(From Lognet 95/1. Used with the permission of The Loglan Institute, Inc.)

Lo Lerci


Letters policy: Unless otherwise stated, letters addressed to logli in general, to The Institute, JCB, or any editor of Lognet will be considered as offered for publication. But it would be good if the writer explicitly offers. We reserve the right to edit letters, mostly just to drop material that has to do with ordering books, etc. Sometimes a given correspondent will have several letters in the hopper, so to speak, and we will combine them into one purely for the sake of clarity. If you are not on e-mail and your letter is a long one, we’d be  grateful if you’d enclose a soft copy on a diskette. We can translate most word-processors into the one we use and having your letters on disk could save us a lot of typing.

The first letter comes from Roy Bigelow, a returning member—returned last year, as I recall—who’d been a student of the ’75 language and who has just made a thorough study of the current version. These are his comments on it.—JCB

Dear Jim,

 I’ve completed my review of the Loglan grammar documentation you’ve so generously sent me. And, as a result, I was able to recreate my Loglan “manual” and cannot resist sending you my unsolicited, and perhaps jejune, comments.

 I know it’s old hat to you, but I applaud all the new developments. With the old documentation, I noted both an inefficient vocabulary and a syntax adequate for discourse but inadequate for prose. Now there’s the whole idea of affixes and cases and so much more elegance and detail in the grammar. Solves so many difficulties. Marvelous!

 I’m so happy to see the developments with respect to interrogatives. When I did my initial version of my syntax, I came up with a formulation which depended on Ie being a punctuation mark which was very complicated. Your [new scheme] is simpler even if it does require additions to the vocabulary. However, I note that rhetorical interrogatives (No? Right? Don’t you know?) remain to be described and neither he, ha, nor ho appear in the grammar nor are properly processed by LIP. [Really, ue! I think you’d better send your LIP back to us for replacing; because mine, Version 2.23 with Grammar #80 in it, handles all three of these interrogatives very well...or at least that’s what a quick check suggests. (I have not, I confessed, thoroughly pursued them into all the nooks and crannies where they appear; but I expect Dr. McIvor, our takrultua (grammarian) now will. You should let him know the circumstances in which you have found he/ha/ho not properly handled.) As for translating rhetorical English No?  Right? and Don’t you know?, I think the same variety of forms exists potentially in L. Try Ei? No ei? Toi dreti ei?or Toi tradu ei? and Ei tu djano (tio)?—JCB] 

 I am happy to see the addition of the I-connectives. They surely are a help in structuring the documents I do. [But] I would like to have a word marking the termination of a group of sentences begun by the sentence separator Ige. Perhaps Igue? The argument is that, in a written document: I is the mark of a new sentence. A group of sentences begun by Ige can be understood to be a sub-paragraph. Subsequent Iges could be understood to mark sub-sub-paragraphs and sub-sub-sub paragraphs, etc. All we need then is a word marking a return to a main paragraph and gue is now that terminator in other textual contexts. [Sounds good to me. But would you make that same argument formally to the Keugru? And include samples of the usages it would implement with either more elegance or more clarity than our present sentence-handling apparatus permits. The K is always pleased to get  proposals from members-at-large...if only to jolt us out of our normal rut of receiving proposals only from ourselves, uu. Address any of the Permanent Members with your proposal. That’s Kirk Sattley, Bob McIvor, or myself . (This goes for any other member-at-large who may be listening in, and who also has a proposal cooking on m’s backburner!)] 

 It bothers me that the language makes no distinction between a logical not (no when it means It is not the case that...) and no [or non-] (no when it appears elsewhere). Also I wish there were an affix with the meaning un-, non-, or a-. Is that being too fine-grained? [The first reason we use the phoneme-string /no/ for all the negatives in the language is that we can. That is, all no’s grammatical roles—and there are many (I used to know the exact number, but it’s slipped me)—are contextually distinguishable. So not using different phonemic forms for the various negatives keeps the number of lexemes down. (Actually, there is one variety of the L negative that cannot be contextually distinguished from the others, and that’s -noi, the bound morph that is used in compound connectives (like anoi) and tense operators (like pacenoina). This is because we wish to distinguish between the bound and free forms in places where both may occur, as in the two expressions da, anoi de and da, a no de (X if Y and X and/or not Y). The second reason is that all of L’s many nos are in fact signs of the same logical operation: negation. So all of them can in principle be eliminated in favor of the sentence-negating one that we translate into E by the cumbersome phrase It is not the case that... . In other words, the sameness of the phonemic expressions of L’s many nos reminds the logli of an important logical fact. Their sameness is part, in short, of the Whorfian “logic-facilitating” apparatus of the language. As for L prefixes for E non-, un-, and a-, they exist; they are nor-, nur/fur/jur- and nor- again.]  

 Also on my wish list, since I have noted a great many other language embellishments, is [there] a suitable word, perhaps a redefinition of an existing word, that can be used in the context of formula descriptions as a substitute for the English where: which introduces a list of variables and their definitions. [I have in the past used leading Eu... (Let us suppose that...) for such uses; but the whole issue of hypotheticals, suppositionals, and counterfactual conditionals is currently under review in the Keugru. So who knows what usages these sages will finally come up with, soi crano! But you can be sure that as soon as they have these new usages sorted out, they will let us logli know what they are. RAM will take over much of the answering at this point.—JCB]

 In going through the Grammar, I find that it is strange that DIO appears in rule <arg6> rather than the rule <argument>. Is there a reason?  [This rule was “grandfathered” to me when I took over, and usually I don’t change what ain’t broke.  However, the usual reason for this is because the lower numbered args need to be combinable in other ways introduced in higher numbered args.  This is usually necessary to avoid ambiguity.  It seems to me that one might wish to combine DIO args with A, AGE, ACI, etc., which come higher in the chain, and that is responsible for their placement. —RAM]

 However, my biggest demurral are the discussions in Loglan 1 and NB3 concerning punctuation. They are opaque to the extent that I was forced to construct my own rationale. When finished, I discovered two things: First, that the language has an incomplete system of “alogical” connectives, the kind with a lower “binding” precedence than the logical connectives but which are needed to construct sentences like Jack, Jane, Joe, walked, ran, flew, from Paris, Rome, London to Vienna, Moscow, Cairo, [respectively] and their many variations and extensions.  [I personally think these are best handled by ze-forms, or other forms of list-designation (the new lou ... luo, for example; LN 95/1:15), but the matter is still under review. It’s likely that Randall Holmes’ article on “The Logic of ‘Respectively’” (LN 94/-3:24ff) hadn’t  appeared at the time you wrote this question; so it may have been at least partly answered for you by that article. But I agree: this is a still-unsettled part of the language, and the people in the Keugru are even now working hard to settle it.—JCB] Second, if we eliminate its use as an[other kind of] alogical connective, there are only two uses for gu; [one is] to introduce parts of sentences common to the preceding part and [the other] as a substitute for a missing argument in a predicate modifier. [I am only aware of one use for gu, i.e., as a general right-closure when its omission would lead to an incorrect interpretation by the parser.  It is redundant, in that more specific closures are available, but needed in certain circumstances (e.g., where the alternative would be a double pause).—RAM]  

I note that a great deal of simplicity and a complete system of alogical connectives may be obtained by defining CI to be an alogical comma-like connective, eliminating the ICI and ACI Lexemes (for which I see no real use) and moving CI from the CA Lexeme to the A Lexeme. I realize that the price to be paid, however, is a massive revision to the grammar. [CI has a very specific usage; to join two words so that they will group as one, roughly equivalent to a hyphen in English.  ICI and ACI have rare but useful roles, to achieve groupings which would be difficult to achieve otherwise. —RAM]

 A comment. Having gone through that exercise and having reviewed my copy of Lognet 94/3, I find that, in the example given for the unordered set, the set descriptor and terminator are, except for convenience, superfluous. Since I don’t have a copy of the referenced Lognet 92/3, perhaps it contains a logical justification for their use. [I believe it does; it’s still in print, so you could order it (for $3) if you wanted to add it to your collection of Lognets.—JCB]

 As updates to the documentation:

 I note that definitions of the words cao and siu are not in the dictionary. [Cao and siu are in my copy of LOD. Does ‘the dictionary’ refer to the 1975 edition? It’s true; they’re not in that one, but they are in LOD. Do you have LOD? If you do, you apparently need an update. (Let me remind all our readers that updates of all the software you can send us in a single package—that is, if we can do it all in one sitting—is only $5.)]

 It appears in the grammar, so I am curious. What does or did nufuju and kin mean? [Back in ’76 and ’77 the formalists amongst us tried to show that all possible orders of arguments of n-place predicates, for n fairly large, could be accommodated by these “compound converses”; but to the best of my knowledge these compounds have never been actually used in speech or text...except of course in contrived examples. Even the simplest of them are quite impossible to decipher on the fly.—JCB]

 Caa, cae, and cau appear as allolexes of token A4 [in the grammar]. I note there are no dictionary definitions for caa and cae, and that cau is defined [as] a case indicator. [My copy of Grammar 80 doesn't show this; what grammar do you have? It looks like you’ve found a “fleeting error”, one that has already been removed. But cau definitely exists and is a case tag meaning by/for before Quantities, Amounts, and Values, and is so-listed in LOD. *Caa and *cae are still unassigned as far as I know.—JCB][Caa and cae were being considered at one time, and were temporarily included in a test grammar. They were mistakenly not removed when the proposed use was discarded for another approach. —RAM] 

 I note that zea appears as an allolex of Lexeme A1. What does it mean? [Both Bob and I remember that someone had decided early on during our 1978-82 machine grammar work, that zea--which was in the 1975 language and was there used to mix predas within predicate strings (see L1, 3rd Edn, p.75)--was “grammatically unnecessary”, presumably believing that ze could perform all its duties. This was a mistake, it can’t; but on the basis of it zea remained for some time unused. Then in 1993, Bob and I rediscovered that we needed zea; so we reintroduced it. However this time we made zea a mixer of whole predicate strings, leaving ze to mix individual predas within a string as well as whole arguments. (To see the current difference between ze and zea in predicate strings, ask LIP to parse Da preda prede zea predi predo and Da preda prede ze predi predo. You will see that ze now mixes predicate words and zea, predicate strings.) This is probably a mistake...an “analogical error” that languages do not tolerate well. If ze mixes whole arguments, as it still does, then it should probably also mix whole predicates, as it now doesn’t. Probably it should. In any case, we’ll discuss this little problem in the Keugru, decide whether it needs fixing, and report back.—JCB/RAM]

 I note that nue appears as an allolex of Lexeme DIE. Are you sure that it is of the same type? [Yes; both are “status markers”—sometimes called “register markers”—built on the Japanese model; see SLK of LN 90/1:5. Die (pronounced [DEE-eh]) means Dear/Darling, as in Djenis, die = Jenny, Darling, and is derived from dipri (precious) while nue [NOO-eh], which is derived from nutra (neutral), means Mr./Mrs./Miss/Sir/Madam as in Smiq, nue = (Mr.) Smith, Sir. Notice that these markers are always postfixed to their operands, which are always some designation of the person whose status relative to the speaker is being “registered”.—JCB] Am I wrong in translating it as I don’t care? [Yes.] As a matter of fact, why are ... the allolexes of [Lexeme] DIE not allolexes of [Lexeme] HOI? [One is postfixed, the other prefixed to its operand; but more important, hoi creates names out of predicates: Hoi Ditca! = Oh, Teacher!; die doesn’t do this.—JCB]

 In the list of discursives in the grammar, I note that a word for “above all/primarily” got lost [try nefi = firstly] and the dictionary definition of the word cao does not list it as being a UI word. [It isn’t; it’s a “disappearing” word, a metalinguistic word, a word essentially outside the language: one that is removed from the string before it is handed to the parser.]

 In the list of case tags, I note that jui has been substituted for mao. [Right. At my request the Keugru replaced mao with jui (derived from jupni = young) because of the very strong association of mao with madzo. Jui seems to work quite well as a tag for “lesser than” arguments...if one uses tags at all, of course, as I tend not to, soi crano.—JCB]  

 The grammar does not consider the word nui. [As far as I know, the string /nui/ is unassigned. Where did you find it?—JCB]

 I think the grammar rule for <headterms> should be rewritten to distinguish between GI and GIO. It makes no sense for sentence qualifiers to follow inverted arguments. [I wonder if the author, who presumably intended goi, is distinguishing grammar from semantics. Grammatically, gi and goi are used in the same way, though their meanings are quite different.—RAM]

 I read with interest your discussion, in the section on Lexeme NI in NB3, concerning the difference between nei when used to designate a quantity and [when used as] a letter[-variable]. Instead of prefixing ne [to] nei, I would have prefixed ... lio. Even in English we refer to a letter appearing in a formula by the words the variable.... [We’ve elaborated this system a little since NB3, which was written in 1987. In particular, our lodtua’s quite recent suggestion—adopted by the Keugru and reported in LN 94/3:16-17—to use me X to mean is one of the current designata of X (thus mela Djan now means is one of the Johns), resolves a morphological ambiguity that had been lurking in the language for some time. For example, /SEnei/ as intended to be heard as the phrase se nei (seven instances of the things currently designated by n, say seven of the children where nei has replaced le nirboi), was indistinguishable from      /SEnei/ when meant to be heard as the single word senei, an expression writable as 7n and the short form of se nirne (seven years), a dimensioned number. Such numbers are used as quantifiers in L; that is, senei (or 7n) means seven years of (some interval), as in Senei livkeo je la Djan = Seven years of John’s life. With this new definition of me, /SEnei/ now uniquely resolves as senei, which is still a quantifier. The other sense of former /SEnei/, which once could have meant seven of the children, is now obtained with /seME-nei/, an expression which now resolves uniquely as se menei = 7 of the n-things, for example, children. Your suggestion of using lio nei for the number n is, of course, correct if we want to designate not a quantifier but a number. Recall that numbers are not used as quantifiers in L, only as arguments. Thus Eu lio nei, konte = Let the-number n be an integer, illustrates the use of lio in forming designations. (The rest of the editorial responses are mine.)—JCB] 

 Speaking of ... Lexeme NI, I noted that NB3 states that the set of compounding rules for NI-compounds has not been written. So I worked out a syntax (which I would be happy to share if you like) for the arithmetic operations that have been defined. It was simple. I just reformulated the rules defined for my trusty spread-sheet. [We would very much like to see your  rules for generating the allolexes of NI. Over the years, several sets have been suggested, but none so far have met the requirements of speech...which, are of course, quite different (being linear and unidirectional) from those that govern blackboard work. The Keugru would be very pleased to look at your rule-set.] Now, as a result of that exercise:

 I would argue that the grammar, if it permitted gaps between the individual words, is complete and the vocabulary is incomplete. [It permits but does not insist on gaps between words.] The syntax I worked out, since it is based only on the commonly accepted mathematical formulations and has no logical content, I deem to be synthetic. What is needed is a symbol indicating the beginning of a formula, analogous to the symbol lio for numbers, [and the symbol] gio for symbolic logic predicates, together with spoken symbols which denote mathematical symbols and functions [as well as] the way they appear on the printed page (pretty printing). Ask any mathematician and he will tell you that it is the symbols on [two-dimensional surfaces] that evoke the appropriate associations and that’s the reason they can’t talk without a blackboard. [That’s also the reason why we need a set of speech-friendly rules—limited as their use might well turn out to be, given the limited capacity of human brains—for actually speaking mathematics. I have in mind such “linearizing” E usages as a plus b the quantity squared, which we easily understand is to be written as (a + b)2, while the utterance a plus b squared succeeds in getting a + b2 written down with equal confidence. There is probably a limit to how much blackboard intricacy can be conveyed in this way, but a successful L “MEX” (mathematical expressions) grammar will take us pretty close to that theoretical—obviously a neurological—limit. I am personally glad to see this topic revived by Member Bigelow. It’s been some years, now, since we put the “MEX Poject” aside for what was then more urgent loglandical business. But it may be time to take it up again.]  

I cannot find a reasonable word for superscript. [How about “high-character(s)” vs. “low-character(s)” as our metaphors? These give the very pretty pair ganlea/damlea. (Letra, by the way,which is the source of -lea, is already defined as a letter/character/symbol in alphabet...; so super- and subscripted numerals are accommodated as well as letterals.)] 

 I find that nea and pea are unnecessary synonyms for nio and pio. [But we need to be able to say nea ne (-1, or negative one) without implying that there is some unmentioned minuend from which 1 is being subtracted, as the incomplete expression nio ne (- 1, or less one) would presumably do.]

 I find that any syntax which is formulated will require the use of bi and kin and thus preclude the use of bi as the logical identity operator. I note that the only change that is essential is a word for the symbol =, one to be used only in mathematical contexts. [This may be true, ue, but I hadn’t thought so. I thought that the context would make clear whether we were talking about mathematical or logical identity. Besides, aren’t the two fundamentally the same? Isn’t a “just another name for” b2 when the equation a = b2 obtains? How does this differ from John is (identical to) the man who came to dinner? Both claim that a single designatum has (at least these two) “correct” designations. The canons of correctness differ, of course, but surely that’s a minor point.]

 Finally, as if the foregoing were not enough, I must tell you a story. Once, long ago, I wanted to write about real estate transactions in Loglan. The dictionary had no word for agent. I considered defining one but, because I felt I couldn’t do a good job of it, rejected that approach. The next try was to adapt a dictionary word. I tried using the nu-fu-ju maneuver. I unsatisfactorily tried the associated words prati, [nurprati], vedma, furvea, [vemtoi], [nurvea], batmi, penti, cmeni, and all the words ... derived from them. My solution ... , since it came closest to denoting the ideas I wanted to write about, [was to]:

[1] Create a new definition for the word vedma, which had most of the arguments I was looking for. I came up with [X (a seller) engages] in a monetary transaction involving Y (a commodity), W (a buyer), H (a monetary consideration), and Q (an agent) [But agents are not always involved in selling!];

[2] Invent a case word-series, /xV/, since there wasn’t one and /x/ was a forbidden consonant; and, finally;

[3] Use Le xu vedma je le xi vedma to designate the buyer’s agent and Le xu vedma je le xa vedma to designate the seller’s agent. Clumsy and unsatisfying, but for me it evoked the needed associations.

 Since it gave me so much trouble, I reviewed that exercise in light of the new materials I have received. I note that things have greatly improved. The dictionary [now] has a word for agent. [Actually, it did in 1975 as well, namely the word dilri = X represents Y in situation Z; but uu, the E part of the definition wasn’t well-developed then, so while represent appeared as a verb on the E-side, the noun forms agent and representative did not.  Uuue, we put you to a lot of trouble! Had you found dilri, two words you could easily have made from it would have been vemdilri for seller’s agent and furvemdilri for buyer’s agent. Or you could have turned these metaphors around and made dilvea, a “representing-seller”, and dilfurvea, a “representing buyer”. This is an ok thing to do in this case only because both orders produce meaningful and relevant metaphors. But reversing the terms of a good metaphor isn’t always—or even usually—possible.] However, if the same situation occurs [again, I note that] we have the zua/-e/-i/-o/-u series. We have nufe. We have the option of inverting the order of the terms [in possessives] and using Le le [preda] gu [prede] in place of Le [prede pe] le [preda],which somehow seems better. Best of all, by considering the zu[V] series of words to be affixes [ah, but we’re not allowed to do that!], I can get their compounds with any other predicate word past LIP. So now I can use Le zuivedma zuuvedma for the buyer’s agent and Le zuavedma zuuvedma for the seller’s agent. [You’re getting it past LIP, alright, but that’s because LIP doesn’t check the well-formedness of your words. It sees these things as PREDAs because they (1) are vowel-final and (2) contain at least one pair of adjacent consonants; but even so, they’re not well-formed. For one thing, the Resolver (when we get one built, ae) will hear the string /lezuiVEDmazu,uVEDma/as Le zui vedma zuu vedma; and this string of lexemes, you would find, does not get past LIP. Won’t you settle for dilvea and dilfurvea, ae?] 

 Being lazy, I cannot fail to note that, if I learn a predicate word together with its arguments and their order, I have a mechanism, albeit pedantic and boring, for inventing synonyms and thus avoiding the need to learn the associated vocabulary. In fact, where the predicate word satisfactorily denotes a context, use of the [case] mechanism rather than [learning new] vocabulary may be preferable.

Thanks for your attention.


Roy V. Bigelow

The next letter is from a British logli, Alex Leith, now living in the U.S.  Alex is the first logli in a long time—perhaps ever—who’s come on board as a Patron with his very first dues! (Nu garti, Aleks!) Alex is now doing a semantical analysis of the whole Loglan vocabulary, using a program called WordNet that uses Hypercard...a necessary part of that long-talked-about “man-machine interface,” I surmise. His letter was first e-mailed to our Loglanists List.

Hoi Logli:

Dr Brown suggested I submit some details. ...

[I was] born in London in 1935, not within the sound of Bow Bells (which is the criterion for a true Cockney), but certainly within the sound of the BBC time signal.

I designed my first language at the age of about 8. The structure was English, with an arbitrarily substituted vocabulary.  I remember only one thing about it: the word for ‘dog’ was ‘hinter’. I was mightily miffed  when on learning some German I found they had stolen my word to  mean ‘behind’. German I learned from a prisoner of war who was working (1944 or -5 this must have been) on the farm in Scotland where I was  evacuated to.  Nobody else would speak to him, but he let me drive the tractor, so ... 

So inevitably at school I “did” languages.  French, German, a smear of Spanish, some Russian. And (though my housemaster told me to stop it and concentrate on my Higher Certificate preparation), I started on Hungarian...from Colloquial Hungarian [by] Arthur H. Whitney; Routledge, Kegan Paul, 1944; 2nd ed 1950. I can tell you those details ’cos I’ve got the book right here, after 40 years.  I think I chose Hungarian because it was the most outlandish language I could find study materials for.  [I] now go there regularly each summer.

Actually, there is what I regard as a family myth, that the Leiths, sometime in the Middle Ages, arrived in Scotland from Hungary. No more details have ever been suggested, but right on the Austro-Hungary border, south of Wien, there is a river called the Leitha. That part of Burgenland was, I suppose, in Hungary before 1918 and Trianon.

Then National Service (2 years compulsory when I was 18). [I] was put on the Russian course. Nine months intensive. Total immersion, really, since the “kursanti” very soon began talking a sort of bastard Russian among ourselves.  (Clockwork Orange must have drawn on this experience.) 

At University I read what in Cambridge they delightfully call “Moral Sciences”.  That is Philosophy, Formal Logic, and Psychology. Philosophy, in Cambridge in the late 50’s, was Logical Positivism. A good emetic.   “My dear sir, what you are asking is not a well formed formula in the language you are using.”  “OK,” I’d think to myself, “but are there Other Minds?” Formal logic, Boolean algebra, Propositional Calculus, has actually served me well later.  

[I’ve] worked in the language-teaching field on and off quite a bit.  Teaching English is a good way to support oneself while travelling.  Greece, Japan. [I] was more interested in designing course materials than in classroom teaching.

In 1975 I returned to Cambridge to do a concentrated year’s Computer Science, and that was home!!!  I’ve been addicted ever since.  

[I’m] now “retired”, and following my own weird in the scope of computers, graphics and languages...  Which is where Loglan is particularly interesting. If we are going to have a computer that understands us (“Hal”), it is going to have to know about the world?  So how do we tell it, and where do we start?

I’d be interested to hear more about the Loglan computer side.

Alex Leith

The next letter is from a member who has been continuously with us since the second time we “went public”,  in 1975. He is Jerome Frazee, who writes to us from a wheelchair, his sister doing the actual typing for him, and who, though he protests that he is “not a logician,” nevetheless publishes articles on logical notations in academic journals. Meet Jerome.

Dear Jim:

Sorry I have not written for so long. I see in Lognet 94/3:7 that Emerson Mitchell and Randall Holmes have been discussing set theory as it might pertain to Loglan. I do not consider myself a logician (as you suggested [I was] in LN 94/2:27) but I want to say that for a long time I have thought that the logic of a logical language should be the algebra of sets. You have evidently noticed some difficulty with the logic of Loglan. [Primarily with packing it all in, soi crano.] I do not know what Mitchell and Holmes have said but in my opinion the algebra of sets more closely fits the natural way we think and is simpler than symblic logic with its quantifiers. (Of course it would have to have the propositional calculus too, but that is not hard. Although most people do think that the if...then... of English is somehow different. It isn’t.) I took it for granted that [set theory] could play no part in Loglan today but now I have decided to say something about it.

With sets all you are really doing is thinking about circles (Venn diagrams) and most people do not find that difficult. I only see two possible problems. (1) Sometimes there is only one circle or even none (the null set in both cases) which might cause some a little trouble until they get used to a circle that is not there. (2) In my opinion, Russell’s paradox enters into the algebra of sets in a way that it doesn’t enter into symbolic logic. You can create the paradox in symbolic logic if you want to, but it doesn’t seem like something is missing if you don’t. (I have finally decided that there is an acceptable—but intellectually unsatisfying—way around this.) In any case it would never make something that is false true or vice versa. Only philosophers sounding the depths of logic could care.

Finally, in your letter [to me] of 18 Nov 93 you said “Do you think History and Philosophy of Logic would give us permission to print this article in La Logli just as it is?” [Jerome had published in that journal an article proposing an elegant new notation for the algebra of sets that I thought would interest logli.] I have tried four time, without success, to get an answer to that question about my article on sets. The first two times they didn’t answer and the third time they told me to contact Dawson Holdings, who handle all their reprints. So I wrote them and they have not answered either...even though I offered to buy some reprints, too! You probably know the law about these things better than I do. [Wes Parsons does, and he’ll see the implied question in this column and no doubt answer it for us.] I would like La Logli to print it alright and am willing to rewrite it (as you suggested) for LL, but I couldn’t do it ... until I finish a couple of other things first.

Hope this finds you and all yours doing well.

Sincerely yours,

Jerome Frazee

The next letter is from Paul Doudna and is a continuation of a letter published in this column in LN 94/3.

Dear Jim Brown:

This is a follow-up of my letter of Feb. 9. I was particularly concerned about the way my analysis of interesting little red school (LN 94/3:10) and little house (p.13) was being interpreted. Let me give a more precise analysis.

First of all, to speak of “a house”, I represent this by “x / house”, which may be read as “something which is a house”. The relation “/” relates a particular “x” to a universal “house”. This is a descriptive phrase, not an assertion that x is a house. It only implies that x may be identified because it has previously been asserted (implicitly or explicitly) that x is a house. The precise meaning of the word “house” is relative to the larger verbal and non-verbal context. That is, a “house” in the suburbs of St. Louis is different from a “house” on the banks of the Amazon River.

Having specified that we are referring to something which is a house, we may wish to further specify that this something is little, that is, “a little house”. I represent this by “(x / house) / little”. This may be read as “x which is a house, which is little”. The first “/” relates a particular “x” to a universal “house” as before. The second “/” relates a particular “x / house” to the universal “little”. Just as the precise meaning of “house” will shift according to the context, so will the precise meaning of “little”. This relativity is what you paraphrase as “little for a house”. I understand what you are saying here and I am not in disagreement with it. I think you are jumping to conclusions to assume that I do not recognize this relativity .

Let me clear up some possible confusion. “(x / house) / little” is precisely equivalent to “x / house / little”. The parentheses are redundant. They are redundant because (in my notational system) “/” may only relate a particular on the left to a universal on the right. *”x / (house / little)” is meaningless (is not a wff, as some might say). Normal English grammar teaches that “little” modifies “house”. But I think it would be more accurate to say that “house” and “little” both modify the same thing. The critical point is that “little” modifies something (“x”) only after it has previously been modified by “house”. Thus the relation between “house” and “little” is not symmetrical. Since it is not symmetrical, it cannot be replaced by the logical AND of class intersection.

If we were to assume that the universals “house” and “little” represented well-defined classes, “a little house” could be represented as “x / (little & house)”, which is equivalent to “x / (house & little)” since “&” is a logical AND and thus a symmetrical relation. We would have to conclude that “a little house” was equivalent to the “a house little(-thing)” (whatever that is). The formula “x / (B & A)” as a representation of an English content word pair “A B” will only work in certain restricted cases. I think we both are well aware that most content words in English represent fuzzy classes, not the well-defined classes of formal logic.

Here is another possibly confusing point. A little house is something which is a house and is something which is little. Conversely, something which is a house and which is little, is a little house. All of which may seem to contradict what I have just said in the preceding paragraph. The problem is that the “and” of English, in most contexts, in not strictly a symmetrical relation. An English “and” does not therefore translate into either the truth functional AND of propositional logic nor the logical intersection AND of class logic. I am sure you are well-aware of this point.

Similar problems exist for OR, IF, and NOT; but that is a whole other subject.

Suppose now that we wish to change the specification of “a house” by saying it is made of brick rather than by saying it is little. We are now talking about “a brick house”. Some grammar books refer to “brick” as a noun adjunct and explain that by saying it is a noun used as an adjective. The problem is that a noun modifying another noun does not work at all like an adjective modifying a noun.

In particular, my simple formula “(x / B) / A” does not work for “brick house”. The first part “(x / B)” is correct but the appended “/ A” makes no sense since a brick house is not a brick. I use a different formula, “(x / B) R (y / A)” to represent “AB” (where A and B are nouns). That is, we are now talking about one object “x”, which is related by relation “R” to a second object “y”, such that x is a B and Y is an A. In particular, “a brick house” is represented by “(x / house) made-of (y / brick)”, which may be read as “something which is a house, which is made of something else which is a brick”. Most noun+noun modification structures can be interpreted this way, although there are obviously many exceptions.

The thing to be noted is that the formula “(x / B) / A” is not ambiguous, whereas the formula “(x / B) R (y / A)” contains an unspecified relation “R” and so is potentially ambiguous. In practice, context, convention, and common sense may make it clear what we probably mean, but it is still potentially ambiguous. All of which brings us to the point of seeing why English is less ambiguous than Loglan in this respect.

What is represented in Loglan as simply one content word juxtaposed with a second content word, in English can be grammatically differentiated into at least two different cases (actually there are more than two). This grammatical differentiation, which does not exist in Loglan, enables us to separate out a special case which is unambiguous (or at least highly restricted semantically).

You indicated in your remarks (in Lognet 94/3) that “little house” is ambiguous (p. 13), but supplied no examples of what these alternative meanings are. You also indicated that there are a finite number of possible “metaphorical types” for “brick house” (p. 13), but you don’t supply a list. I personally think that “little house” has one meaning whereas “brick house” has potentially an infinite number of meanings. But even if I’m wrong, it would seem to be obvious that adjective+noun modification is more semantically restricted than noun+noun modification. And as long as that is the case, a language which grammatically differentiates these two cases must be less ambiguous than one that combines the two cases without any grammatical distinction. [Surely this much must be granted.—JCB]

I thought, in my previous letters (naively, it seems) that I was simply stating the obvious, but apparently not. Somewhere along the line my point got lost. Of course, if it is accepted as an article of faith—I hope my judgment is not too harsh—that anything that English can do, Loglan can do better, then it really doesn’t matter what I say. [Of course we do not accept “as an article of faith that whatever E can do, L can do better”! When this is true, we try to demonstrate it; and when it’s false, we are pleased to wonder why. And of course it does matter what you say...even as an outsider, as I trust you’re discovering.] 

Following my comments on the distinction between one-object and two-object metaphors (p. 13) you add the observation: “This distinction, though interesting and important, simply does not go far enough toward generating a theory of metaphorical types.” I agree with your observation, more or less; but so what? The distinction is one that occurs in English grammar but not in Loglan grammar. Because the distinction exists, we know that a “little house” does not mean a house containing small furniture, a house built with little rooms, a house designed by a short architect, a house built for midgets, a house located on a little hill, a house with small windows, a house containing miniature portraits, ad infinitum.

In the Loglan counterpart cmalo hasfa, these extraneous meanings cannot be logically eliminated. They may be eliminated by the user if he associates the Loglan words with the corresponding English words, and (unconsciously, perhaps) associates the semantic restrictions of the English with the possible meanings of Loglan. Ask a human Loglanist if a cmalo hasfa is little, [and] he will respond: “Obviously, yes.” Ask a computer and it might respond: “That’s one possibility. All I know is that it is a house and is related in some way to [being] little.”

As to the observation that I have not generated a “complete theory of metaphorical types”, obviously I have not. I only pointed out two possible modification patterns to explain why sometimes English is less ambiguous than Loglan. The two formulas, “(x / B) / A” and “(x / B) R (y / A)”, have wide application but there are many other possibilities.

For example, I remember when the ambiguity of “lion hunter” was discussed years ago. I don’t know if this ambiguity has been resolved or not in the current version of Loglan. [No. It awaits a typologically complete analysis of the varieties of metaphor. But surely we can guess the type of this one: it is of the “served+server”, “target+shooter” type, of  the type generally known as the “patient+actor” type. But you’re right; we know this only by plausibility analysis. There are no “bowels of the language” where we can go to discover what kind of metaphor lion hunter is. Perhaps one day there will be...at least an archive where the type of any metaphor underlying a complex predicate in L may be discovered.] But here are some examples of how deep structure (if you’ll pardon that term) can be used to differentiate related meanings:

1. Lion (which is a) hunter, a hunter which is a lion. 2. Lion (‘s) hunter, a hunter of lions. 3. Lion (which is) hunted, something hunted which is a lion. 4. Lion (‘s) hunted something hunted by lions. 5. lion hunt, a hunt for lions. 6. Lion hunt, a hunt by lions.

1. (x /a hunt) / lion. 2. x /a (hunt .o (y / lion)). 3. (x /o hunt) / lion. 4. x /o (hunt .a (y / lion)). 5. << hunt .o ‘lion >> OR << hunt .o (y / lion) >>. 6. << hunt .a ‘lion >> OR << hunt .a (y / lion) >>.

I will not attempt to explain my syntactic notation here. I only wish to illustrate that there’s a lot more to a “theory of metaphorical types” than what I have written about so far. Also, it should be clear why I regard this as a structural question, even though it does not concern the surface structure. I make no claim to having worked out a complete theory (of anything). ...

Paul Doudna

Reader Doudna makes a very useful point in this letter, namely that natural languages—which all restrict in one way or another the kinds of word-pairs that constitute legitimate metaphors—often offer much stronger clues as to what kinds of metaphors their users intend than are provided by a predicate language like Loglan. This is true, of course, because in a predicate language all predicate pairings are grammatically permissible. Good. Let’s nail that point down. In just the way that Paul Doudna suggests, English adjective+noun constructions permit only a restricted range of metaphorical types to be expressed (but not just one, note), while English noun+noun metaphors admit a much larger range of types. Loglan predicate pairs, in contrast, admit the largest metaphorical range of all...the largest possible range, in fact; for that is part of its deliberate design. So it is doubly important that we logli supply our “metaphor facilitating” language with a mechanism—even if it remains a largely covert one—by which a speaker may, when s chooses, specify the kind of metaphor s has in mind. Presumably every speaker would know what he intended; so making inquiries of our metaphor-using speakers as they spoke would be an option we could choose to use whenever we needed to...whenever, in short, a speaker spoke “mysteriously” to us, perhaps in startlingly new metaphors. Then, with such a mechanism available—the linguistic mechanism itself would be a set of typologically exhaustive  infixes, the null allolex that would make them optional being part of that set, as I’ve described elsewhere—we could make a specification of the sort of metaphor intended by the word-maker a standard part of every dictionary entry for a complex predicate, that is, for every predicate composed of other, simpler words.  And we could use the same mechanism to interrogate our poets when they, as is the wont of poets, invent new metaphors on the fly.

I have the beginnings of such a catalog...not of the infixes that might one day represent them; I have assigned none of those yet; but of the dozen or so distinct types that these infixes would, among others, be required to specify. I have been assembling this catalog for years. I’m 74 this summer; it’s about time I turned the project over to someone else, soi srisu. I would be pleased to do so to any logli who would like to pursue it further, and ultimately come up with that “exhaustive set”—hopefully one that would cover all the word-invention routes used by any language (for that’s what metaphors are)—to which we could then set out to assign CVV-form infixes.  Hopefully we’ll have some left by the time the project is done!