So far, lerfu words have only appeared in Lojban text when spelling out words. There are several other grammatical uses of lerfu words within Lojban. In each case, a single lerfu word or more than one may be used. Therefore, the term “lerfu string” is introduced: it is short for “sequence of one or more lerfu words”.

A lerfu string may be used as a pro-sumti (a sumti which refers to some previous sumti), just like the pro-sumti ko'a , ko'e , and so on:

In Example 17.21 , .abu and by. represent specific sumti, but which sumti they represent must be inferred from context.

Alternatively, lerfu strings may be assigned by goi , the regular pro-sumti assignment cmavo:

There is a special rule that sometimes makes lerfu strings more advantageous than the regular pro-sumti cmavo. If no assignment can be found for a lerfu string (especially a single lerfu word), it can be assumed to refer to the most recent sumti whose name or description begins in Lojban with that lerfu. So Example 17.22 can be rephrased:

(A less literal English translation would use “D” for “dog” instead.)

Here is an example using two names and longer lerfu strings:

Perhaps Alexander's name should be given as ru'o.abupyky instead.

What about

Does this mean that A gives B to C? No. by. cy. is a single lerfu string, although written as two words, and represents a single pro-sumti. The true interpretation is that A gives BC to someone unspecified. To solve this problem, we need to introduce the elidable terminator boi (of selma'o BOI). This cmavo is used to terminate lerfu strings and also strings of numerals; it is required when two of these appear in a row, as here. (The other reason to use boi is to attach a free modifier – subscript, parenthesis, or what have you – to a lerfu string.) The correct version is:

where the two occurrences of boi in brackets are elidable, but the remaining occurrence is not. Likewise:

requires the first boi to separate the lerfu string xy. from the digit string ro.