When you make a mistake while speaking, whether in your wording or your grammar, you don't normally bother to correct it — if you even realise you made a mistake in the first place. That's because natural languages are fairly redundant (for this very reason!); and we normally rely a lot more on context than on what we actually hear, anyway. If we do catch ourselves making an error, we stumble out a correction that will do the trick, without going into details like how many words should be cancelled: again, context is almost always more than adequate. So if I say

I downloaded and learned some Esperanto vocabulary. Er, Lojban vocabulary.

context and common sense dictate that Lojban vocabulary is meant to replace Esperanto vocabulary. But what if it was meant to replace some Esperanto vocabulary? Or downloaded and learned some Esperanto vocabulary? We wouldn't normally care, in natural languages.

But Lojban is Lojban precisely because it is not a natural language. And this kind of imprecision does not sit well with how the language was designed. So Lojban allows you to be more precise about what words you are correcting. Whether it is actually too be precise to be useful — well, that's something for usage to determine. But the tools are available, if you want them.

si erases the immediately preceding word. If you want to erase two words in a row, you say si si after them. So the correction above would be in Lojban

.i mi te benji je cilre loi bangrnesperanto valsi si si lojbo valsi.

The problem with si is, you have to count words. This can get tedious, and you shouldn't have to keep a transcript of your words when you want to correct yourself. The other correction word Lojban offers is somewhat more helpful: sa erases a phrase. It works by taking the word following it, which starts the phrase to serve as the correction. It then goes back in the sentence, looking for the last time you used a phrase starting with the same word. (Same selma'o, actually.) Once it finds the last such phrase, it replaces all text from that phrase up to sa with the phrase following sa. For example:

.i mi te benji je cilre loi sa .i mi cilre loi lojbo.

The correction following sa is a sentence; you know that, because the first word after sa is the sentence marker, .i. So the sentence following sa replaces the current sentence up to and including sa. Or consider:

.i mi mrilu fi do ca le purlamdei sa ca la reldjed.

The correction is ca la reldjed. 'on Monday'. So what it replaces is everything from the last phrase beginning with ca: ca le purlamdei 'yesterday'. The English version would be "Yesterday I mailed you... actually, it was Monday."

Tip: Of the Lojban erasure words, sa is not as widely known as si, and another, unofficial solution has arisen on IRC (Internet Relay Chat) to the problem of correcting a word in the sentence after you've completed that sentence. (People on IRC tend to type faster than they should, so this kind of problem arises pretty frequently.) The solution is to repeat the error word, then erase it with si, then give the correction. Strictly speaking, that's not how si is meant to work — it only makes sense to a computer parser if the erasure is within the current sentence; but you'll see this on IRC fairly often.

Exercise 3

Apply the required erasures to the following Lojban sentences.

  1. .i mi viska le si la djan.

  2. .i mi viska la djan. si si si catlu la djan.

  3. .i mi viska la djan. sa catlu

  4. .i lenu lebna loi lojbo valsi cu nandu sa nu vimcu loi lojbo valsi lo jufra cu nandu

  5. .i mi .e lemi pendo cu zvati le barja sa .e la ranjit. cu zvati le barja ca lenu do zvati le gusta