't intend—so you have to pay attention.

In natural languages, a phrase like "red-headed fish-eating cat" can be interpreted in several ways. But Lojban forces a single, specific structure.

That's what we mean by Lojban's syntactic precision. Tanru are a perfect example of that.

Lojban doesn't mess around, does it?

While the structure is precise, the meaning of a tanru is actually quite flexible. It usually just means "a Y that has something to do with X" or "an X-ish Y."

Humans fill in the usual reading.

In ordinary situations, trust your intuition.

A tanru functions as a single selbri. Its place structure is determined by the rightmost word (the "anchor" or "head" of the chain).

Hrrrgh...

Good morning! Why the long face?

I was just going over my notes, and I realized I only know a handful of sumti—and most of them are just pronouns like "me" or "you."

That's a fair point. Today, let’s learn how to turn any word into a sumti.

Exactly. How do I talk about a "woman" or a "cat" as a thing? Is there a word for "the" or "a"?

Remember our predicates ninmu and mlatu?

"Woman" and "cat". Yep, got it.

Those are brivla. If you wrap a brivla (or even a tanru) in lo ... ku, it turns into a sumti meaning "the thing(s) that fit the x₁ slot of that predicate."

So lo is like an article? And what's ku for?

ku is a terminator—it tells the listener exactly where the description ends. Lojban is very explicit about its structure.

So lo ... ku acts like brackets around a description.

lo ninmu ku nelci lo mlatu ku The woman likes the cat. / Some woman likes some cat.

Precisely. lo [selbri] ku basically means "some thing that is (or does) [selbri]."

So with words we know:

• lo blabi ku — something white
• lo citka ku — an eater
• lo sanli ku — someone standing
• lo cadzu ku — someone walking
• lo klama ku — someone going
• lo sipna ku — a sleeper
• lo gleki ku — someone happy
• lo tatpi ku — someone tired

Just think of lo as "pulling out" that first place slot (x₁). And this works for tanru too:

la .soran., la .latcmatcad. — is la also an article?

Yes. la before cmevla (name words—the dotted, consonant-final pieces like .soran.) means “the thing called …”

No ku after la?

There are complications…

Never mind then!

Each cmevla (name word) must end in a consonant. If your name ends in a vowel, you just add one—usually s or n.

So my name becomes .soran. Self-intro… la .soran. cmene mi?

Not quite! cmene here is the predicate “x₁ is the name of x₂”—not the same idea as a cmevla (the name-shaped word .soran.). The x₁ of cmene has to be the name text (often marked with zo), not “whoever la points at.” la .soran. picks out the person; saying what you’re called uses patterns like zo .soran. cmene mi.

So how should I introduce myself?

To say "My name is Sora", you can use zo (which we'll cover later) or me (which we'll see in Lesson 15):

Simplest is probably mi'e.

Wait, doesn't Lojban distinguish between "the cat" and "a cat"?

Not in the way English does. lo is our workhorse—it's very general and covers both cases.

"When in doubt, use lo." Got it.

There's also le, which we use for "the specific one(s) I'm thinking of." It's still referring to the x₁ slot, but the speaker is singling it out.

Like saying "that one thing I was talking about..."

Roughly. Practically speaking, if you can use le, you can almost always use lo instead. le is just a more specific tool in your kit.

When in doubt, lo—I think I can remember that!

Next, let's look at how we can manipulate our selbri—which, in practice, opens up a whole new set of tricks for our sumti.

Manipulation, huh? Sounds like we're getting into the advanced stuff!

It’s actually very intuitive. The SE series of words can be attached to a selbri to swap its x₁ slot with one of the others (x₂, x₃, x₄, or x₅).

"Swap"? How does that work in practice?

Example with klama:

I see! So se swaps places 1 and 2, te swaps 1 and 3, and so on.

Exactly. These conversion words don't change the underlying meaning—they just change the perspective of the sentence.

Is it really worth the extra words?

Definitely! For one thing, if you only care about a recipient or a tool and don't care about the giver or the user, it saves you from having to use a bunch of zo'e:

Hmm, I guess that is cleaner.

But the real magic happens when you combine it with lo! This lets you turn any place slot into a description or title:

Whoa, that's a huge power-up for our descriptions!

Oh, I get it now! That's what you meant by "widening sumti tricks."

Bingo. Whenever you're stuck, just remember: with se/te/ve/xe, you can use lo to extract any slot from x₂ to x₅.

We've covered a lot of ground—simple sentences, tanru, articles, conversion words... I think I'm starting to get the hang of this!

Good! Today we're looking at the FA series. This is mostly a trick for readability and making sentences more compact.

The "FA series"? What's that about?

Example:

Right, we saw that in the previous lesson.

It feels a bit clunky with all those zo'e's, doesn't it?

I guess. But without them, mi klama ti would just mean "I go to this place" (x₂), rather than "using this tool" (x₅).

Exactly! FA words let you assign a following sumti to a specific terbricmi (x₁, x₂, …), no matter where it appears in the sentence:

Oh, so we can just skip all the middle "zo'e" slots by tagging the one we actually want!

Just one thing to keep in mind: Conversion (SE) is always applied before the FA tags. So if you use both, the FA tags count according to the new converted terbri.

(1) I give (something) to you. (2) The man speaks in Lojban. fo puts la .lojban. into the x₄ slot!

Right! And (3) is the real test. Here's the key rule: after you tag a sumti with a FA word, any following untagged sumti just follow in order from that point. They "fall in line."

Okay, let's see. The selbri is ciska (writes). The first sumti is "the child" (x₁). Then we see fi, which puts the yellow paper in x₃. So the next word, pinsi (pencil), must fill x₄!

So: The child writes on yellow paper with a pencil.

Spot on! You just combined articles, tanru, and FA tags all in one go.

I'm on fire!

Naturally!

We're almost halfway through! This is a great time to learn about abstractions—what we call "noun clauses" or "that…" clauses in English.

I see the pattern:

• lo nu [sentence] kei ku — the event where [sentence] happens.
• lo du'u [sentence] kei ku — the proposition that [sentence] is true.

Exactly. While English often just uses "that" for both, Lojban distinguishes between an event and a proposition (along with several others).

Wait, what exactly is a "proposition"?

Think of it as the "content" of an idea—the kind of thing you believe, doubt, or claim as true.

A bit fuzzy, but I think I get the gist.

Lojban dictionaries will tell you whether a word normally takes nu or du'u in its slots. You'll develop a feel for it as we go.

The general pattern is: lo [Abstractor] [sentence] kei ku

The word in the middle is called an abstractor (or a NU-series cmavo). We'll start with just nu and du'u.

There are several others like ka, ni, si'o, and so on, but let's stick to these two for now.

It still feels a little... well, abstract.

Let's have some practice!

Roughly:

1. I'm sad that the woman died.
2. Eating that apple pleases the bird.
3. The man is certain (of the proposition) that the apple is tasty.
4. Studying Lojban is interesting to me.
5. I think you are beautiful.

Excellent! You can think of them as building noun clauses with a clear closing tag (kei).

kei ku feels like a lot of typing... So nu [sentence] kei is basically a selbri by itself?

Technically, yes—abstractors take a full sentence and build a new predicate out of it. But in practice, you'll almost always see them wrapped in lo ... ku to turn them into arguments (sumti).

So nu means "...is an event" and du'u means "...is a proposition." If you say I don't need to overthink it, I won't!

I've been on a bit of a decluttering kick lately.

Oh? Do tell.

Tossing out old junk just feels so good!

It's a rare sight to see you being so tidy! Want to try some decluttering in Lojban while you're at it?

What do you mean? Is there junk in the language too?

The "junk" that most beginners (and even pros) want to get rid of are terminators—all those repetitive ku's:

lo ninmu ku klama lo tcadu ku lo zdani ku lo zarci ku lo karce ku The woman goes by car from home via the shop to the city.

Ugh, five ku tokens in a single sentence! That's a lot of work.

Well, Lojban actually lets you omit terminators whenever your intended meaning is clear to the listener. The golden rule is:

If dropping it doesn't change how the sentence is parsed, you can throw it away.

That sounds like a bit of a judgment call...

Here are the ABCs of Elision:

1. End-of-sentence terminators: You can almost always drop these.
2. ku: You can usually drop this, except when it's right before the selbri.
3. kei: Be very careful with this one. Don't drop it unless it's at the end of the sentence.

The key is that your "brackets" must still be unambiguous to the listener.

So:

lo ninmu ku klama lo tcadu lo zdani lo zarci lo karce

Wait, why can't we drop ku right before the selbri?

Look at what happens if you do:

lo ninmu klama lo tcadu …

That turns it into a tanru ("woman-type goer")! The listener won't know where the description ends and the main verb begins.

Precisely.

So that one ku has to stay? That's kind of annoying.

There's a better way! You can use cu—the selbri marker. It tells the listener "the main verb starts here," and it effectively closes all open descriptions before it.

Now that's much cleaner!

And what was the deal with not dropping kei in the middle of a sentence?

Words like nu and du'u are greedy—they will keep swallowing the following words into the sub-clause until they hit a kei. If you drop kei, they might swallow the rest of your sentence! cu can also act as a stop-gap, but it's safer to keep the kei visible unless it's at the very end.

Got it: lo/le descriptions are easy to close, so ku is often gone. But abstractions (NU) are greedy, so keep the kei around until the very end.

Tense in Lojban is straightforward: just put a PU-series word immediately before the selbri.

Simple enough. But what do you mean by "tag"?

ca, pu, and ba are what we call tags. They "stick" to the following word and give it a specific time or place context. For now, we'll just stick them onto the selbri.

Got it. They're like little context stickers!

You can get even more specific by adding distance words after your time direction: zi for "near" and zu for "far."

Direction plus distance—that's a neat little system.

That creates a compound tag. And we can take it even further by adding duration: ze'i for a short duration and ze'u for a long one.

So the order goes: Direction → Distance → Duration. I think I can remember that.

The exact same pattern works for space too! In Lojban, time and space use the same parallel structure (we call both "tense" in a broad sense).

The template is: Direction → Distance → Dimension (area/region).

FAhA (direction), VA (distance), VEhA (extent).

That's a lot of little words to stack up, but at least the template is consistent.

Wait, we didn't use any of these tags in the earlier lessons. What tense were those sentences in?

Tense is completely optional. If you don't use any tags, the time and place are just understood from context—much like how Japanese leave things like singular/plural or tense vague when it's already obvious.

But will people actually understand me?

Of course! Natural languages leave out massive amounts of detail all the time, and we still communicate just fine.

Alright, so the lesson is: omit the tags when it's clear, and add them when you need to be precise.

Continuing our look at context tags, let's talk about aspect (or ZAhO cmavo). This tells you which phase of an event you're referring to.

Which phase? You mean like... stages?

Think about the act of running:

• The state before you start.
• The moment you begin.
• The act while it is ongoing.
• The moment you stop.
• The moment you finish.
• The state after you've finished.

Lojban has a specific tag for each of these "slices" of an event.

I see. So for something like "eating an apple," we could talk about being about to eat, starting to eat, being in the middle of eating, stopping, finishing, or having already eaten.

Exactly! And just like tense tags, these aspect tags go immediately before the selbri.

Right — it really is the tense follow-up: tags go before the selbri. So my try at the pattern would be:

mi pu'o citka lo plise I’m about to eat the apple.

mi co'a citka lo plise I begin eating the apple.

mi ca'o citka lo plise I’m eating the apple.

mi co'u citka lo plise I finish eating the apple.

mi ba'o citka lo plise I’ve already eaten the apple.

You can also combine tense and aspect! The standard order is tense first, then aspect.

mi pu zi ba'o citka lo plise A little while ago, I had just finished eating the apple.

Alright, let's try these! 1: la .soran. ca pu'o sipna 2: la .kocon. ca'o pinxe lo djacu ku lo rirxe ku 3: le melbi cipni pu zu ba'o cliva 4: Wait, "That man over there"... is that vu nanmu (far man)? So, lo vu nanmu ku?

Spot on. lo vu nanmu ku means "the distant man" or "that man over there."

Then 4 is: lo vu nanmu pu zi co'u jmive.

Exactly! Another nice side benefit: putting tense or aspect tags before a selbri also marks the start of the verb, acting much like the cu we learned earlier. It prevents "accidental tanru" from forming.

Oh, cool! So if a selbri is tagged, it can't be mistakenly swallowed as part of a compound word on its left.

Correct. And just like before, once you've built your tagged selbri, you can wrap the whole thing in lo … ku to turn it into a description.

So we can basically nominalize an entire phased event. That's powerful!

We've seen how tense and aspect tags work before the selbri. But these tags have another major job: they can be placed before a sumti, acting much like prepositions in English.

So far tags sat on the selbri — e.g. le nanmu ba zi ze'u ca'u vu ve'u bajra.

Right. You can also say things like:

I see. So these tags are versatile—they act like adverbs when they're on the selbri, and like prepositions when they're on a sumti.

Exactly. And here's a shorthand: if you want to use a tag with an unspecified place (zo'e), you can just use tag + ku. It turns out ku is a very busy word!

Does the tagged sumti modify the whole sentence, then?

Generally, yes. You can think of the tagged sumti as providing the context for the entire bridi. For example: "I meet you; and this whole event took place before some reference time."

We have another family of tags called BAI—these are modal tags derived directly from brivla. The meaning of the tag corresponds to the x₁ slot of its source word. (And just like before, you can use se/te/ve/xe to swap those slots!)

Ah, so fi'e means "by the creator" because it's derived from the x₁ of finti (to create).

Precisely! BAI tags pull out the x₁ of their source brivla. And if you use those swapping words (se, te, etc.), you can access the other slots, too. It's exactly the same logic we used with lo se klama.

Whew, that's a lot of new vocabulary to digest.

Sorry about that! I might have gotten a little carried away. Since we're officially at the halfway point, now is probably a perfect time to take a breather, review your notes, and try to commit some of these to memory.

Let's take a break from tags for a bit and talk about copulas—the "be-verbs." Specifically, how to say things like “A is B” when B isn't just a simple predicate.

You mean like when we're setting two things equal to each other?

Something like that. Remember, we already have sentences like ti mlatu (this is a cat) or ti bloti (this is a boat), which don't need any extra words for "is."

True. But I still don't know how to say something like "This is Sora" or "That is the specific boat we were talking about."

Good point! Lojban handles this in a few ways. Let's start with me.

When you say A me B, you're saying one of three things: that A is among the items in group B, that A is identical to B, or that A's group is a subset of B's group.

Basically, think about "the things A points at" and "the things B points at":

• If every thing A refers to is also something B refers to, you're using me (it's either a member of B or a sub-group of B).
• If A and B refer to exactly the same set of things, that's identity (co-reference).

In simple terms: everything that is A is also B.

If I tried to draw that “A sits inside B” idea, it might look like this:

Be careful! me is not used to talk about literal parts of a whole (like a wheel being part of a car). For that, we use the word pagbu.

lo vi klupe cu pagbu le minji This screw is a part of the machine.

Got it. pagbu for the "part-of" relationship, and me for "membership" or overlap.

If you want to express strict identity (A and B are exactly the same thing), we use the word du.

Up until now, we've mostly dealt with straightforward statements. But what if you want to say that something isn't the case? It's time to tackle negation.

The simplest way is to place the word na immediately before the selbri. This basically says, "It is not true that..."

Just like the tense tags we saw earlier, placing na before the selbri also clearly marks where the verb begins. It can even take the place of cu.

That's remarkably simple! Is there an opposite word? Like a way to say, "Yes, it really is true!"

There certainly is! That would be ja'a. It works exactly like na, but it affirms the entire sentence instead of negating it.

Nice — paired particles that work the same way.

So, na and ja'a deal with the truth of the entire sentence. But we also have the NAhE-series of words, which allow you to tweak the meaning of the selbri itself (similar to how SE works).

These words let you build new shades of meaning: the opposite, a neutral midpoint, or just something "other than" the standard definition.

Keep in mind that NAhE words don't act as sentence markers like na or cu do. The [NAhE] + selbri combination is still treated as a single unit and can still form part of a tanru.

From the definitions alone, na and na'e don’t look that different…

na negates the entire claim (it says the whole sentence is false). na'e simply points to a different predicate ("not green" still implies we're talking about a color—it's what we call "suggestive negation").

Oh, I see. So ti na nanmu means "This thing is not a man" (it could be a rock for all we know). But ti na'e nanmu suggests it's something like a man or has a gender, but specifically is not a man.

In general, na is your safe default for negation. Use the NAhE series when you want that specific "it's another kind of..." nuance.

Negate with na, refine with NAhE. Simple!

Today's topic: questions! There are two main families: truth questions (yes/no) and fill-in-the-blank questions (for things or actions). First, a quick test, Sora—translate: "I am not a human."

mi na remna—but hey, I definitely am human!

Good—glad to see you haven't forgotten na. To ask a yes/no question, just place the word xu at the beginning of the sentence (or right after the .i marker).

What kind of example is that...? And what's an "attitudinal"?

Attitudinals express emotions or attitudes—we’ll cover them in detail in the finale. For now, the important question is: how do you answer? You can use go'i to mean "Yes, that's right" or na go'i for "No, that's not it."

(Wait, so she wasn't your mother?) So go'i basically repeats the previous statement, right?

Exactly. Grammatically, go'i acts as a selbri that means "the same as the last sentence." So na go'i just means "that sentence isn't true." This is usually all you need for answering yes/no questions.

Got it. English yes/no flips on negative questions — how does Lojban handle that?

It's a common stumbling block! In Lojban, go'i echoes the exact previous statement. If you're asked a negative question like "Is it true you're NOT my mother?", a plain "yes" or "no" can be confusing. To be perfectly clear, you can use ja'a go'i to strongly affirm the positive version of the sentence.

For "fill-in-the-blank" questions, use ma. You just place ma exactly where the thing you're asking about would go. You can even use several in one sentence! The listener answers by giving the missing information in the same order.

To ask when, where, why, or how, just combine a tag with ma. For example: ca ma (when), bu'u ma (where), ki'u ma (why), or ta'i ma (how—from tadji, meaning method).

mo replaces the selbri: “What is x₁ / what is going on?” Answer with a selbri — same idea as ma, different slot.

Exactly. mo can even be used inside tanru—for instance, mo nanmu means "What kind of man?"

Watch zvati vs stuzi place structure when asking “where is the restaurant?” — ma stuzi le gusta vs le cukta pu zvati ma.

Now let's look at how to urge someone to do something: commands.

To make an imperative sentence, you just replace "do" with the word ko.

So putting na before the selbri with ko means "don't..." That last example is interesting—it takes a moment to wrap my head around it.

ko can fill any terbricmi — the command is “make this bridi true.”

So instead of ordering someone to love you, lo prenu cu prami ko is more like "Behave in a way that results in people loving you."

ko alone already spans “do it” through “please do”; that covers most imperative needs.

If you want to add more nuance—like making a polite request, a suggestion, or granting permission—you can use attitudinals at the start of your sentence:

(Plus pei for “may I?” patterns — see below.)

Some sentences keep do instead of ko — proposals and permissions aren’t always straight imperatives.

We’ll go deeper into the nai suffix in the finale. But generally, if it's a suggestion or a request rather than a direct order, it's very common to use do instead of ko.

You can also use the word pei to turn an attitudinal into a question:

I see. So nai flips the attitude, similar to how to'e flips a brivla. This is getting deep!

Relative clauses allow you to attach extra information to a sumti. You can use a full sentence with poi or noi, or a simple link to another thing with pe or ne.

The key difference is that poi restricts the group you're talking about, while noi just adds an incidental description.

Whether you place the relative clause inside the description (lo [selbri] poi … ku) or after it (lo [selbri] ku poi …) can slightly change the meaning. For most everyday talk, either works, but check the refrence grammar if you need to be extremely precise.

We widen selbri → practically sumti again (like Lesson 8).

Déjà vu!

It's a similar pattern: be and bei allow you to fill the slots of a brivla while it's still being used as part of a description.

If you want to turn "eating an apple" into a single property, you'd say: citka be lo plise. This creates a new concept: "eats-apples."

So lo citka be lo plise would be "an apple-eater"!

Why bei after the first sumti?

If we used be for every link, complex nested structures would become impossible for the parser to untangle. bei ensures that the groupings stay perfectly clear.

These be-phrases can even be used inside tanru, and you can use be'o to stop the chain before it accidentally swallows the next word in your description.

Explore the toolkit: Notice the subtle differences between:

• ti lumci be lo dinju be'o minji
• ti dinju lumci minji
• ti me lo minji poi ke'a lumci lo dinju
• ti minji lo nu lumci lo dinju

Time for numbers! The digits are pa, re, ci, vo, mu, xa, ze, bi, so, no (1–9, 0). Notice that most of the vowels follow the a, e, i, o, u pattern (though mu breaks it slightly by using u).

To make larger numbers, you just string the digits together: pano (10), recivo (234), or renopamu (2015).

There's a lot to memorize here, but for now, focus on the digits, the article li, and how to use them with lo. You can pick up the more advanced quantifiers as you need them!

Next up: quotation! We'll look at how to quote Lojban directly and indirectly, plus some handy tricks for referring back to things you've already said.

Before we dive into the deep end, here's a non-logical connective: jo'u. This is what you use when you just want to group two things together like a bundle, meaning "X and Y together."

Finally, we have attitudinals. These are "contextual stickers" that attach to a word to express how the speaker feels about it, but they don't change the underlying grammar of the sentence. They're what make Lojban feel "alive."

And that, my friend, was the main course—congratulations! You've officially made it through the basics of Lojban.

.ui cai—We actually did it! I can't believe how much we've covered.

It really is an impressive list: pronunciation, basic sentences (bridi), compounds (tanru), place structures (SE and FA), abstractions (NU), time and space (tense), aspect (ZAhO), modal tags (BAI), negation, questions, commands, relative clauses, identity (me/du), internal links (be), numbers, quotations, connectives, pronouns, and attitudinals... .i pei—how are you feeling about it all now?

.ui—To be honest, it's been a ride! I love that there's no messy inflection to worry about, and in some ways, the logical structure feels even simpler than English once you get the hang of it.

coi .kocon. ...Oh, hey Sora! I didn't expect to see you here. mi'e .sevan. .i mi citno mensi la .soran.

Sevan! You're only showing up now? We're already finished!

ki'e .sevan.—Thanks for bringing that over. Now that we're all here, we can officially bring this edition of First Lojban to a close.

je'e do. (No problem!)

But wait! There's one last thing—a little pop quiz to see how much you've really learned. Grab a pen and some paper! The exam is here, and once you're done, check your answers against the answer key.

I got an 81! That's a solid A in my book!

Nicely done! You've officially completed the basic track.

If you're still hungry for more, there's also the "Back Alley"—two extra "dungeon" chapters (Lesson 27, Lesson 28) for those who want a little more logic-flavored fun. But for everyone else, it's co'o for now!

co'o

Welcome to the "Back Alley"! This is a little philosophy and logic gym tucked away next to the main Lojban course. It's totally optional, but it's a lot of fun if you like to think about how language and thought fit together.

When we make existential claims (saying that something exists), the truth of what we say depends entirely on our domain of discourse. For example, saying "there are two yellow things" might be true if we're only talking about a specific picture or a scene in a movie, but it would be false if we were talking about everything in the entire universe. Whenever you see a quantifier, always ask yourself: What is the frame we're counting within?

Ah, I see. So "a bird exists" usually just means “there's a bird in our line of sight,” not necessarily that a bird exists in every possible corner of the cosmos.

Since we're just scratching the surface here, I highly recommend checking out the quantifier chapters in the CLL for full tables and more formal exercises. This little "Back Alley" is only here to point you toward the machinery!

Wait — there's more?

Hey, think of it as New Game Plus. You've cleared the main story; now we get into the hidden rooms.

…Who are you again?

Sevan. Your sister? We live in the same house?

Oh. Right. When did you get into Lojban?

About two lessons before you. Anyway, first topic: letter words.

You already know that Lojban uses Latin letters. Back in Lesson 2, you learned how to name them:

• Consonant letter names: [consonant] + y + . → by., cy., sy., zy. …
• Vowel letter names: . + [vowel] + bu → .abu, .ebu, .ibu, .obu, .ubu

A word used as a letter name is called a lerfu (letter/symbol word).

So by. is the Lojban name for the letter "b" and .abu is the name for "a". Got it. But I never heard about the period . or apostrophe '. What are those called?

Good catch. They're special:

• Period . → denpa bu ("pause-letter")
• Apostrophe ' → .y'y. (literally the cmavo .y. spoken as a lerfu — two glottal pauses around the schwa)

So denpa bu is literally the "waiting letter." A pause. That's kind of poetic.

Also works for letters Lojban doesn't normally use, like h, q, or w. Since ' already sounds like an "h", people often write h as .y'y. bu ("an h-like letter"). And q is typically .ky. bu ("a k-like letter"), since q in most languages sounds like k.

The real power tool here is bu. Stick bu after almost any Lojban word, and it becomes a lerfu — a symbol standing for that word.

Common examples you'll see in practice:

So .ui bu is literally the emoji :)? That's delightful. Can I put bu after any Lojban word?

Almost. A few words eat up whatever comes right after them before bu gets a chance to act:

• zo, zoi, la'o, lo'u — all of these grab the next word or next passage as a quotation before bu can form a lerfu.

So zo bu doesn't make a "zo-letter"; it quotes the word bu itself.

Edge case, sure. I'll remember: quotation words come first.

Here's something subtle. Consider this sentence:

Wait — "a gives bc"? Not "a gives b to c"?

Exactly. by. cy. is a sequence of lerfu and sequences are treated as a single sumti — the string "bc". So the sentence has only two arguments: the giver (a) and the thing given (bc).

Word spacing doesn't matter: by.cy. and by. cy. both mean the letter string "bc".

So how do I say "a gives b to c"?

Use boi to explicitly end the lerfu (or number) sequence:

boi cuts the lerfu string at that point. So cy. stays separate, filling x₃ of dunda as the recipient. Okay, I'll remember that one.

boi works with numbers too. If you ever see something like li pa by. (the number 1 followed by the letter b), writing li pa boi by. keeps them from fusing into a hybrid string.

Last topic: how to Lojbanize acronyms like NASA, FBI, CD, or BPFK.

The official recipe:

So every letter is just spelled out in Lojban order, smooshed together, with a final consonant tacked on?

Exactly. The last consonant makes it a proper cmevla (remember: name words must end in a consonant). As long as it's intelligible, you have some flexibility — some people drop all the bu parts and just use the Lojban letter names directly, others invent shortcuts. Communicability is the real rule.

Huh. So Lojban texters probably have their own abbreviation styles by now.

Almost certainly. The language grows with its speakers.

You know what bugged me about Lesson 3? Koshon said cmevla can have stress somewhere other than the second-to-last syllable, and then Sora just yelled "let's ignore that!" and we moved on.

Classic Sora. Okay, let's fix it. This lesson is about two things: marking non-default stress, and the consonant clusters that are never allowed anywhere in Lojban.

You know the default: stress the second-to-last syllable, ignoring y.

But names from other languages don't always fit that pattern. Lojban lets you override it: capitalize the letters of the stressed syllable.

Just capitalize the whole stressed syllable?

Technically it's enough to capitalize just the vowel, but capitalizing the whole syllable is easier to read. Either way is valid.

Some writers also use diacritical marks — like la .akásakas. — but that's non-ASCII and not everyone can type it. Capitals work everywhere.

So the two forms la .akasakas. and la .aKAsakas. represent the same name with different stress placements, not two different words.

Exactly. It's an accent record, not a spelling distinction.

One practical point: whether to preserve the original accent is up to personal taste. Some people think Lojbanization should change the accent (it's part of adapting the word). Others want to keep the original as close as possible. Both views are reasonable.

Now the important list. Lojban avoids certain sound combinations — usually because they're easy to mishear or because they don't fit the word-boundary system. Here are the key forbidden patterns:

That's a lot. Do I have to memorise all of those to make a cmevla?

For everyday Lojbanization you mostly need rules 1 and 2 — they cover the vast majority of cases. The others are rarer. And the fix is always the same: insert a y between the two problem consonants.

So y is the universal lubricant between stuck consonants.

Exactly. And two more vowel rules to watch when making name words:

So "Tokyo" (東京 = Toukyou) would be .tokiion. to avoid the kyo glide becoming a semivowel?

Yep, or .tokion. if you're fine with the kyo-glide. Either is defensible. The point is that the parser can still tell the words apart.

Try a few yourself:

Lojbanize the following names. Remember: the word must end in a consonant, no forbidden clusters, apostrophes between non-i/u vowel pairs.

Okay:

1. .kiioton. — kio would make ki a semivowel before o, so I doubled the i. Or just .kiioton. / .kioton. if I'm okay with the glide.
2. .mikun. or just .mik. — straightforward.
3. .djans. or .djang. wait, final ng → I need a consonant… .djangn. — hmm, gn is allowed (voiced+nasal). Actually just .djans. or .djan. is fine too.

All reasonable. Option variety is a feature, not a bug. The community will converge on popular forms for well-known names.

So wait — Sevan showed me earlier that you can type .o'imuXAGjisofyBAKnicuZVAtilePURdi (all one string, no spaces) and a Lojban parser can still tell what every word is?

Yes. That's not a trick — it's a core design property called audio-visual isomorphism (AVi for short).

Iso-what-now?

"Isomorphism" means "same shape." The idea is:

• Any properly spoken Lojban utterance can be uniquely written down.
• Any properly written Lojban text can be uniquely read aloud.

This is stricter than most languages. In Japanese, for example, a string of hiragana like くるまでまとう could parse as "waiting by car" or "waiting until [he] comes" — the boundary between words is ambiguous. Lojban forbids that.

Recall the three word types (Lesson 3):

• brivla — has a consonant cluster in the first five letters, ends in a vowel.
• cmavo — no internal consonants except possibly the very first, ends in a vowel.
• cmevla — ends in a consonant.

These shapes were engineered so that once you read a stream of sounds (including where the stress falls), there is exactly one way to cut it into words. Every word boundary is recoverable from the sound alone.

That's why you can ditch spaces if you include stress marking?

Exactly. Spaces are a courtesy to the reader, not a grammatical necessity. Lojban's design makes them optional — as long as stress is explicit, the string is still unambiguous.

(That said: please use spaces when writing to humans. Reserve the spaceless form for puzzles and Twitter character limits.)

Now you can fully understand why certain dots are required:

1. Before vowel-initial words — .i, .abu, .iu, etc. Without the dot, a preceding final vowel might run into the opening vowel and create ambiguity about where one word ends and the next begins.
2. After words ending in y — .y. is a cmavo; the trailing dot keeps it from being read as the start of the next word.
3. Around cmevla — la .soran. has a dot at each end, clearly marking the name as a single unit separate from surrounding words.

All three rules are there for the same reason: to keep word boundaries clear.

So "omajinaï" (Koshon's word for the dots back in Lesson 2) was actually "word boundary markers"!

Right. Koshon skipped the explanation to keep Lesson 2 simple. Now you know.

This design has a few practical consequences:

So the word-shape rules I grumbled about learning aren't arbitrary — they're load-bearing architecture.

Exactly. Lojban is less like a collection of rules-for-rules'-sake and more like an engineered system where each constraint earns its keep. Take out one rule and something else breaks.

That's actually… kind of cool? It's more like a proof than a language.

Some people say exactly that. Whether that's a feature or a bug is left as an exercise for the learner.

So far I've been collecting cmavo like pocket lint: lo, le, nu, be, poi, noi, cu, ko'a, fa, se … they're everywhere. Is there any system?

Yes! Each cmavo belongs to a selma'o — a syntactic subcategory. Words in the same selma'o behave identically in terms of grammar: you can swap them out without causing a parse error (though the meaning changes, of course).

And what's the selma'o called?

Each selma'o is named after one of its representative members, written in capitals. For example:

• The article class (the "sumti-former" cmavo like lo, le, la …) is called LE.
• The tense class (time-distance markers like zi, za, zu …) is called ZI.
• The abstractors like nu, du'u, ka … are the NU class.
• The selbri negation cmavo na, ja'a are the NA class.

You've actually already been using these names! SE for se/te/ve/xe, FA for fa/fe/fi/fo/fu, COI for coi/co'o/ki'e…

Oh — that's why Koshon kept saying "SE class" and "FA class." They were just giving me the selma'o names all along!

Exactly. The names aren't invented arbitrarily — they're always the simplest or most common member of the class, uppercased. You can find them in any Lojban dictionary next to each cmavo entry.

Within a selma'o, any member can replace any other and the sentence will stay grammatical. Only the meaning differs:

All three are in LE selma'o, so they all "work" in the same grammatical slot — just with different references.

Right. Same slot, different meaning. That's exactly what selma'o tracks.

One caution: "article" in the English-grammar sense doesn't map perfectly onto any single selma'o. What English calls articles spans multiple Lojban classes. Selma'o is strictly about syntactic behavior, not English grammatical terms.

Some selma'o are large enough to sub-divide by meaning. The convention is to append a number to the selma'o name:

• KOhA1 — personal pronouns: mi, do, mi'o, mi'a …
• KOhA2 — demonstratives: ti, ta, tu
• KOhA3 — reflexive/anaphoric: ri, ra, ru, ko'a, fo'a …

All of these are still KOhA (same grammar), but the sub-number tells you which semantic cluster they belong to.

So the big picture is: selma'o tells you the grammar, the number tells you the meaning flavour.

Nicely put. And if you ever want to really map out Lojban's grammar systematically, going through the major selma'o one by one is a great approach. It also helps you notice when you're missing vocabulary — you might know the NU abstractors nu and du'u but realize you've never seen si'o or ni.

I know all of those! I just didn't know they had names.

Now you do. And whenever you encounter a new cmavo in the wild, your dictionary will tell you its selma'o — which instantly tells you how to use it, even before you know what it means.

Can I ask something I've been wondering about for a while? Those "x₁ loves x₂" definitions — they're in English. But Lojban isn't English. How do I know the English gloss captures the real meaning?

You don't. Not completely. And that's worth talking about.

When you look up a brivla, you see something like:

prami

x₁ loves x₂

That "x₁ loves x₂" is a translation — a Japanese, English, or Russian window into a Lojban concept. Every translation carries the source language's baggage:

• English "loves" has tense built in (present tense default); Lojban prami doesn't care about tense.
• English separates "love" from "like" culturally; Lojban speakers use prami more broadly.
• English "love" is a verb — implying a doer and a receiver in a specific power arrangement; Lojban relations are more neutral.

So the definition is a hint, not a theorem.

Right. The real meaning of a brivla is the relation it names — a pattern in the world — not the English words used to describe it. When you internalize a brivla, you eventually stop translating and just know the shape of what it connects.

Think of it this way: citka names the "eating" relation. Whether that's "eating", "consume", "to eat", "eats" doesn't matter. The relation connects eater and food. That's the concept.

So "minimum description, maximum interpretation" applies to the definitions themselves too, not just the sentences.

Exactly. The place structure tells you what kinds of things are involved, not exactly how to translate every nuance into your native language.

Each brivla has its slots because that kind of situation usually involves those participants. A "buying" relation (pleji, vecnu) typically involves a buyer, a seller, a thing sold, and a price — so those are the places. A "waiting" relation (denpa) involves a waiter, an event waited for, a state while waiting, and a trigger — so those are its places.

Less-central participants don't get core slots — they get tags (ba, fi'o, etc.). The design choice is: put the most essential participants in the numbered slots, leave the rest for tags.

And you never have to fill every slot. You say what you mean and let context fill the rest.

Lojban's motto, really: minimum necessary statement, maximum interpretive latitude. Don't say more than you mean.

Back in Lesson 28, you used zo'u to write the prenex of an existential statement. But zo'u has a second job: it's a topic marker.

Topic as in… what the sentence is about?

Yes. In most languages, topics and grammatical subjects often coincide, but they're different things. The grammatical x₁ slot in Lojban is whoever fills the first place of the predicate — that might not be what the speaker considers the "topic."

Anything placed before zo'u (any sumti, with or without tags) is marked as the topic of the following sentence.

So the elephant sentence — x₁ of clani is the nose, not the elephant, but the elephant is what we're talking about. The topic sits before zo'u, and then the normal predication follows.

Exactly. Compare the same content without explicit topic:

Grammatically equivalent claim, but the topic/emphasis is different. The zo'u version says: "Here's what we're talking about (elephants), and here's what I'm saying about it (the nose is long)."

In practice, modern Lojban speakers often omit explicit topics and let context carry the weight. But zo'u is there whenever you want to be precise about what the sentence is about, independently of what fills x₁.

So zo'u is doing double duty: prenex-separator in Lesson 28, topic-marker here. Same word.

Same word, same mechanism — you're always separating a "preamble" from the "main claim." In the prenex case, the preamble introduces quantified variables. In the topic case, it introduces what we're discussing. Both are just different kinds of "setup before the sentence proper."

Long time no see! Sevan asked me to guest-lecture on tanru grouping.

Koshon knows this better than I do. Also, Sora listens to Koshon.

I do not… okay maybe a little.

Recall from Lesson 6: tanru group left to right by default.

blabi mlatu zdani = ((blabi mlatu) zdani)

That means "a white-cat house" — the house is cat-like (white-cat-type house). Not "a white (cat-house)" (a house for white cats).

Right, and we got burned by that once. So how do we change the grouping?

Three tools: bo, ke/ke'e, and co.

bo placed between two selbri binds them tighter than the default. Think of it as the multiplication sign in arithmetic: 2 + 3 × 4 evaluates differently from (2 + 3) × 4, and bo is the × sign.

Oh! So blabi mlatu bo zdani finally puts the house in charge of being white, not the cat.

Exactly. bo is the quick way to say "glue these two together first."

If bo is the × sign, ke … ke'e are actual parentheses. Everything between them is treated as one unit before any surrounding selbri are considered.

When would I use ke instead of bo?

When grouping more than two pieces. Watch:

So ke'e is the right terminator for ke. Not ku or anything else.

Right. And ke'e is often optional in practice — if it's at the very end of the tanru before a sumti, the parser can figure it out. But when you're nesting multiple brackets, explicit ke'e is much safer.

co is the trickiest of the three. It does two things together:

Why would you ever want to reverse the direction? That sounds confusing.

It becomes useful when you want to attach a long modifier phrase (one that includes be/bei sumti slots) to a simpler head. Without co, that gets unwieldy:

Oh — co lets you keep the main concept (klama — "goes") on the right where it's easy to read, and pile the modifiers to the left of co. The arguments after co then belong to klama, not to the stuff on the left.

That's the third feature of co: the sumti following co's right-side selbri fill that selbri's places, not the overall tanru's. It's a convenience for readable tanru with embedded sumti.

Now for something fun. Write a short poem using only selbri — no sumti, no tags. Just predicates strung together with .i between them. You can use bo, ke, and co to shape your tanru. The brevity forces you to let the predicates carry everything.

Koshon calls this tanru pemci — predicate poetry.

Here's an example:

No nouns, no pronouns, no time words — just pure predicates. And it works. Because Lojban predicates already contain the "is" and "does" without needing extra words.

Tense is optional. Number is optional. Subject is optional. Tanru is ambiguous by design. All of that ambiguity becomes atmosphere in poetry.

Okay I'll try one:

.i barda mlatu
.i cmalu mlatu
.i mutce melbi
.i gleki munje

(Big cat. Small cat. Very beautiful. Happy world.)

I love it.

…I also love it, but I'm not going to say so.