(A page from the Loglan web site.)

(From Lognet 93/2. Used with the permission of The Loglan Institute, Inc.)

Sau La Lodtua
(From the Logic-Worker = Logician)

by M. Randall Holmes

My apologies for not providing a column in the last issue; contrary to popular belief, professors are busy people!

I'm still not ready with anything very concrete this time, so I thought I'd give some brief remarks on my opinion of the significance of Loglan and its present condition from my position as a mathematical logician (and amateur philosopher, you will find). These are largely inspired by a recent dialogue I had with Alan Gaynor.

Alan Gaynor's thesis is that Loglan is a new and revolutionary "implement of reason." I won't go into the details of his argument; but this claim forced me to think out my own position on this subject yet again.

I disagree with Gaynor in the following sense: I do not think that Loglan is capable of expressing essentially more than a natural language like English. This means, obviously, that I expect negative results from tests of a strong version of the Sapir-Whorf hypothesis using Loglan! While Loglan might facilitate certain kinds of thinking to some degree, I do not believe that it implements any basically new kind of thinking, or even improves existing kinds of thinking to the point where qualitative improvement in thought can be expected.

One of my sources for evidence for this claim is Quine's book Word and Object, in which he gives an analysis of the logical structure of natural language (using English, of course) which looks like an engineering blueprint for Loglan! This seems to indicate (as I will claim explicitly below) that, far from being basically different from existing natural languages (at least some of them), Loglan is a refinement of natural language.

Another source for my doubt is my experience as a mathematical logician. Loglan implements first-order logic fairly well, including quantifiers, although there is a major design flaw (now irreversible) in the choice of SVO order over a verb-initial or final order, which gives the subject an unwarranted special role, compounded by the breakdown of SVO into S{VO}; the sutori arguments should be attached to the subject if to anything, and certainly not to the verb! But this follows natural language (at least English and other Indo-European languages!) faithfully.

The next step (which, as I will explain, is the last step, in my opinion) is the expression of higher-order logic, or, equivalently, of type theory or set theory. Loglan is more capable in this direction than I originally thought, but not ideal. It does have the ability to construct sets by abstraction (the set of all x such that...) and it has just been augmented with finite sets and sequences (I think), something the need for which was not originally anticipated. However, these features were implemented as afterthoughts. Loglan is at the point where it is no longer possible to add major new features; it is running out of little words!

The excellent implementation of first-order logic is not a major improvement over natural language; natural languages (at least some of them) have had reasonable capabilities (at least potentially) in the area of propositional connectives and quantifiers since the point in (possibly recent) prehistory when relative clauses were perfected. An anthropological question: do all existing languages have relative clause constructions adequate to first order logic? If not, we do share the world with people whose logical capabilities are essentially hampered by their language. Loglan is greatly aided by the fact that its designers explicitly understood first-order logic, which no one did at the time that natural languages were developing these features. I believe that this was the most recent Whorfian revolution, but it occurred (for us, at least) in recent prehistory when the implicit capabilities were implemented in natural language, not when propositional and first-order logic were explicitly laid out.

The capabilities to handle type theory or set theory in natural language seem to be more limited; also, the semantics of these areas are much less well-understood (on an implicit level) by ordinary speakers of natural languages. But Loglan was not engineered from the outset to handle this area, either; as I said, its features for handling sets and sequences are "afterthoughts." I don't think that Loglan's capabilities in this area will cause marked Whorfian effects. Actually, my prediction is that these features of the language will be widely misused!

I think that the level of type theory or set theory may provide the platform for an engineered language which will cause Whorfian effects, but that in any event either the latest Whorfian revolution (relative clauses) or this hypothetical one in the near future will be the last one. Quine, in Word and Object, gives some grounds for the belief that natural language already supports this level; he handles class abstraction (the construction of sets) using relative clauses. But issues related to sets or classes (to a greater extent than issues related to first-order logic) have caused logical and philosophical errors on the part of great thinkers right up to modern times; an engineered language which encouraged the correct use of these concepts might indeed cause Whorfian effects.

It is also possible that the clarity and unambiguity of Loglan really will make it possible to learn to think on this hypothetical next level, and that the ambiguity of natural language will be found to confuse the issue enough that their theoretical ability to implement this level is too remote to realize in practice for the ordinary speaker.

My experience as a mathematician leads me to believe that in any event either the last one or the next one will be the absolute last. The reason for this is that we see in mathematics, beyond the level of Russell's theory of types, a hierarchy of stronger and stronger theories to which we cannot find an upper bound (due to Godel's results), but all of which are expressed in the same language (the language of set theory). There is a hierarchy of strength of theories which goes on as far as we can "see," but the strength comes from the adoption of new axioms, not the adoption of a more expressive language. We do study more powerful languages...ones we cannot speak because their sentences are allowed to be infinite!

Nonetheless, in spite of the pessimistic remarks above, I think that Loglan is potentially important. It is an engineered language, designed on scientific principles; it is the first language of this kind designed to be spoken rather than used as technical notation embedded in natural language discourse. Its lack of ambiguity may make it possible for more people to explicitly understand and implicitly use formal reasoning on the level of predicate logic (since natural language allows propositional connectives to be confused with the mixture operation and quantifiers to be confused with mass term formation, for example, although I wonder how much this occurs in practice); it might turn out that it will help more people over the hurdle set by the last Whorfian revolution (in which case, once again, my pessimism above needs to be qualified!).

Loglan's importance lies in the fact that it is the first deliberately engineered language; I believe that others will follow. Loglan seems to have arrived at a position where it is an adequate language but does not admit much in the way of extensions. This is not a call to re-engineer Loglan! As everyone knows, if we don't stop re-engineering, no one will learn to speak the language, and it will be stillborn. Once Loglan is actually spoken and used, the experience of linguistic engineering can be re-applied to designing further languages. There are advantages of engineered languages independent of the Sapir-Whorf hypothesis; they should be easy to learn, usable as a computer interface, and free from ambiguities and the resulting confusions.

Copyright 1993 by The Loglan Institute, Inc. All rights reserved.

Send comments and corrections to:

djeimz AT megaseattle DOT com