The following cmavo are discussed in this section:

The term “anaphora” literally means “repetition”, but is used in linguistics to refer to pronouns whose significance is the repetition of earlier words, namely their antecedents. Lojban provides three pro-sumti anaphora, ri, ra, and ru; and three corresponding pro-bridi anaphora, go'i, go'a, and go'u. These cmavo reveal the same vowel pattern as the ti-series, but the “distances” referred to are not physical distances, but distances from the anaphoric cmavo to its antecedent.

The cmavo ri is the simplest of these; it has the same referent as the last complete sumti appearing before the ri:

The ri in Example 7.30 is equivalent to repeating the last sumti, which is la .alis., so Example 7.30 is equivalent to:

Note that ri does not repeat le ri kumfa, because that sumti is not yet complete when ri appears. This prevents ri from getting entangled in paradoxes of self-reference. (There are plenty of other ways to do that!) Note also that sumti within other sumti, as in quotations, abstractions, and the like, are counted in the order of their beginnings; thus a lower level sumti like la .alis. in Example 7.31 is considered to be more recent than a higher level sumti that contains it.

Certain sumti are ignored by ri; specifically, most of the other cmavo of KOhA, and the almost-grammatically-equivalent lerfu words of selma'o BY. It is simpler just to repeat these directly:

However, the cmavo of the ti-series can be picked up by ri, because you might have changed what you are pointing at, so repeating ti may not be effective. Likewise, ri itself (or rather its antecedent) can be repeated by a later ri; in fact, a string of ri cmavo with no other intervening sumti always all repeat the same sumti:

Here the second ri has as antecedent the first ri, which has as antecedent le tricu. All three refer to the same thing: a tree.

To refer to the next-to-last sumti, the third-from-last sumti, and so on, ri may be subscripted (subscripts are explained in Section 19.6):

Here rixire, or “ri-sub-2”, skips la .rik. to reach lo forca. In the same way, riximu, or “ri-sub-5”, skips la .alis., rixire, la .rik., and lo forca to reach lo smuci. As can clearly be seen, this procedure is barely practicable in writing, and would break down totally in speech.

Therefore, the vaguer ra and ru are also provided. The cmavo ra repeats a recently used sumti, and ru one that was further back in the speech or text. The use of ra and ru forces the listener to guess at the referent, but makes life easier for the speaker. Can ra refer to the last sumti, like ri? The answer is no if ri has also been used. If ri has not been used, then ra might be the last sumti. Likewise, if ra has been used, then any use of ru would repeat a sumti earlier than the one ra is repeating. A more reasonable version of Example 7.34, but one that depends more on context, is:

In Example 7.35, the use of ra tells us that something other than la .rik. is the antecedent; lo forca is the nearest sumti, so it is probably the antecedent. Similarly, the antecedent of ru must be something even further back in the utterance than lo forca, and lo smuci is the obvious candidate.

The meaning of ri must be determined every time it is used. Since ra and ru are more vaguely defined, they may well retain the same meaning for a while, but the listener cannot count on this behavior. To make a permanent reference to something repeated by ri, ra, or ru, use goi and a ko'a-series cmavo:

allows the store to be referred to henceforth as ko'a without ambiguity. Example 7.36 is equivalent to Example 7.21 and eliminates any possibility of ko'a being interpreted by the listener as referring to Alice.

The cmavo go'i, go'a, and go'u follow exactly the same rules as ri, ra, and ru, except that they are pro-bridi, and therefore repeat bridi, not sumti – specifically, main sentence bridi. Any bridi that are embedded within other bridi, such as relative clauses or abstractions, are not counted. Like the cmavo of the broda-series, the cmavo of the go'i-series copy all sumti with them. This makes go'i by itself convenient for answering a question affirmatively, or for repeating the last bridi, possibly with new sumti:

Note that Example 7.38 means the same as Example 7.26, but without the bother of assigning an actual broda-series word to the first bridi. For long-term reference, use go'i cei broda or the like, analogously to ri goi ko'a in Example 7.36.

The remaining four cmavo of the go'i-series are provided for convenience or for achieving special effects. The cmavo go'e means the same as go'ixire: it repeats the last bridi but one. This is useful in conversation:

Here B's sentence repeats A's within an abstraction (explained in Chapter 11): le si'o mi go'i means le si'o mi klama le zarci. Why must B use the word mi explicitly to replace the x1 of mi klama le zarci, even though it looks like mi is replacing mi? Because B's mi refers to B, whereas A's mi refers to A. If B said:

that would mean:

The repetition signalled by go'i is not literally of words, but of concepts. Finally, A repeats her own sentence, but with the x1 changed to do, meaning B. Note that in Example 7.39, the tense ba (future time) is carried along by both go'i and go'e.

Descriptions based on go'i-series cmavo can be very useful for repeating specific sumti of previous bridi:

Here the go'i repeats le xekri mlatu cu klama le zarci, and since le makes the x1 place into a description, and the x1 place of this bridi is le xekri mlatu, le go'i means le xekri mlatu.

The cmavo go'o, nei, and no'a have been little used so far. They repeat respectively some future bridi, the current bridi, and the bridi that encloses the current bridi (no'a, unlike the other members of the go'i- series, can repeat non-sentence bridi). Here are a few examples:

(Note: The Lojban does not contain an equivalent of the my in the colloquial English; it leaves the fact that it is the speaker's son and daughter that are referred to implicit. To make the fact explicit, use le bersa/tixnu be mi.)

For good examples of nei and no'a, we need nested bridi contexts:

Finally, ra'o is a cmavo that can be appended to any go'i-series cmavo, or indeed any cmavo of selma'o GOhA, to signal that pro-sumti or pro-bridi cmavo in the antecedent are to be repeated literally and reinterpreted in their new context. Normally, any pro-sumti used within the antecedent of the pro-bridi keep their meanings intact. In the presence of ra'o, however, their meanings must be reinterpreted with reference to the new environment. If someone says to you:

you might reply either:

or:

The ra'o forces the second mi from the original bridi to mean the new speaker rather than the former speaker. This means that go'e ra'o would be an acceptable alternative to do go'e in A's statement in Example 7.39.

The anaphoric pro-sumti of this section can be used in quotations, but never refer to any of the supporting text outside the quotation, since speakers presumably do not know that they may be quoted by someone else.

However, a ri-series or go'a-series reference within a quotation can refer to something mentioned in an earlier quotation if the two quotations are closely related in time and context. This allows a quotation to be broken up by narrative material without interfering with the pro-sumti within it. Here's an example:

Of course, there is no problem with narrative material referring to something within a quotation: people who quote, unlike people who are quoted, are aware of what they are doing.