(A page from the Loglan web site.)

(From Lognet 90/3. Used with the permission of The Loglan Institute, Inc.)


By Robert A. McIvor

Originally, gu served as a right-hand closure in almost all cases where one was needed. Gu was all that was necessary in simple cases since the machine grammar was capable of inserting these closures where the machine grammar required them but where human speakers would not normally feel the need to insert them. More precisely, in order to produce a provably unambiguous grammar without unduly multiplying the number of rules needed, rules were inserted that required right-hand closures. However, the grammar was made sufficiently redundant that the computer was capable of inserting missing closures to make machine Loglan provably unambiguous.

While this grammar served well for normal conversational purposes, it soon became evident that in order to make da's meaning plain when more complex structures were afoot, the speaker or writer might have to close off a structure in present time that had been begun far back in the sentence. The only means we then had for performing such distant closures was to speak or write a string of explicit gu's...enough of them to close off all the intervening structures before the final gu in the string would be understood correctly by the parser as effecting the desired distal closure. To use a rather extreme example, suppose one wished to say, without rearrangement, He said that he hated the brother of the teacher that went to France last summer (i.e. the teacher that went) to me (the person addressed). In Loglan, if we had only one kind of comma to effect the closure that is necessary before the mi argument can be properly interpreted, we would have to say Da cutse lepo da tsodi le brudi je le ditca ji godzi la Frans, na le pasko cimra gu gu gu gu gu gu gu gu gu mi. Let us consider this formidable requirement in detail

The first gu replaces the PAUSE that may occur after any descriptive predicate, in this case after le pasko cimra. The second gu replaces a PAUSE that is found at the end of any instance of the grameme <terms>; in this case <terms> is composed of the single modifier beginning with the PA lexeme (na le pasko cimra). The third gu now replaces a guu--which is one of the explicit commas which we have in fact added since the time when our only option was to speak a string of gu's, but any of these new commas may still be shortened to gu if context doesn't warrant their explicitness--which closes the <termset> of the predicate godzi. [The expressions in angle-brackets are grameme names--Ed.] The fourth gu replaces a gui (another of the explicit commas) which closes the ji-clause. The fifth gu replaces a PAUSE that closes an <argument + argmodifier> structure (Le brudi je le ditca), not to be confused with the PAUSE which closes le ditca, nor the gue which closes the je-phrase itself. The sixth gu replaces the gue (yet another explicit comma type) which closes the je-phrase; the seventh, the pause that goes with a descriptive predicate (if modified, the pause follows the modifier), the eighth replaces the guu which terminates the tsodi <termset>, and finally, the ninth replaces the guo (yet another explicit form) that terminates the lepo-clause. At last we can speak the mi in the second place of cutse, and the parser then puts in the implied guu that will close the cutse <termset>.

Obviously, it would be as ridiculous to expect a human speaker correctly to insert all these gu's as to expect a human listener to decode them all on the fly. However, replacing all these gu's with a single guo to terminate the lepo-clause will enable the parser--and we hope also the human listener--to parse the whole sentence correctly. This single explicit comma works in this case because, since the first term in the cutse <termset> was a lepo-clause, a terminator for the lepo-clause will suffice to close everything inside it. Once a speaker begins such a clause, he has only to remember to use a lepo-clause closure for each additional lepo-clause embedded in it. In this instance there were none; so one guo suffices. This maneuver might not always give the simplest solution, but it is probably the easiest to apply in speech.

A similar argument applies to the other closures. There are now four of them: je-clauses are terminated with gue; ji-phrases and -clauses are terminated with gui; lepo-clauses are terminated with guo; and <termset>s are terminated with guu, which has just been added to the explicit comma set. Any of them can be replaced by gu when explicitness is not required.

For the writer equipped with LIP, another approach to learning good punctuation habits is to write the sentence up to the point where closure is required, parse it with LIP, examine the string of implicit closures which will be supplied by the parser, and try out the last of them as an explicit punctuator. If this does not give the parse desired, add more closures as necessary.

Although in the sentence studied here guo gives the most economical punctuation, the comma pairs gue guu and guu guu used in its place would also give the correct parse, as the parser is able to insert what is then the implied guo correctly. Either gue or guu alone causes mi to be interpreted as the second argument of tsodi. This type of closure is difficult to predict without either good knowledge of the grammar, or considerable experience in using the parser.

In conclusion, the simplest approach to correctly punctuating intricate sentences in Loglan is to use a bit of forethought; anticipate when backtracking to a distal antecedent will be necessary, and then use the closure appropriate to the grammatical type of the distal structure to be terminated. For example, in the above sentence, if it was the brother that went to France, the speaker should insert a gue after je le ditca to avoid having to say gu gu gu.

Copyright 1990 by The Loglan Institute, Inc. All rights reserved.

Send comments and corrections to:

djeimz AT megaseattle DOT com