Chapter 5. Pronouns & Back-References#

Why Lojban Needs Its Own Pronoun System#

English pronouns like he, she, it, they carry implicit information about gender and number. Lojban does not use grammatical gender at all, and number is optional — so the English pronoun system doesn't map cleanly onto Lojban. Instead, Lojban has several series of pro-sumti, each organized around a specific function.

---

The mi-series: Personal Pronouns#

The most fundamental pronouns refer to the participants in the conversation.

Neither mi nor do specifies singular or plural. A spokesperson saying "we believe …" would use mi on behalf of a group.

The "we" pronoun series is more precise than English: Lojban distinguishes whether the listener is included (mi'o = you and I; mi'a = we without you). This eliminates a common ambiguity in English: "We're going to the store" — does that include you?

ko is the command particle. It can go in any sumti slot:

ko klama le zarci Go to the store! (literally: make "you go to the store" true)

mi viska ko Make "I see you" true! = Be seen by me! / Show yourself!

ko kurji ko Take care of yourself! (ko in both x₁ and x₂)

---

The ti-series: Demonstrative Pronouns#

Three pronouns refer to things pointed at in physical space:

ti

this — something near the speaker

ta

that — something at a medium distance (often near the listener)

tu

that yonder — something far from both speaker and listener

ti melbi This is beautiful.

mi klama tu I'm going to that (far-away thing).

Note: ti, ta, tu are true pronouns — they replace the whole argument. To say "this boat" (the boat near me), use a tense tag, not ti:

le vi bloti — the nearby boat (vi = "here")

ti noi bloti — this thing, which happens to be a boat

---

The di'u-series: Utterance Pronouns#

Sometimes you want to refer to something that was said rather than something physical. The di'u-series refers to utterances — sentences, sentences, passages of text:

do na nelci loi mlatu .i di'u jitfa You don't like cats. That [previous utterance] is false.

di'u refers to the sentence as a piece of text. If you want to refer to the situation described (not the words themselves), use la'e di'u ("the referent of the previous utterance"):

mi prami la .djan. .i mi nelci la'e di'u I love John. And I like that [= the situation of my loving John].

Contrast with:

mi prami la .djan. .i mi nelci di'u I love John. And I like [the sentence "I love John"]. — liking the sentence, not the fact

la'edi'u is often written as a single word. It's very common in Lojban text.

---

The ko'a-series: Assignable Pronouns#

Lojban's equivalents of he/she/it/they are the ko'a-series — ten generic pronouns with no inherent reference. You assign them explicitly with goi.

Assigning with goi — forethought form:

la .alis. goi ko'a cu klama le zarci .i ko'a cu blanu Alice (hereafter: ko'a) goes to the store. She is blue.

Assigning with goi — afterthought form:

la .alis. klama le zarci goi ko'a .i ko'a blanu Alice goes to the store (call her ko'a). ko'a is blue.

Both la .alis. goi ko'a and ko'a goi la .alis. work identically — goi is symmetric.

Unlike English pronouns, ko'a-series pronouns carry no gender, no animacy, and no number — they refer to whatever you assigned them to. You can assign them to people, objects, places, or even abstract things like events.

---

The ri-series: Relative Back-References#

Sometimes you want to say "the aforementioned …" without choosing a ko'a slot. The ri-series points back to recently mentioned sumti:

ri

the previous sumti (the most recently completed argument)

ra

an earlier sumti (vague — not the one ri refers to)

ru

a much earlier sumti (earlier than ra's referent)

la .teris. cu klama le zarci .i ri blanu Terry goes to the store. It [= the store, the most recent sumti] is blue.

Be careful: ri refers to the previous sumti in the text, not necessarily the subject. In the example above, ri is the store (last completed sumti) not Terry. If you want Terry:

la .teris. cu klama le zarci .i la .teris. cu blanu

or assign:

la .teris. goi ko'a cu klama le zarci .i ko'a blanu

ri is convenient for quick back-references; ko'a is for careful, unambiguous longer-range references.

Scan (parallel to the ti-series for distance): ri / ra / ru repeat sumti by recency; go'i / go'a / go'u repeat whole main-bridi patterns the same way (see The go'i-series later in this chapter).

ri before the surrounding sumti is finished: ri always copies the last complete sumti before it — not a phrase that is still open. A useful pattern is “in someone's room”:

la .teris. cu sipna ne'i le ri kumfa Terry sleeps in Terry's room. Here ri = la .teris. (the possessor), because le ri kumfa is not yet complete when ri appears.

The same meaning, spelled out without ri:

la .teris. cu sipna ne'i le la .teris. kumfa

So ri cannot refer to le ri kumfa itself — that would create a self-referential tangle. Inner sumti (here la .teris. inside the larger description) count as more recent than the outer phrase they sit in.

Why you still say the second mi: ri mostly skips mi, do, assignable ko'a…, zo'e, and ri itself. To say “I love myself”, repeat mi — ri will not stand in for the second slot:

mi prami mi I love myself.

For “x₁ of this bridi” in a specific place, vo'a is often clearer; see the section vo'a: The Reflexive below. This pair just shows why ri is not a substitute for another mi.

ti and ri: You may use ri after ti / ta / tu if you have just pointed at something new. Two ri in a row with nothing between them still pick the same antecedent:

la .teris. cu viska le tricu .i ri melbi Terry sees the tree. It [= the tree] is beautiful.

la .teris. cu viska le tricu .i ri du ri Terry sees the tree. It is the same as itself. (Both ri = le tricu — a deliberate “chain” of ri.)

Quotations: ri inside lu … li'u normally does not reach out to sumti in the surrounding narrative; see Chapter 17 for quotation rules and nearby-quote exceptions.

What ri counts and skips:

• ri counts every completed sumti in order of appearance — including ti/ta/tu (demonstratives), since you may have changed what you are pointing at.
• ri skips most KOhA cmavo: mi, do, ko'a-series, ko-imperatives, zo'e, and the ri/ra/ru words themselves.
• A sumti nested inside another sumti (e.g. inside a le description) counts based on its start position, not its container. So inner sumti are "more recent" than the container that wraps them.

Subscripted ri for exact targeting: When you need to skip back further than one sumti, use xi + a number:

lo smuci .i lo forca .i la rik. pilno ri xi re A spoon. A fork. Rick uses [the second-back = the fork].

ri xi re skips one and picks up the next-to-last sumti; ri xi mu skips four and picks up the fifth-from-last. In practice this is cumbersome in speech — prefer assigning a ko'a slot for anything beyond a one-step back-reference. Older texts sometimes write rixire glued together; it is the same as ri xi re.

When subscripts are awkward, use ra: The same scene, but with a vague “some earlier thing” instead of counting:

lo smuci .i lo forca .i la rik. pilno ra A spoon. A fork. Rick uses [some previous sumti — not la rik.].

Here ra usually means lo forca (the fork): the listener uses context. If even ra / ru feel fuzzy, assign once: le forca goi ko'e .i la rik. cu pilno ko'e.

Ordering rules for ra and ru:

• ra refers to a recently used sumti that is not the one ri would pick. If ri has not been used, ra may be the last sumti.
• ru refers to a sumti further back than ra's referent. If both ri and ra are used, ru must be even older.
• A chain of consecutive ri (no other sumti between them) all refer to the same sumti — each ri takes the previous ri's referent as its antecedent, which is the same underlying sumti.

---

vo'a: The Reflexive#

To say the same thing fills two slots of the same bridi (like English "herself", "himself"):

vo'a

the same as x₁ of this bridi

vo'e

the same as x₂

vo'i

x₃; vo'o = x₄; vo'u = x₅

la .alis. cu prami vo'a Alice loves herself.

mi tavla do fo vo'a I talk to you in (my own) language. (vo'a = x₁ = mi)

---

zo'e: The Unspecified#

zo'e is not exactly a pronoun but is worth reviewing here: it fills a place with "something real but unspecified":

mi citka zo'e I eat something (unspecified).

Trailing zo'e slots at the end of a bridi can simply be omitted:

mi citka = mi citka zo'e (eating something unspecified)

But when you need to skip a middle place while filling a later one, zo'e makes the skip explicit:

mi klama la bastn. la .atlantas. zo'e le karce I go to Boston from Atlanta via (unspecified route) by car.

---

Names and Vocatives#

Names in Lojban always end in a consonant and are surrounded by pause marks. They come after la when used as sumti:

la .alis. cu tavla la .djan. Alice talks to John.

To address someone directly (a vocative), use a greeting particle followed by the name:

coi

hello / greetings

co'o

goodbye

doi

O [address marker, no greeting implied]

coi .alis. Hello, Alice.

co'o .djan. Goodbye, John.

doi .teris. ko klama Hey Terry, come here!

doi sets the value of do for the rest of the conversation:

doi la .alis. mi prami do [Addressing Alice:] I love you.

After doi la .alis., do refers to Alice until changed.

mi'e is the self-introduction vocative:

mi'e .djan. I am John.

---

Pro-bridi: The broda-series and go'i#

Just as ko'a-series pronouns replace sumti, Lojban has pro-bridi that replace entire predicates or bridi.

The broda-series (broda, brode, brodi, brodo, brodu) are placeholder relation words, assigned with cei:

la .alis. cu gerku cei broda .i la .djan. broda Alice is a dog (call that "broda"). John is-broda. = John is also a dog.

go'i is the most commonly used pro-bridi: it repeats the previous main bridi, optionally with new sumti overriding the old ones:

mi klama le zarci .i do go'i I go to the store. You [go to the store too].

xu do klama le zarci .i go'i Are you going to the store? — Yes. (go'i alone = "the previous bridi is true")

na go'i No. / That is not so. (negated go'i)

---

ce'u: The Abstraction-Focus Pronoun#

When you build a ka property abstraction (from Chapter 12), you sometimes need to indicate which open slot the property is "about". ce'u is that marker — the lambda variable of Lojban:

le ka melbi — the property of being beautiful (ce'u is implicit in x₁)

le ka ce'u melbi — the property [of x] of being beautiful (same, but explicit)

le ka mi prami ce'u — the property of being-loved-by-me (ce'u fills x₂)

Without ce'u, x₁ of the abstraction is assumed to be the open slot. With ce'u explicitly placed, you can build properties for any slot:

do cnino mi le ka ce'u melbi — You are new to me in the quality of being beautiful.

ta mutce le ka ce'u barda — That is very big. (literally: that greatly has the property of bigness)

ce'u is used with gismu like ckaji (has property), simlu (seems), mutce (very), traji (superlative), and wherever a ka abstraction needs to specify its open place.

---

The go'i-series: Pro-bridi for Repeating and Referring#

go'i (covered briefly in the pronoun chapter) is one member of a larger family of pro-bridi — particles that substitute for whole predicates. The full series:

All of these can take new sumti to override specific places of the referred bridi:

mi klama le zarci .i do go'i I go to the store. You do too. (go'i = "go to the store", x₁ changed to do)

mi klama le zarci .i do go'i le zdani I go to the store. You go to the home. (go'i with x₂ overridden)

nei — this very bridi — is the opposite move from go'i: go'i replays the previous sentence; nei refers to the current predicate relationship while you are still inside it. That is what makes it useful for recursion and reflexive wording: you can nest the ongoing bridi inside itself (often as le nei “the/event of this bridi”) without naming the predicate again.

ra mipri le nei They hide themselves. (le nei packages this bridi as a sumti so the patient of mipri is tied back into the same clause — a natural reflexive).

Longer prose often uses le nu le nei … (an abstraction whose inner text still means “this bridi”), which is how translators handle English reflexives and self-referential descriptions in one smooth Lojban clause.

go'o — a later bridi (forward reference). Rare in textbook prose; one pattern is promising something that will be said next:

mi nupre le nu mi go'o I promise the event of my doing [what I will say next].

co'e (selma'o GOhA) is the pro-bridi for unknown/unspecified predicate — like zo'e but for selbri:

mi co'e le zarci I do [something unspecified] to the store.

le go'i — descriptions from a previous bridi's places:

Any GOhA word can be wrapped in le (or another descriptor) to extract a specific place of the repeated bridi as a sumti:

le xekri mlatu cu klama le zarci .i le go'i cu melbi The black cat goes to the store. The [same thing = the black cat] is beautiful.

le xekri mlatu cu klama le zarci .i le se go'i cu barda The black cat goes to the store. The [x₂ = the store] is large.

This lets you refer back to specific arguments of a previous sentence without repeating them or assigning ko'a slots.

go'e — the bridi before last:

go'e (= go'i xi re) is especially useful in conversation where two speakers alternate:

A: mi ba klama le zarci A: I will go to the store. B: mi nelci le si'o mi go'e B: I like the idea of my going [to the store — repeating A's bridi]. A: do go'i A: You will [go to the store] too.

Here B's go'e repeats A's sentence; A's final go'i repeats B's sentence. Tense particles are carried along automatically.

ra'o — updating pro-cmavo in reported speech:

When you quote or repeat someone else's words, the assignments they made (ko'a = X, broda = Y) were made in their context. ra'o tells a GOhA to re-evaluate pronoun assignments for the current speaker's context instead of copying them verbatim:

la alis. cusku lu mi prami ko'a li'u Alice says "I love [ko'a]." ra'o [Re-evaluate: for the listener, "mi" = Alice, "ko'a" = whatever Alice assigned it to.]

Without ra'o, go'i after reported speech would re-assert literally with the original speaker's referents. ra'o triggers context-shift so the repeated bridi makes sense from the new speaker's vantage point.

---

da, de, di — bound sumti variables (logic)#

da, de, and di are the first three bound sumti slots — “thing-1”, “thing-2”, “thing-3”. They are meant for general statements (existence, “for all”, “if…then”) with quantifiers and often a prenex zo'u. Storytelling usually prefers names or ko'a; you reach for da when the point is logical form, not a specific individual you are tracking.

A minimal taste (details, scope, and negation live in Chapter 21):

da zo'u da klama Something goes. (existential reading: there is something that goes.)

su'o da poi gerku zo'u da blabi Some dog is white.

su'o da su'o de zo'u da prami de Something loves something. (two existentials in the prenex.)

If you see da in older text without an obvious quantifier, it often abbreviates su'o da (“at least one thing”). Don't confuse da with ko'a: ko'a is your label for a definite referent; da is a pattern variable under quantifiers.

---

du — identity (“is the same as”)#

du is a selbri meaning identity: the sumti it connects are the same entity. It is the everyday “X is Y” when you identify or define, not merely compare.

ko'a du le nanmu ko'a is the man — you are saying what ko'a is (answering “what is ko'a?”).

la .alis. du le ninmu Alice is the woman — one person, two descriptions.

By contrast, mintu and similar selbri make a claim about similarity or sameness-of-kind between things you already have in mind — not the same speech act as a bare identity definition. (du is related historically to dunli, but dunli has extra places for how they are equal; du does not.)

du also appears as mathematical equals in Chapter 18 (li … du li …). The same “same value / same thing” idea, in number land.

---

da'o — cancel assignable references#

da'o (selma'o DAhO) clears assignable back-references: ko'a–ko'u, fo'a–fo'u, broda–brodu after cei, and similar — so none of them inherit an old referent. It does not change who mi and do are (speaker and listener); it does not erase ri's discourse rules (there is nothing to “store” for ri anyway).

Use da'o when:

• you join a conversation and want to signal “don't assume my ko'a is yours”;
• a long stretch used many assignments and you want a clean slate before a new block;
• you are switching exercises or examples in a textbook.

Topic markers ni'o / no'i can also reset discourse context (see Chapter 17); heavy ni'o sometimes appears together with da'o at a major section break.

la .alis. goi ko'a cu gerku .i ko'a cu melbi Alice, called ko'a, is a dog. It is beautiful.

da'o .i ni'o la .djan. cu klama le zarci [Drop assignable pronouns.] [New topic:] John goes to the store.

After da'o, ko'a has no value until you assign it again with goi or cei.

---

Lujvo from pro-sumti rafsi (advanced)#

Some cmavo in KOhA and GOhA have rafsi (combining forms) for building lujvo. This is specialist material — you can read Lojban fluently without ever using it — but it explains odd dictionary entries.

• Idea: attach a pro-sumti rafsi inside a lujvo as if it were an internal argument of the underlying predicate. Example from the reference grammar: donma'o (“you-cmavo”) ≈ second-person pronoun — glossed as cmavo be zo do (a cmavo for “you”). See the Lujvo from pro-sumti rafsi section in this chapter above.
• zi'o and other “place tricks” in lujvo use zi'o-rafsi patterns; the full convention is in Chapter 12 (abstractions) and Chapter 14 (how lujvo are formed and scored).
• co'e, du, and bu'a also have rafsi; compounds built from them pick up context-dependent meanings, like any vague tanru.

For the cmavo-to-rafsi table, Chapter 14 lists morphology details; jbovlaste lists rafsi when you need to check a coinage.

---

bu'a: Selbri Variables#

bu'a, bu'e, and bu'i parallel da, de, di, but they stand in for selbri (relations), not sumti:

da bu'a de — something stands in some relation to something else

da bu'a de .ije de bu'a da — if A relates-to B then B relates-to A (symmetry)

They appear in logical statements (prenexes) when you want to make claims about all predicates of a kind or about the existence of some relation, without naming it:

su'o bu'a zo'u mi bu'a do There is some relation between me and you. (I relate to you somehow)

ro bu'a zo'u ganai da bu'a de gi de bu'a da For every relation: if A-relates-to-B then B-relates-to-A. (claim about all symmetric relations)

bu'a-series variables work exactly like da-series (bound by quantifiers in the prenex), but they fill the selbri slot rather than sumti slots.

---

Anaphoric Pro-sumti and Pro-bridi: Full ri-series and go'i-series#

The ri-series back-references are based purely on recency in the discourse stream:

Key points:

• ri skips mi, do, and the ko'a/fo'a-series — but it does count ti/ta/tu (demonstratives), because you may have changed what you are pointing at since the last use.
• ri also skips zo'e, lerfu strings used as pronouns, and the ri/ra/ru words themselves.
• In la .teris. cu klama le zarci .i ri barda, ri = le zarci (the last completed sumti before ri), not la .teris.. Names with la are counted when they appear as complete sumti earlier in the string.
• Chaining: two ri in a row with no other sumti between them pick the same referent — see ri du ri above.

To avoid ambiguity in complex text, prefer explicit ko'a-assignments over ri.

---

Summary#

• Names: end in consonant, written with pause marks .name., used after la
• Vocatives: coi (hello), co'o (bye), doi (O …), mi'e (I am …)
• la'e di'u = the situation described by the previous sentence (not the sentence itself)