(Originally appeared in Lognet 90/2)

In Defense of Loglan Morphology

by James Cooke Brown

Part 1 — Answer

Once in a blue moon someone produces a critique of Loglan that deserves the most careful answer we can make to it. Our editor Rex May’s article is such a critique. Careful, systematic, friendly to the spirit of Loglan, it examines Loglan morphology point by point and finds it wanting. This article is my answer to that critique. I enjoyed writing it. For one thing, it gave me a chance to explain some things about the early engineering of Loglan that I’ve never had an occasion to discuss in print before. But the real payoff came in some calculations I made at the end in order to illustrate the paper. What they showed me was that even English-as-a-second-language would be less learnable on a world-wide basis than what Rex is calling Classical Loglan would be. This was news to me. My daughter says that it was lying on the surface. ’Of course Loglan is more learnable than English!’ snorts she. I wonder. Had you seen that? Anyway, thanks go to Rex for pushing me in the direction where this result was apparently “lying around”.

Rex’s article was originally titled “Yet Another Morphological Revolution”. I retitled it “A Critique of Loglan Morphology with Suggestions for a New International Language” because that is what it genuinely seems to be. Clearly it is not a serious proposal for yet another morphological revolution in Loglandia, as it throws most of our morphological babies out with the bath. For example, while it would continue to distinguish little words from other morphs, Rex’s proposed morphology would not distinguish predicates from names; its connectives and attitude indicators would not be distinct from other little words; and, and its compound predicates would be distinguishable from its simple ones only by hyphenation, which flies in the face of the Zipf principle (which says that more frequently used expressions, in this case, the compounds, should be shorter than less frequently used ones, their deriving metaphors), or by a very costly system of arhythmic stresses. In addition, Rex’s morphology would not have Loglan’s virtually algorithmic capacity to borrow words, and its borrowings, even when possible, would not be distinct from its other words. His morphology would also not permit the incorporation of natural names. Finally, having abandoned composite derivation altogether, the learnability of Rex’s language would be less than half that of Loglan. All of these fairly pleasant, and at any rate distinctive, features of Loglan—including its impressive leamability—would have to go. So this is not a revolution Rex is proposing but a departure for another country. I’ll devote Part 2 of this paper to demonstrating each of the points I’ve just made. But in Part 1 I’ll consider only Rex’s critique of current Loglan. Some of the points he makes have independent validity and in any case must be answered.

Let’s turn first to the list of Loglan features Rex sees as “flaws”.

Flaw #1. Our proportionality measure of memorability. Rex is apparently unaware of how useful the [eek] of week in likta really is. Few people are. But I studied this matter quite carefully in the late ’50s. I divided the words we had made by then—we must have had four or five hundred at the time I began these studies—into five or six groups on the basis of the proportion of some English clue word each contained: likta’s .67 of week, for example; godzi’s (it was gotso then, but the English fraction is the same) 1.00 of go; marka’s quite different 1.00 of mark; dzoru’s 0 proportion of walk, or of anything else in English that might be associable with walking; barda’s .50 of reward; dilri’s .22 of represent; korce’s .75 of cord; sakta’s .33 of saccharine, yet with no countable part of sugar which is what sakta means; and so on. Sometimes, as in the case of sakta, the mnemonic word was not the same as the key word in its translation but merely associable with it on semantic or historical grounds. This sort of indirection didn’t seem to matter at all. Indeed, I found that the clue word can be considerably more remote than saccharine is from sugar—as remote, for example, as vaccine is from cow for the Englishman learning Spanish vaca—and still serve its memory-jogging function.

Having classified the then-made words in this way, I randomly selected an equal number of them from each of the 4 or 5 “proportionality” classes in which I had sufficient cases, and scrambled them up on test lists of 25 to 50 items. I mimeographed these lists and handed them out to my subjects—students in my research classes at the university, more than likely, as well as to some hapless visitors to my home—reading the chosen list aloud to them, and pointing out the phonemes that were shared—e.g., ‘[LEEK-tah], week, notice the shared [eek]’—if any were. I also pointed out the clue-word/key-word association—e.g., sacharine/sugar—if any was involved. After this single exposure—which thus served as a model of “one-shot learning”—I tested my subjects for either recognition learning (responding week to likta) or recall learning (responding likta to week), using different subjects for the two tasks. Some subjects were tested after short intervals (a few hours) and some after long (a few days or weeks). I ran these little experiments several times and in several patterns, using several different lengths and scramblings of the stimulus lists. It was from these studies that I perfected our methods for scoring trial words, as well as confirming that the “proportionality effect” I had been assuming actually existed. For the results, as I recall, were always at least consistent with the “proportionality hypothesis”, i.e., that the probability of remembering either a new Loglan composite primitive when given its meaning, or its meaning when given the word, is a simple linear function of the proportion of phonemes of some English clue-word that are found in it. Indeed, in some tests the probability seemed to be well-estimated by the proportion itself. The matching phonemes must occur, I found, in the same order in both words if there were three or more of them, and in the same C/V pattern if they were only two; see L1, p.416 for other details. I confess I never investigated this relationship with other than native English-speaking subjects. But such a fundamental relationship is not likely to vary very much with the native language of the learner...provided, of course, that at least some phonemic patterns in the learner’s native language are shared by the sutori language to be learned. I didn’t publish these results for two reasons. One, they’re obvious. They tell us only what any alert learner of foreign languages already knows, namely that da’s vocabulary learning is helped by whatever familiar phonemic patterns da encounters in da’s new language: helped a lot if the similarity is great, a little if it’s little. Two, these studies were routine engineering tests. I did them to check my calculations and to improve the way I made them. No engineering laboratory publishes that sort of routine confirmation of its slide-rule work. Still, now that the job is done, I invite anyone who would like to explore and refine these results—perhaps by putting them in some larger context—to do so. The Loglan primitives now exist in larger numbers and so do their clue words; so the same studies could now be more sensitive. It might be fun to repeat them with new subjects thirty years later, or explore the same Loglan-learning effects in those who speak other languages than English. Write me if you’re interested in exploring this.

So Rex is just wrong about this one. He’s apparently not aware of what is happening in his own head when he learns Loglan words, or why it is they are so easily learned. Few people are. But Loglan has been from first to last an engineered language, not merely a designed one. If I have adopted some strange feature, let’s say that weird 5-letter primitive form that Rex finds so dull, nine times out of ten it was because I had found out, either by formal analysis or by experimental work, and often by both, that it did what I wanted it to. There are exceptions to this, even embarrassing ones. The complex-formation system I went to press with in 1975 is one of them. That was a case where a set of perfectly reasonable assumptions were not tested before publication and turned out to be at least in part false. The failure of non-unique fragments of the primitive roots of complexes to serve as clues to the meanings of those complexes—a failure experienced by myself and all my apprentices in 1977-78—was the biggest but not the only flaw in the 1975 morphology; and it was the one that sparked the Great Morphological Revision of 1978-82. Our experience is described in my 1983 article in TL6/1, “The Great Morphological Revision”, as well as in other places. GMR was a step-by-step engineering project from first to last, as those who endured its “taste tests” with me probably remember only too well. What a bore! But also what a solid language it produced.

Oh, by the way, Rex’s remark that “we quantify things that really can’t be quantified” is, in this case, wide of the mark. Psychologists can quantify animal learning performances, including those of human animals, and have been doing so rather successfully for 100 years. Rex obviously doesn’t trust science, or at least not that kind (although I’ll bet he drives over bridges instead of swimming alongside them!). And that’s his privilege, of course. But most of the people who admire Loglan do so in part because it has been so patiently and, in the main successfully, engineered. That is to say, they admire the way both pure and practical science have been applied to both its design and its construction.

Flaw #2. I’ll try to make short work of this one. The “tradeoffs” Rex speaks of aren’t really. They are simply the result of measuring—again, an engineering result—the diminished returns I was getting for undiminished effort. To show Rex this, let’s consider the extreme case. Suppose we wanted to make Loglan words maximally memorable (recognizable and recallable) to the people of this Earth (as in fact we do want to do, but within the limits imposed by finite resources), and had decided, therefore, to look for overlapping sets of clue words in all the world’s languages. There are about 8,000 currently spoken human languages, by a recent count. We would thus increase the work of word-making by at least three orders of magnitude, and probably more like ten, over the amount of work required to handle 8 target languages. So adopting such a perfectionist goal would probably mean that Loglan would never get done. But then we look at the numbers. We notice that even after such Herculean labors we would not have increased the average memorability of Loglan words, as measured over the world’s population, by more than a few percentage points over what we had already achieved with the first eight languages! This is a remarkable but demonstrable result; and it derives from two circumstances. One, the number of primary and secondary speakers of human languages beyond the seventh or eighth is relatively small compared to the massive numbers we acquire with the first eight. (These comprise about 80% of the planetary population. Remember we are counting all the primary and half the secondary speakers of each language.) Two, if we include languages beyond the first eight, the added languages will be increasingly unlike the major ones that preceded them. (Arabic, for example, was the ninth language but is like none of the preceding eight.) That means that languages beyond the eighth are less and less likely to display phoneme sequences that match those already mobilized, and thus to precipitate the synergically larger-than-expected memorability increases that occur when minor and major languages are related. Consider, for example, the way the “small” languages French and German on the eight-language list interact synergetically with their “larger” relatives Spanish and English. Now notice that Japanese, which is also a small language but sui generis, does not enjoy this synergy...not even with Chinese. So, of the eight target languages, Japanese makes the smallest contribution to Loglan although it is not the smallest language. After eight, more and more languages are like Japanese and fewer and fewer are like French and German. Again, this is something I found out by experiment. I decided not to include the ninth language, Arabic, precisely because I found, by actually remaking a test sample of words with Arabic added, that adding it would have augmented memorability only negligibly while increasing the work of word-making non-negligibly, that is, substantially...at times, it seemed, geometrically. This is a classic instance of diminishing returns, a phenomenon familiar to all design engineers. One does not design bridges to be as strong as they can be for to do so would be to encounter infinite costs. Somewhere long before infinity one stops augmenting even the most desirable properties of the objects one is designing.

I’m afraid it makes no sense to say (as Rex does) that my stopping to augment “neutrality” at the point I did stop, namely before including Arabic—actually, it was not neutrality I was augmenting, but the probability of a human person anywhere on the planet finding something usefully memorable about a Loglan word in some language da already knows...which is an even deeper notion than neutrality—had the effect of “seriously compromising [Loglan’s] neutrality”. One does not compromise the safety of a bridge by stopping doing neglible things to it. One stops doing negligible things to something precisely because doing so does not compromise it.

Rex’s alternative list of target languages is very interesting, but his way of using it is, alas, fatal for his project. But his 28-language list wouldn’t have improved my chances of reaching the word-making goal I had set for the Loglan project any more than an 8,000-language list would. That goal, let us remember, was to increase the average memorability of Loglan words for all the inhabitants of this planet to a point where that average could not be significantly (i.e., non-negligibly) bettered. No other list of target languages does this more than negligibly better than the list I chose did. And Rex’s method of word-making, I’m sorry to say, cuts my learnability result by more than half. But that is something we will take up later, when we examine Rex’s proposed language on its own terms.

Flaw #3. That [ng] is not a Loglan phoneme. But [ng] shouldn’t be a Loglan phoneme for the simple reason that it does not have the privilege of being used initially in syllables in very many widely-spoken languages. Obviously in Loglan, every consonant must have that privilege. Even in the few large languages in which it is phonemic, e.g., Mandarin Chinese and English, for instance, [ng] may not be initial. Thus sing but not *ngiss; chang but not *ngach. In some African languages, /ng/ is a permissible initial, witness the Ngoro-ngoro crater. It also occurs initially in Maori and some Southeast Asian and Siberian languages, but that’s about the size of it. So including it in Loglan would therefore either force some rare oral gymnastics on our international crowd of speakers, or introduce a complication for them by giving them a consonant that can’t be used initially. Neither internationally rare sounds nor distributionally restricted ones are acceptable in an international language. Therefore, there is no [ng] phoneme in Loglan. [Ng] might well be acceptable in Rex’s new language, though, because its rules (as we shall see in Part 2 of his paper) guarantee that [ng] will never be initial. As an allophone of Loglan /n/, of course, [ng] is fine. In fact it is quite common in that role in the target languages.

So I’m afraid it’s not true that [ng] is a “full-fledged phoneme” in the target languages as Rex says it is. It would be a “half-fledged” one in Loglan and therefore ok as an allophone but unsuitable for at least this international language. Besides, no single letter is ever used to represent [ng] in the Latin alphabet. That alone could be a decisive reason for excluding it.

By the standards that have governed the adoption of new Loglan phonemes in the past—for example, h in 1977, c and j in 1959—[ng] is a poor candidate. It would reduce legibility; it would not increase memorability significantly; and it would give us no new little words. Or if it did, it would do so at the expense of requiring our target population—who are still four-fifths of the human population, remember, almost none of whom know how to pronounce [ng] initially now—to learn to say [ngah ngeh ngee ngoh ngoo]...a pleasant exercise for the amateur linguist who is planning a trip to the Ngoro-ngoro crater, I suppose, but not one likely to endear us to the inhabitants of Desmoines or Vladivostok.

Fault #4. The admission of /y/ (schwa). Well, I admit that this one caused me some worry...not its use as a consonant-buffer, which is nonproblematic, but its use as a hyphen. Schwa, pronounced [uh], is the “reduced vowel” of all the Germanic and Slavic languages. That is, it is what happens to a normal vowel in these languages when it moves into an unstressed position. (Note what happens to American [ah]—[aw] in British speech—when it moves from a stressed syllable in the noun [PRAH-jekt] to an unstressed one in the verb [pruh-JEKT].) It is therefore almost unexpungeable from the efforts of native speakers of these languages to speak other languages, for example, Spanish and Loglan, that do not reduce vowels. So native English-speakers end up saying [KAH-suh] for Spanish casa no matter how we train them. Only long residence in a Spanish speaking country can teach one to stretch one’s mouth around that final syllable and say [KAA-saa]. As a consequence, I worried—and in fact, still worry—about the morphological ambiguity that might be introduced into words like matmymatma by such speakers now that we have decided to use y as a hyphen. How is this complex to be distinguished from the phrase matma matma by people who can’t produce a proper /a/ in unstressed syllables? Well, there is one hope here, and that is stress. That’s another difference between the two productions besides these hopelessly muddied a’s: /matmyMATma/ vs. /MATma-MATma/. The complex word has one stressed syllable; the phrase has two. Admittedly, one stress will always be weaker than the other; but there are nevertheless two distinct stresses in the phrase and only one in the word. Can our computers be taught to hear such minuscule differences? I hope so. I even think so. And if so, then we can get by with this use of schwa as a “gluing” morph in complexes...even in the Germanic and Slavic dialects of our language. But if they can’t be, we may end up regretting using schwa as a hyphen, and fall back on the continuants.

But adding schwa was a powerful move, formally. And once the need for it as a consonant buffer was seen in 1982—in which role, as even Rex agrees, its choice over all other vowels is unimpeachable—then replacing the more awkwardly polymorphic r/n/l-hyphen, which is what we had been using in these places, with the simple monomorphic y was a move urged on me by practically everyone...well, by Dr. McIvor and Birrell Walsh in 1983, and again by Bob LeChevalier in 1985, all with good arguments. So in the end, I agreed with them that using y as our intraverbal hyphen as well as a consonant buffer was a formally simplifying move (if only people could pronounce their unstressed a’s unschwa-ishly!); and y appeared for the first time as a hyphen in the language described in Notebook 3 in 1987.

But I almost agree with Rex on this one. Hyphen y comes closer than any other of his alleged flaws to being a real flaw in the language. But even if so, it is a very pretty one! Like the birthmark on the face of a beloved, one grows to positively love it. At the same time one continues to feel uneasy—even a little belligerent, perhaps— about how others will perceive it! And I confess I do.

Flaw #5. The ciV (and other CiV and CuV) words and syllables. It’s not fair to call this one a flaw. This is simply terra incognita, a place where the most serviceable rule-structure has not emerged yet. I could have run experiments with logli volunteers to determine what the most pleasant or natural thing to do with the “optional disyllables” is. But that would have taken a lot of time, delaying the revision of Loglan 1 and GPA by possibly another year, and as far as I can yet see, there are no serious structural or learnability issues involved. Words resolve easily enough whether the i/u-initial vowel-pairs are pronounced as monosyllables or as disyllables. So I decided to leave this question open, to “let usage decide”. Usage hasn’t decided yet...although I have observed my own speech and that of a few others, and so have my own bets laid. Obviously the simplest rule—one that covers all bases—is to pronounce them all disyllabically! But I doubt very much that that will turn out to be the best rule. I can’t quite imagine teaching myself to say [EE-ah] for ia or [OO-ah] for ua. Can you? Any opinions beginning to shape up on how to pronounce the optional disyllables? Let’s hear them.

Flaw #6. Writing the diphthong [ow] as ao. This is another non-flaw. It is an obvious and well-grounded writing rule; one followed, by the way, by the designers of Pinyin, which is now the official Romanization of Chinese. So [ow], a sound that occurs frequently in Chinese, is written ‘ao’ in that language too when Latin characters are used. Writing [ow] as ao is an especially useful thing to do in Loglan. It provides a clear and easily understandable distinction between ao and the disyllable [ah-oo], which is the pronunciation of yet another permutation of the five Loglan vowels taken two at a time, namely au. There are no better, more distinct, or more consistent interpretations of the two permutations ao and au than pronouncing them as [ow] and [ah-oo]. Moreover, if you say [ah-oh] repeatedly and quickly, what you get is [ow]. Saying [ah-oo] repeatedly and quickly does not give that result Try it. Rex is wrong on this one. It is true that if you play a recording of [wah] backwards, what you get is [ow]. But the [w] allophone of Loglan u that figures in the monosyllabic pronunciation of ua is not the same sound acoustically as the [oo] allophone that figures in the disyllabic pronunciation of au; and it’s acoustics that recorders listen to, not phonemics. These are subtle matters, but quite clear ones once one develops an ear for this sort of listening.

Flaw #7. The advantages of writing [y] as an allophone of i and [w] as an allophone of u are obvious and quickly stated. Doing so cuts down on the size of the phoneme set that has to be learned; it is an economical use of letters; and it aids in the identification of affixes by allowing [tweh], for example, to be written tue, which thus identifies it as a combining form of tugle, and this will be true despite the fact that different allophones of Loglan u occur in the two morphs. Writing [y] as i and [w] as u also avoids the troublesome ambiguities and irregularities that are associated with pronunciations of the letters V and ‘w’ among peoples who use the Latin alphabet. Collapsing the sounds [y] with [ee] and [w] with [oo] to make two phonemes out of four avoids all this and has no real disadvantages. The semivowels [y] and [w] are nearly always called for before vowels—yes; only “nearly always”; for there’s that terra incognita of the optional disyllables still ahead of us—and no one has any trouble learning that ia is pronounced [yah]. In fact, it is phonologically instructive to make the discovery that i before a actually yields the sound of ‘Yah’, just as it is instructive to learn that a before i yields the sound of English ‘eye’; for those diphthongs are what emerge when [ee-ah] and [ah-ee] are spoken repeatedly and rapidly. So using ‘i’ for [y] and ‘u’ for [w] is, in fact, an attractively spartan feature of the language, part of its “stark simplicity”. This is hardly a flaw.

Flaw #8. Ah, those “procrustean primitives”. Rex saw the reason for them before the (nearly) free-form borrowings were added, but now he doesn’t see why the primitives, too, cannot have the same free forms. The main reason, Rex, is that they would then not be recognizable as primitives, nor would the borrowings be recognizable as borrowings; and it is very important in Loglan that both these types of predicates be recognizable for what they are...it is part of that “learnability by adults” business that we are trying to maximize. Moreover, if the forms of primitives were really free (as not even Rex proposes, by the way), then they would be indistinguishable from little words or little word compounds on the one hand, and from names and borrowings on the other (a distinction which he does want to collapse); and then the whole point of Loglan morphology, namely the ability of the newcomer to recognize the approximate grammatical and semantical role of each new word da encounters from an examination of its shape alone, would be lost. This is one of the babies we don’t want thrown out with the bath...and neither does Rex, it will turn out, although some of our babies do slip by him. But back to the functions of the 5-letter form. Believe it or not, an even more important and more ancient reason for adhering to the Loglan 5-letter primitive form is that it is probably the very form-constancy of Loglan primitive predicates—which are the most important predicates in the language, remember, the ones that everyone learns first—that makes them so remarkably easy to learn. In my experience, and in the experience of the handful of others whose learning I’ve been privileged to monitor over the years, the average learner can master about 800 Loglan primitive predicates in two or three weeks of one or two hours of flashcard work a day, probably even less if using MacTeach 2. I do not believe this can be done with the primitive notions of any other language, including Esperanto, despite the fact that, at first glance, Loglan primitives do not look easy to learn. Why, then, are they so easy? That’s a fascinating question; and there are many possible answers. But the one that strikes me as worth investigating first is that, compared to “natural, free-form” words, there are so many things about the Loglan primitives that you don’t have to remember when you’re trying to learn one. That it is five letters long, for example. That’s a constant. That it has exactly 3 consonants and 2 vowels. That’s a constant. That, like a Loglan little word, it is C-initial and V-final, but unlike that little word, there’ll always be a consonant-pair and it’ll be either initial or medial. And so on. This bundle of easily remembered features apparently helps the learning process in the most remarkable way. It may allow learners to concentrate on the differences within the form, the internal features that distinguish one primitive from another: godzi from gotri, for example. But da can forget about all the features which all primitives share. That’s one possible hypothesis about why learning Loglan primitives is so easy. Another possible explanation is that the 5-letter form is also a remarkably generous one. In fact, it’s a deliberately designed “clue transporter”. So an awful lot of multilingual cues can be piled into it. It can contain a sizeable part of each of two short Chinese and English words end-to-end, for example, as in supta (English soup plus Chinese tang), or henji (hen plus ji); or it can accommodate 6 or 7 natural words piled up on top of one another, as in matma, mrenu and the famous blanu. Far from being distracting to the monolingual learner, the presence of this polyglot cargo in the words da is learning is apparently fascinating to the internationally oriented logli.and most of our logli are. Witness Stephen Rice’s recent experience with them, for example. Indeed, who can forget that the Chinese word for chicken is ji after learning that the Loglan word is henji? No shorter word-form would allow us to cram so much of this truly planetary information into our words, information that we hope is going to prove just as useful to the logli on the other side of the planet as the English part of that cargo is proving useful to us. This is the kind of “cultural neutrality” that Loglan offers, the warm synergetic kind. And that is what the generosity of the 5-letter form provides for. Besides, when one looks at the design problem from the purely formal point of view, one discovers (actually rediscovers, for this is no accident, of course) that the two 5-letter forms we currently use are the shortest C-initial, V-final forms that are distinguishable from little words or little word compounds; see Note 6 on page 39 of the 3rd Edition of L1 if you want to examine this argument... those of you who still have that tattered little book. If we want that feature—and I assume we still do—these are the shortest forms we can use to contain our primitives. Luckily they are just long enough to allow a good deal of usefully composite learning information to be packed into them. Moreover, since they are combinatorially extremely numerous—numbering nearly 100,000, it turns out—our primitive word-space need never be crowded. Either Rex doesn’t understand these things about the 5-letter form, or he is indifferent to its functions. But flaw in the structure of Loglan, the 5-letter primitive is not. It may, in fact—as we will not see fully until we examine its effect on learnability in Part 2 of this paper—be one of the language’s most successful morphological features.

Flaw #9. Allomorphs, or the provision of short combining forms—abbreviations, if you like, sometimes several of them—for each of the hundreds of productive primitives that figure most frequently in our metaphors. Again, Rex appears not to understand the architecture of this unique system, or the large and virtually unprecedented linguistic function which it is designed to serve. That function is, in a nutshell—and I hope I have not put this too immodestly—to make the spontaneous construction of an infinitude of complex predicates available to future logli through the universal human art of combining old notions in new patterns to suggest new meanings. There is nothing like this in Esperanto. If Rex thinks there is, then he doesn’t understand what the Loglan system is or does. The only two languages in our target set that approach Loglan’s metaphorical “cornucopiousness”—if you will allow me that pleasant neologism (a Whorfian effect if I ever saw one!)—are the two “compounding” languages Chinese and German. But these languages do not go far enough in metaphor-management. Chinese and German metaphorical words are often awkwardly long, and the morphologies of these languages offer no certain clues to word boundaries in the spoken form. The Loglan affix system solves both problems. It shrinks the defining metaphor to make the complex, typically by 40%. Thus rodja madzo sensi madzo becomes rojmaosesmao, and drops 8 phonemes in the process. And the strings of elements that make the complex, being affixes and not words, cannot be mistaken for strings of words. Finally, the procedures used by logli to make these and other types of words guarantee that (1) complexes will always be identifiable as such, and (2) their boundaries will always be known.

The long-range goals of this system of affixes are admittedly fairly grand. They are to make it possible for logli to invent or incorporate into their language hundreds of thousands of metaphor-based words—perhaps, in the end, googolplexes of them—all capable of having been spontaneously invented by some gifted speaker in da’s moment of inspired need, as well as instantly deciphered by the first as well as all subsequent auditors of da’s metaphor, and in this way be collectively capable of allowing humans to express their diverse experiences on this planet in their international language, and perhaps their experiences on other planets as well. Esperanto will not do that. Nor will any other language of which we know. So here again is a first for Loglan, a capacity unmatched by any other language, natural or constructed: the promise of infinite metaphoric—one might even say “meteoric”—growth.

Rex was not among us while the Great Morphological Revision was going on. So he didn’t read the articles and notebooks in which the functions and features of the emerging system were analyzed and discussed. I do agree, though, that it is a huge system, not easy to take in all at once, and that its heart is not worn on its sleeve.

I myself am still learning about this immense system. For example, I apparently had a misconception about how it should be learned, one that has recently been replaced with a new insight. Let me share that insight with you. Apparently one does not have to learn the entire set of affixes in one fell swoop before one can make or decipher complexes with confidence...not, that is, the way one does seem to need all the primitives to be on deck at once, so to speak, to have the full semantic pallette of the language at one’s fingertips, before one can experience that sense of chromatic closure that enables one to paint new complexes with assurance. But one doesn’t have to know all the affixes to do this...not even, as Dr. McIvor and I have recently found out, if one is going to judge the quality of old metaphors and bang new affixes together into new ones when the old ones seem a bit pallid. Even under those relatively demanding circumstances, one seems to learn of the affix system just what one needs of it for current purposes. The important thing is that one know the pallette, the set of primitive notions that can be, and usually are, representable in this abbreviated way. Then, when you decide you might want to use one in a word-painting, and you’re not quite sure yet what its abbreviations are, you look it up. The 4th Edition serves very handily for that. The curious thing is that any abbreviation that one has ever used to make or understand a complex in this way seems to remain indelibly in one’s mind. The affix-learning experience abounds, in short, in one-shot learning. Why is that? Probably because, from a context of active use in word-making, one doesn’t need to memorize the primitive/affix pairings that one encounters. One doesn’t need to memorize, for example, that tel- means terla (terrestrial) when one has just discovered that telbie means “Earth-edge”, and then finds out that that’s some “ancient” logli’s word-painting for the line that separates Earth from sky. That discovery, and its embodiment in the newly encountered word telbie, is so much more interesting than relating any affix to its primitive that we just don’t bother to do the latter. Anyway, the relating seems to be done automatically for us while our attention is held by the metaphor. That brings me to another curious point about affix-learning. The better the metaphor in which one first uses, or encounters, a new affix, the more indelible its trace seems to be. Who can forget that telbie means “Earth-edge”, for example? The metaphor has a vividness that practically vaults one into space. And that experience seems to lend a similar indelibility to the otherwise ho-hum facts that -bie means bidje and tel- means terla. Learning affixes in this active, contextual, painter’s way seems to be all the affix-learning one needs to do. So MacTeach 3 does not seem now to be the most efficient way of learning Loglan affixes. Making new complexes and deciphering old ones does. M3 presents the affixes nakedly, out of context, in long silent ranks. So the labor of learning them in this laborious, repetitive, albeit “rectangular” way (and that may still be a useful property, for brush-up purposes, for example) may very well prove more than we need to spend. The optimal solution to this problem still lies ahead of us. We don’t know yet how to induce affix-learning in that “miraculous” one-shot way. But we now have clues. I invite anyone who has an interest in exploring these kinds of learning problems to explore this one...and tell us what da learns.

Flaw #10. Esthetics. I’m afraid I find this all pretty subjective, pretty much a function of what Rex finds, or fails to find, interesting in the architecture of the language. If Rex is “bored” by an encounter with yet another Loglan 5-letter primitive, one like najda, for example, but “refreshed” by encountering a word like iglu, could it be that the information being conveyed by these two shapes does not interest him? That the first is a primitive, and therefore likely to be in some sense a “universal” notion, and more likely to be involved in the metaphors of the language? That the second is a borrowing, and therefore in some sense a “local” or specialized notion? That one is, therefore, not likely to find igloos in Africa, but that 5-letter najdas are likely to be found all over the place? I wonder if Rex is also disinterested in the fact that all the V-, CV- and CVV-form words he encountered in “Classical Loglan” were logical or grammatical words? In a word, the structure words of the language? Presumably not, for this information survives in the ingenious morphology Rex has designed. But does he find it useful to know that V-initial little words are either logical connectives or expressions of attitudes? And that the latter are always VV in form and the former never are? Apparently not, for these distinctions do not survive. But to my eyes and ears, at least, letting these messages seep in through my morphological pores is part of the fun of doing Loglan: spotting just these structural and etymological and semantical and logical and attitudinal clues that shape the message as one goes along. If we need to teach “Loglan Architecture” to our new logli to put them in touch with this aspect of the Loglan experience—teach courses on the model of “Music Appreciation” ones, or Art History ones like “Understanding the Baroque”—then we should probably do so. But I get the feeling while reading Rex’s account of his own esthetic response to the language that he is missing something that some of the rest of us, at least, are seeing...or that he’s just uninterested in what we’re seeing. I admit that Loglan architecture has become a little baroque, that the days of “stark simplicity” are over. But even that’s a challenge...both to perceive and understand how and why the language has become what it has. When musical forms advance from unison to four-part choir music, there are some problems of perception at first; but the eventual rewards of understanding the new forms are esthetically fairly great. The move from 1975 Loglan to 1989 was that kind of advance. It was admittedly an intrication. But later on there may be fugues.

On the other hand, some of Rex’s “esthetic” plaints are real enough. I, too, have trouble with those new “gluing” h’s. It may be that we will have to go back to using r as our gluing morph in borrowings. R is far easier to both pronounce and hear in these gluing spots than h is: artomi, for example, easier than athomi. The only trouble with using r as glue is that it tends to bear too much semantic freight. That is to say, more than just occasionally an r pops up in a word where it could mean something other thah a spot of glue; and that may well mislead us. (Is the r in arteri a gluing r or a “real” one? This sort of problem comes up much more frequently with r than with h; thus ethili happens but is rare.) The resolution of this delicate tradeoff was tipped in the h-direction for me in 1988 when I asked my panel of science advisors which they preferred: a list of about fifty r-glued science words I had prepared for them, or the same fifty words glued up with h? In other words, 50 choices like ?artomi or ?athomi for atom. Unanimously and uniformly my judges chose the h-words. Their consensus impressed me. But it was a sight test. No speaking or listening was suggested or required. It may be, as we confront the difficulty of both hearing and speaking h in these gluing contexts, that despite the indubitable ease of reading and writing h, we will go back to the more mellifluous r...or find some fresh solution. I regard the book on this issue as still only half-closed.

As for Spanish-speakers not being able to use their jota for Loglan h...why not? Both Spaniards and Russians do this when they are learning English. Why not with Loglan? The fact that their [kh] sound is the same as the sound of the x that is irregular in standard Loglan doesn’t really matter. At first x and h will be the same sound in their learners’ dialect of Loglan as it is in their learners’ dialect of English. Later the two will separate.

The main problem with h will be for and with the French, who have no h of any kind! What will our French friends do with athomi? Whatever it is they do with ‘h’s when learning English or German, I should imagine. (Frequently they leave them out.) But our use of h in borrowings is not, after all, the only use we logli make of it. Logli will be saying Tu he? in la Belle France, too, I expect. Will it sound like Tu ’e? there?

The vocalic use of the continuants m n r l is not confined to English, as Rex seems to be suggesting. Moreover, the vocalic values of continuants are required for making good imitations of the full spectrum of natural names...something Rex may not be interested in, but we still are. Even those who do not have vocalic continuants in their native languages can be easily taught to make them by using their own consonantal values as starting points. In fact, once made in this adventurous way, the vocalic versions of these consonants will often sound comical to such ears. For example, to British ears, sustaining an American /r/ often makes it sound like the growl of a dog. Funny, but certainly not difficult to imitate. Even between a consonant and another vowel it’s not difficult to sustain one of these continuants, as in the American pronunciation of the name ‘Thoreau’ [thrr-OH], A Briton would say [THAW-roh], or perhaps [thaw-ROH] following the French. And how clearly, then, the disyllable that results contrasts with the monosyllabic [throh] throw in which the /r/ is sternly consonantal. So there is no real problem here, not for anyone who is familiar with these sounds as consonants...which is everyone in the target population but the Japanese, and they lack only l. A little drill, perhaps, may be necessary to teach logli who don’t use these consonants as vowels in their native tongues how to make vowels of them here. But there’s nothing articulatorily awkward about that. There are no new positions to master, which would be awkward. And, as I say, they tend to sound funny to such ears; and that is sure to help.

As for pauses, they occur significantly in all languages. They’re just hard to notice, because they haven’t been put systematically into the writing system of any language...yet. In Loglan they will begin, at least, to make their debut on the printed page, and so become noticeable. As for their necessity in Loglan, there is simply no other (economical) means of preventing form-masking at consonant-consonant word junctures, as may take place between names in Loglan ([TAHM.-MAWR-uh-suhn]), or at vowel-vowel junctures, as takes place before V-initial words {[lah.igh-LEEN] and [leh.EEG-loo]). It is possible that Rex has not perceived the problem here, and so doesn’t see the need for its solution.

Fault #11. Names. Let’s make short work of this one. In Rex’s own language, which for brevity I’m going to call “Rexlan”, names will be morphologically undistinguished just as the names of a natural language tend to be similar to its other words. But we logli invite names to enter our language from every linguistic quarter; and we try to distort them as little as possible when they get here...just to make the persons who bear them, or who live in the places that bear them, feel at home when they arrive in Loglandia. Now it happens that this noble project requires that we give names some unobtrustive distinguishing feature in our language, something very like that final consonant. Again, it is perhaps because Rex doesn’t see our problem—or see it as a problem—that he doesn’t like our solution to it. The problem, Rex, is that when viewed in planetary perspective, natural names exhibit an absolute riot of morphological forms. Yet, in an international language, we want to respect that riot, not banish it by applying linguacentric incorporation rules: like the Britisher introducing da’s French visitor Richard as [RIHT-chuhd], for example, instead of saying [ree-SHAHR]; or more arrogantly still, my calling Deutschland ‘Germany’ if I am English, ‘Alemania’ if Spanish, and ‘Germania’ if Italian. In Loglan, we renounce such ethnocentric arrogance. We import each natural name into our language by reproducing its local pronunciation as faithfully as our phonemes will allow. Given that the non-names of our language are beautifully regular, and that we wish to lift all words, including names, intact from the speechstream, what do we do? There are very few things we can do, morphologically, to distinguish such riotous objects in a stream of non-riotous ones; and all of them will involve giving names some invariant and unique feature. The most workable of these—and I believe I have studied most of the available mechanisms—is to give them a consonantal ending. Since all other Loglan words are vowel-final, that does it. But what this move entails that Rex seems to object to is that we pause after that final consonant. But the pause is just as necessary as the consonant! Otherwise (1) a name may get stuck to some other name (the “Jimbrown” phenomenon), or (2) its final consonant may be borrowed by the following word and it will cease to be a name (the “Jimrenu” one). Hence that odd-looking final consonant must be followed by a pause in careful speech—in talking to one’s computer or one’s accountant, say—just to preserve the useful oddity of our names. (Of course one can omit those name-marking pauses when speaking with error-tolerant humans, but not all humans are error-tolerant; for example, learners and teachers aren’t.) So here again, what Rex complains of is no flaw.

I’m afraid there’s another feature of Rex’s treatment of names that suggests a more serious misunderstanding, and that concerns the role of names in languages generally. In most natural languages (French being the only exception that I know of, and then it’s only a semi-exception because the distinction I’m speaking of appears only in its written form) we do not distinguish very carefully between the use of a name as a singular term (Mr. Chrysler) and what is very often the same word used as a general term (a Chrysler car). In Loglan, being a logical language, the distinction between singular and general terms is fundamental. It must be categorically observed in both the written and the spoken form. In fact, the name/predicate distinction may be the logically most fundamental one in the language, paralleling as it does the distinction between designation and predication itself. So we are pretty careful about this one. That is why we have an ordinary imitative name for the individual that is the English language (Inglec) but a slew of words of an entirely different morphological type for the predicate senses of English: gleca, gleci and gleco. Every logician will applaud us for this. Rex, it seems, wants to throw this distinction away and go back to the muddiness of natural language. This shows perhaps more clearly than anything else he’s said that Rexlan is not Loglan, and isn’t meant to be.

[To be continued.]

Copyright © 1990 by The Loglan Institute. All rights reserved.

Thanks to Mike S. for proofreading the OCR files for this article.