(Return to the Table of Contents)
... The name of the song is called "Haddocks' Eyes",' [said the White Knight.]
'0h, that's the name of the song, is it?' Alice said, trying to feel interested.
'No, you don't understand,' the Knight said, looking a little vexed. 'That's what the name is called. The name really is "The Aged Aged Man".'
'Then I ought to have said "That's what the song is called"?' Alice corrected herself.
'No, you oughtn't: that's another thing. The song is called "Ways and Means": but that's only what it's called, you know!'
'Well, what is the song, then? said Alice, who was by this time completely bewildered.
'I was coming to that,' the Knight said. 'The song really is "A-sitting On a Gate": and the tune's my own invention.'
-Lewis Carroll
Notes 56
Loglan is a language which was originally devised to test the Sapir-Whorf hypothesis that the structure of language determines the boundaries of human thought. An important implication of this hypothesis is that the widely differing structures of individual human languages must therefore set very different formal limits on the historical potentials of the various human cultures that are, in a sense, contained in them. Glimpses of these limits were seen in the data of comparative linguistics by Edward Sapir in the 1920's, and a hypothesis which explained their structural origin was proposed by Benjamin Lee Whorf in the 1930's.1 By the early 1950's, the many detailed implications of their theory of linguistic relativity had begun to occupy the attention not only of other linguists,2 but of psychologists and sociologists as well.
During the 1950's a good deal of cross-cultural experimental work was done on the psychological fringes of the hypothesis, and while many of the results were corroborative, not all of them were, and of course none were decisive.3 The Whorfian phenomenon, if indeed it existed, was apparently so deeply imbedded in the reality-shaping mechanisms of both language and culture that there seemed to be no way of disengaging it for a decisive test. What was wanted, if this important idea was not gradually to be surrendered as essentially untestable, was a device capable of separating the presumed linguistic cause from the predicted cultural effect. But in nature such uncontaminated devices do not exist.
Work on Loglan began in 1955. As a sociologist with a background in both social psychology and philosophy, my own interest in the Sapir-Whorf hypothesis centered around the symbolic mechanisms--the logics, metaphysics and epistemics--by which human individuals contrive both their self-images and their world-views. An important theory of the symbolic process as it eventuates in selves and world-views had been proposed by the American philosopher George Herbert Mead (1934); but Mead's work was uninformed by comparative linguistics. Clearly if these symbolic mechanisms differed among languages as much as Sapir and Whorf thought they did, then the worlds and selves of persons living in different cultures ought also to differ systematically.
But these matters differ systematically anyway, and which difference is due to what cause is nearly impossible to say. For it is nearly impossible to distinguish observationally between the structural effects of the primary language spoken by a people and the content of the culture about which and within which they speak it. Bilingualism offers some opportunity for disentangling these effects. But true bilinguals are rare; their cultural biographies are usually wonderfully idiosyncratic and hence incomparable; and there is a disconcerting sense in which the bilingual has escaped the Whorfian predicament anyway. To unravel the purely linguistic effects of given languages on human thinking from the effects of the general cultural milieux in which thinking takes place seemed, therefore, to require an experimental approach. In particular, it seemed to require the deliberate introduction of a reasonably culture-free second language with known formal properties into a laboratory-like setting in which its effects on the behavior we call thinking could be precisely gauged.
But how does one import anything as massive as a human language into the laboratory? Well, how does one experiment with lightning? The first thing you do is reduce its scale. You try to reduce this magnificent natural phenomenon to a manageable spark.
But could the size of a human language be significantly reduced without destroying its essential character? Most of the linguists I read and talked to thought not; but one could not know the answer to this important question beforehand. All one could do was try. Then, if the contrived phenomenon, the manageable linguistic spark, behaved in every relevant way like a human language--that is, if people could actually speak it, think in it, generate and transform ideas in it--it was just possible that its manipulation under conditions of control would permit us to make certain limited inferences about the natural phenomenon itself. In particular, not only the Sapir-Whorf hypothesis, but many other questions about the three-way interaction between languages, cultures, and human mental processes seemed amenable to this kind of small-scale, linguistical experimentation.4
In 1955 there was no experimental linguistics. But the then very rapid development of theoretical and mathematical linguistics seemed to be transcending the earlier descriptive stage of that science, and hence to offer clear if tentative guidelines for the construction of just such a culture-free language as my strategy required. A number of essential features of linguistic structure had recently come to light, or were strongly suspected; the major components of the language mechanism had been identified; and except for the semantic levels of linguistic structure--then defying exact analysis as they still are5--the interactions between the lower levels of the rule structure were rapidly being reduced to exact, mathematical descriptions. Guided by these descriptions it should be possible, it seemed to me, to construct a system which, while distinctly smaller than a human language, bore all those other features of the natural objects which were then thought to be essential.
But if the constructed language were to serve as a laboratory instrument--something to "release", as it were, or at least increase the probability of observing, Whorfian effects--it would obviously not do to imitate any natural language or group of languages too closely. What was wanted, apart from smallness, was not a typical human language, but a severely atypical one. For if the Whorfian effects of second-language6 learning turned out to be elusive--and compared to those of primary language learning, which we had decided were impossible to isolate, we could certainly expect them to be minor--they would probably not be revealed in a single culture in a single try. In fact, we could probably expect the complete pattern of any given second language's Whorfian effects to emerge only against the background of many primary cultures, and perhaps only then after many tries. Thus a language constructed to release measurable Whorfian effects when learned as a second language should offer fairly large structural contrasts with all the first languages that might be involved. Since any natural language might eventually be involved, what was required was a diminutive but nevertheless genuine human language which was easily learned by adults and which differed from all natural languages in some scientifically interesting way. In short, some decisive and probably functional difference between the constructed language and at least a wide range of natural languages would have to be found.
The most promising way to create such a difference, it seemed to me, was to exaggerate some natural function of human language, that is, to increase the functional adequacy of some complex of linguistic structures in a way that would have a strong independent likelihood of enhancing the measurable performance of its learners on some specified set of tasks. Besides, in its original formulation the Whorf hypothesis is a negative one: language limits thought.7 One way of disclosing such phenomena is to take the suspected limits off, more precisely, to push them outward in some direction in which removing limits would have predictable effects. So it was settled. The diminutive language should also be a functionally extreme one in some known or presumable way: an extremely poetic one, say, or an extremely efficient one, or extremely logical.
Now there is very little scientific knowledge about the literary functions of language, and while a lot is known about efficient codes, it is hard to relate this property to Whorfian effects. Enhancing and clarifying the logical structures of the diminutive language, however, seemed to answer all the requirements of the project. There is a very considerable body of knowledge about the formal properties of logical systems; and a hyperlogical linguistic structure should have a clear and interesting Whorfian effect if it had any: namely the facilitation of certain identifiable kinds of thought. Not only that, but a language which only faintly promised such a mind-enhancing effect would almost certainly prove attractive to a large body of potential learners, namely students. Thus the idea of Loglan as a hyperlogical or thought-facilitating language had a very natural birth.
Or rather, rebirth. For the dream of a logical language is, of course, a very old one...at least as old as Leibniz and probably much older. Some efforts had indeed been made in the pre-scientific era of language study to construct such a language. 8 But like nearly all such early essays in language-building--coming, as they did, before the huge, unconscious mass of the language structure was known, or even suspected--these efforts too had failed. Then too in 1955 the philosophical as well as the linguistic stage seemed to be set for taking up this age-old human project once again. Logicians had made great strides in the analysis of a wide variety of scientific and mathematical forms of thought,9 and at least one school of analytic philosophers, the so-called "ordinary language" school, had made the analysis of conversational forms the focus of a very considerable philosophic effort.10 Thus the art as well as the science of language analysis had, in the preceding decades, developed in a way that seemed to lead quite naturally to the resumption of the logical language project.
The name 'Loglan' was derived before the language itself was built, from the two English words 'logical' and 'language'. But the claim invested in this metaphor is in fact narrower than the wide word 'logical' suggests. Loglan is logical only in the sense of purporting to facilitate certain limited kinds of thought: namely those kinds which proceed by the transformation of sentences into other sentences in such a way that if the first are true so also are the second. We might also expect it to minimize, or help prevent, the errors that are usually made in performing such deductive operations. But these are fairly modest senses of the word 'logical'. We might have meant to convey by it the much stronger claim that Loglan is a deductive system, in the sense that geometry and formal logic are. To support such a claim we would have had to show that Loglan had a set of elementary notions and elementary operations from which all its complex notions and complex operations had been rigorously derived. But we do not make this claim. Derivation in Loglan, as in the natural tongues, is by metaphor, not by formal definition. In fact we take the familiar mechanism by which new meanings are spontaneously created by a speaker or writer combining old words in new ways to be one of the essential properties of human language, and hence one which we must not remove if Loglan is to be a veridical member of its genus. For surely one of the most striking behavioral distinctions between using a language and using a deductive system is that the speaker of a language is at liberty to extend its semantic field by instant metaphor in any direction that he chooses. This is a move that is denied the geometer and the logician, as they well know. The users of deductive systems must introduce new terms by formal definition or not at all. Clearly, Loglan could not be logical in this or any other sense which deprived its speakers of the essential moves of speech.
There are other senses of the English word 'logical' in which Loglan isn't. It is not, for example, wholly consistent; nor could it be and remain optimally logical in the transformational sense implied. Loglan is more consistent than most languages. But though it is a small language, it is a large system; and large systems, like large minds, tend to be intolerant of consistency. Neither is it "reasonable" nor "self-evident." Like the meanings of the simplest words in all spoken tongues, the basic meanings of Loglan are essentially arbitrary. Learning them must be undertaken in the spirit of boundless innocence with which one approaches any foreign tongue. Yet the reader will find, I think, that the narrow transformational mode in which Loglan is logical is, in the end, a very rich one, and one that richly distinguishes it from all the natural tongues with which he may be familiar. For the selection pressure on the evolution of all surviving languages has almost undoubtedly been greatest in the nursery and the marketplace, not the study, and certainly not in that all-too-recent habitat of some of them, the scientific laboratory. So in selecting these relatively new functions of language for optimization--new, that is, in the long time-scale of language evolution--we have perhaps overstepped the bounds of nature. Yet it is precisely this that we intend. For by making Loglan an extreme instance of the genus Language--by extending it along a single dimension of language structure as far as it would go and still be speakable--we have given ourselves a far better chance of observing the Whorfian phenomenon than if we had contented ourselves with a culture-free imitation of some average human tongue.
So let us be clear about our scientific strategy. By making Loglan hyperlogical we have intended to maximize, or at least greatly increase, the probability of observing the Whorfian effect. We are in pursuit of a simple existential: Does the phenomenon exist? Then, if it does exist, there are many more subtle things we might wish to do with it, or about it.
Building these logical structures into Loglan has been less demanding linguistically than most linguists I have talked to have supposed. The list of transformation structures I have taken to be necessary, if perhaps not quite sufficient, for these purposes is a short one. It includes speakable provisions for (i) the propositional calculus, including the unique determination of connective scope; (ii) the apparatus of quantification theory, including a clear distinction between bound and unbound variables; (iii) clear distinctions between all known modes of designation and description; and last, and most tellingly, perhaps, (iv) a word-classification scheme that (a) allows all claims to be expressed in the predicate calculus and (b) treats all predicates indiscriminately except as they are distinguished by the number of their places. This means that Loglan has no nouns, verbs, adverbs or adjectives in any ordinary sense, but only predicates and multi-place ones as required. Now this, among all the ploys adopted by logicians, is perhaps the most far-reaching in its implications for language structure. I suspect it will turn out also to be the most troubling, and yet ultimately the most freeing, feature of Loglan grammar for the English-thinking mind. It is also what makes Loglan processable by machines.
The logically-trained reader will have observed that this list of essential transformation structures is not only a short one, but far less ambitious than it might have been. There is no notational provision in Loglan, for example, for a theory of types...or for any other scheme for removing the paradoxes to which the absence of a hierarchical notation quickly leads. Moreover, the notational provisions for a class calculus are rudimentary. So Loglan in its present form is less than an ideal vehicle for "speaking symbolic logic." Perhaps it could have been made to approach that ideal more closely; I do not know. Perhaps it will in time be brought closer to that ideal by the inventions of its users. But the maximization of Loglan's logicality, in my view, has not been nearly as important in the early days of the language as the retention of the essential ways of speech. By providing this elementary logical machinery in a form which, because it is speakable, may be directly accessible to the human mind, I have hoped to show that some behavioral consequences of a thought-facilitating kind ensue. Certainly the effect produced will not be the greatest that could in principle be achieved, if Whorf is right, by a language that was even more hyperlogical than Loglan. But these simple transformation structures may be enough to cast the Whorfian links between linguistic structure and the mental life into clear relief. And that is certainly enough for a first step.
Apart from the thought-facilitating functions of Loglan, the language is also meant to be a manageable laboratory instrument: teachable, measurable, controllable; its structure transparently observable both at the moment of introduction into any experiment and in continuous change. Loglan has a number of properties which bear on this complex instrumental function, but chief among them is the matter of scale.
Did we succeed in making Loglan small enough to be "a manageable linguistic spark?" I confess I do not yet know the answer to this question...even after thirty years. But Loglan does seem to be easily learned,11 and on every formal parameter it is agreeably small. The number of its grammar rules is an order of magnitude less than has come to be expected of natural grammars from recent work. Fewer than 200 two- to five-term rules are required to define its domain of permissible utterances; and this contrasts very favorably with the three- to six-thousand rule partial grammars that have been written in machine-translation work with languages like English and German.12 Moreover, the number of its elementary predicates is small: 800-odd as compared to an uncountably larger number for English, although perhaps not so small as Chinese, which is said to Construct its complex predicates from an even smaller list of radicals. The number of structure words in Loglan is also reasonably small: about 120 monosyllabic morphs and their compounds suffice for all the grammatical and logical work of the language. The number of "lexemes" or word classes is about half of what is expected in a natural language: less than 70 as compared with 133 in an early grammar of English.13 Its phonology is also at the small end of the human range: 27 phonemes as compared to the 45 found by a conservative count in English. So on comparative grounds Loglan Would seem to be manageably small. While the size of a language is not the only factor that determines the speed with which it is learned, it is undoubtedly an important one; and all my early teaching trials have suggested that Loglan is indeed very rapidly learned.
Another feature of the language that reflects its intended use as a laboratory instrument is its cultural neutrality. Partly this has been achieved by what we have come to call its "metaphysical parsimony," or the fact that its grammar presupposes a reasonably small set of assumptions about the world perhaps the smallest possible set, on our present understanding of language structure. This feature also supports the thought-facilitating functions of the language in some obvious, and in some not so obvious, ways. But its original purpose was to guarantee the metaphysical14 neutrality of the language for speakers of widely different native tongues. Thus any speaker, from any culture, should find it possible to regularly express in Loglan what he takes for granted about the world; and he will be able to do this without imposing--or what is perhaps more to the point, without being able to impose--these assumptions on his auditor. Thus, Loglan has many optional grammatical arrangements, but very few obligatory ones. There is no obligatory tense system, for example, as there is in English, nor is there an obligatory gender system as there is in most European tongues, nor is there an obligatory epistemic inflection of the verb as in Hopi. But both tense and epistemic operations exist as optional "inflections" of the Loglan "verb."
Still another element of Loglan's cultural neutrality reflects its intended use in cross-cultural experiments and possibly also as a medium of international translation. To this end I have tried to make the sounds of the basic words of the language equitably familiar to persons of very different language backgrounds. Its sounds and word-roots, for example, have been drawn with strict impartiality from the eight most widely spoken tongues. Of these eight, three are Oriental: Hindi, Japanese and the Mandarin dialect of Chinese. The other five are more likely to be familiar to readers of this book: English, Spanish, Russian, French and German. In this phonological sense Loglan is ready to be used as a "world language;" speakers of these eight languages comprise well over three-quarters of the present population of the earth. What is more germane, native speakers of any of these eight source languages will be able to hear many clues to meaning in Loglan speech. So cross-cultural comparisons of the results of learning Loglan as a second language will be possible across a broad range of native languages.
To maximize the total amount of phonological familiarity to be found in Loglan required that we use a composite system of vocabulary derivation for its "primitive" words. (By 'primitive' I mean words which are used within a language to derive its complex terms and are not themselves internally derived.15) For example, the Loglan word for 'blue' is blanu. And blanu is derived partly from Chinese 'lan', partly from English 'blue', French 'bleu' and German 'blau', and more remotely from Hindi 'nila', Spanish 'azul' and Russian 'galuboi'. Only Japanese, among the eight source languages, has no phonetic affinities with this particular Loglan word. Not all words incorporate so much familiarity, of course. But our word-building procedures have roughly maximized the total amount of it to be found in the language.16
A fourth instrumental property of Loglan bears on a functional relationship between languages that I hesitatingly call accommodation. We want the language to be small; yet we also want it to be very large. We want it to be large in the specific semantic sense of accommodating all that we might wish to say in it, either in response to urgencies developed within the language or because of those prior semantic urgencies that originate in the fact that we already speak other languages. We want, in short, to be able to speak Loglan not only like a Loglander, but also like a Trobriander or an Englishman, a Frenchman or a Chinese. This is a large order, and I am perfectly certain that I have not satisfied it. Yet the very effort to satisfy such a grand criterion has proved rewarding. More than any other functional property of the language, the installation of this one, even incompletely, has involved years of work. And the work is incomplete. It is on this point more than any other that I expect to be informed by the publication of even a fourth edition of this volume of the extent to which the grammar of Loglan does not permit the expression of meanings to which its speakers find themselves driven...whether, as I say, because of impulses generated within the language or from semantical needs coming from outside.
Yet the remarkable fact is that a very large amount of accommodation has already been built into the language without expanding its grammatical structure to those massive proportions to which linguists have accustomed us with their studies of natural grammars. One way of putting this is that the domain of permissible Loglan sentences is very large...so large, in fact, that English, say--or rather that subset of Loglan sentences which can be put into one-one, or more likely, many-one semantic correspondence with English sentences (for English is ambiguous)--fits into a very small corner of it. Yet the rules that define that vast domain are not numerous; not nearly as numerous as the rules of English are. If I am right in this observation--and of course I may not be, for Loglan may not turn out to accommodate the semantical field of English nearly as well as I think it does--then the possibility exists that for some important reason the grammars of the natural languages are far larger than they need to be. Far larger, for example, than is grammatically required to express the semantical field which they in fact engage.17
The last formal property of Loglan I wish to mention is its freedom from syntactic ambiguity. I originally conceived this property as an obvious essential for a maximally transformable tongue...obvious, because transformations proceed in one direction from one interpretation of an ambiguous sentence, and often in a quite different direction from another. Moreover, we manage to steer around the very considerable ambiguity of the natural tongues by entertaining only the most plausible interpretations of the sentences we hear. Consequently most natural thinking follows plausibility routes. But this is unfortunate. One wants to be able to think implausibly and yet clearly; for many important arguments have absurd conclusions. Unfortunately it is quite difficult to think nonsensically in an ambiguous language. Whence comes not only the logician's enforcement of all rules of clarity but also his strange love of nonsense, a love which more "natural" minds feel verges on the bizarre...as, for example, in the jabberwocky world of Lewis Carroll.18
Loglan's freedom from ambiguity has one surprising and possibly dysfunctional correlate. It was once feared that the helpless clarity of its sentences might prove a chilling feature of the language to its poets. One literary scholar (William Empson 1930) had shown that English poets, anyway, appear to use the rich fund of syntactic ambiguity in that language for poetic purposes. Thus if an English poet can contrive a line that says forty things, and each of these forty things suits her poetic purpose in some way, then she has accomplished at one stroke what a more humdrum writer might take yards to say. And then not say it. So grammatical clarity may therefore unfit Loglan for poetry just as it fits it for thoughtful discourse. Yet nonsense, too, furnishes literary delight; and an unambiguous language makes absurdities clearly sayable. In addition, for a basic structural reason we have not discussed yet, metaphor-making in Loglan is remarkably facile. The net result of these pluses and minuses on the Loglandic poetic act is not known yet. We have had some poets, but not enough yet. Again, as scientists, we have an interesting set of possible outcomes to observe.19
In sum, and as far as its intentional properties are concerned, Loglan is (1) formally small, almost certainly smaller than any natural language; (2) transformation-facilitating, and in that sense logical; (3) maximally recognizable over a very broad population base, and in that sense impartial; (4) metaphysically parsimonious and hence culturally neutral; (5) largely but probably still imperfectly accommodating of at least some natural tongues; and (6) syntactically unambiguous. What all this may mean in the laboratory is, as yet, anybody's guess. But my own sense is that we have, for the first time, a small-scale model of a human language that may be worth experimenting with.
The kind of laboratory for which Loglan was originally designed is the social psychological laboratory. What social psychologists normally investigate experimentally are the effects on individual psychology, including thinking, of social experiences, often contrived ones. How Loglan might be used in such a setting to study the Whorfian effects, if any, of second-language learning, is a complex matter that deserves and gets a separate chapter in this book, namely Chapter 7. But there are now other kinds of scientific laboratories in which Loglan might prove useful. One of them is the artificial intelligence laboratory. Let us consider briefly how Loglan might be used there.
Artificial intelligence (AI) workers have applied computer-modeling techniques to a broad range of intelligent performances, i.e., those that apparently involve a considerable amount of internal processing before some relatively simple but highly flexible output is produced...for example, a chess move. These phenomena now include the diagnosis of diseases, the generation and proving of logic theorems, the playing of games like chess and backgammon, and some apparently automatic behaviors which, like human pattern-recognition, also appear to involve intelligent processes. We may expect in time that the AI technique will be applied to itself, i.e., to that intelligent process of which the output is a computer program; see Section 1.5 on the machine-man interface below.
In all AI studies the principal tool is a computer program. The investigator's goal is to write a program that will cause a computer to produce outputs--medical diagnoses, moves in chess, proofs of theorems, discovered patterns--that imitate with reasonable fidelity the performances of some intelligent organism given the same inputs, usually man. The most commonly accepted rules of AI research are (1) that the inputs given the computer be strict analogs of the inputs received by the performing organism, insofar as this is known, and (2) that the outputs produced by the computer be strictly comparable with, or even as good as or better than, the organismic performances. What goes on between input and output is up to the investigator. To produce its imitative outputs the program may or may not be provided with a model that imitates some suspected natural process or tests some theory. In any case, it is by altering the program, and by observing the consequences of each alteration on the computer's output, that the AI investigator slowly improves the match between its output and whatever is known about the living organism's range of responses. Just as similar computer models are now being used to study weather systems, galactic evolution, plate tectonics, and other complex natural processes that are also difficult or impossible to observe directly, so AI models are being used to study intelligent processes. The aim, of course, is ultimately to understand ourselves.
Loglan could contribute to that understanding. Being a fully described and demonstrably speakable human language, Loglan can provide AI investigators with at least the beginning of a model of how human speech generation and understanding works. Loglan grammar is not only known but already written in a machine-parsable code. So it is itself the beginning of an AI program. Also, like all formal grammars, the current machine grammar of Loglan is infinitely malleable: its rules can be rewritten in an unlimited number of ways that will still parse the same sentences. So a wide variety of hypotheses about how human speech is internally generated can be tested by a grammatical model simply by altering its rule structures in systematic ways. Such alterations could continue until clues about what improved the model's fit with human speech began to emerge.
We do not yet have recorded output from fluent Loglan speakers to supply AI laboratories.20 Besides, it is likely that AI workers interested in testing detailed models of the speech generation process will prefer experimentally induced bodies of verbal output to the records of spontaneous speech that we have collected for linguists. But in the next few years a generation of truly fluent speakers of Loglan is very likely to emerge. Enlisting some of them as subjects, evoking speech from them under controlled circumstances and measuring the physical parameters of that speech--the pace and pattern of its hesitations, the locations of its blurts and pauses, its variations in tone and rate, and, as controls, the measured response strengths of the various individual words in their vocabularies--could easily provide a human standard against which the infinitely manipulable AI output could then be modeled.
Why could not such an investigation be conducted now with natural speech? Because the first complete grammar of a natural language has yet to be written. If we are interested in matching human speech in any language with the output of a computer model, that model must contain not only some analog of the human speech generator itself, but also an analog of the particular language in which the speech is occurring, i.e., a grammar of that language. If the grammar we install in the speech-generating program is incomplete, then the model will be incomplete; it cannot be expected to account for the full range of speech productions in that language. Human speech is essentially spontaneous. To constrain it, say to some subset of a language like English for which one thought one had a partial grammar, would be to turn it into something else, something that was not speech. Besides, if the partial grammars now in hand are any indication, when a complete grammar of a natural human language is finally written, it will be far too large for programmatic manipulation in the AI lab. Natural languages are very large affairs. Like whales and elephants, they make poor experimental animals.
Thus, more than anything else it is the small size, formal completeness and machine parsability of Loglan grammar that seem to suit it for manipulation in the artificial intelligence laboratory.
Another possible computer-related role for Loglan Is at what is called the "interface" between humans and their computing machines. At the moment this interface is occupied by a wide variety of computer software, very little of it succeeding in making the use of computers comfortable for the uninitiated. Despite a considerable commercial effort to make this software "user-friendly", the art of using computers remains an arcane one, full of magician's mumblings and sorcerer's symbols. Much of the arcanery appears to have arisen from the incorrigible literality of the machine coupled with the passion of professionals for compact code. The result is a frequently absurd situation in which a forgotten semicolon or an unmatched parenthesis can mean the loss of a night's work.
Ideally, as all the science fiction has it, one should be able to talk to one's computer: to put questions to it, to clarify one's purposes, to ask what's wrong when uncertainties over semicolons arise, to discuss with it the various ways in which a problem might be approached, to make choices among the sets of strategical alternatives it might present to you, or to answer questions about its understanding of your problem or about the units or degree of precision in which you wish the solution expressed. In other words, to make the machine-man interface truly comfortable for humans and yet continue to be instructive for machines, we need a language in which the requirements of both humans and machines are met.
Loglan may be such a language. We have seen that it is utterly unequivocal grammatically. One consequence is that we humans become aware of what we are actually saying when we talk Loglan. So a Loglan-speaking human is much less likely to say one thing while meaning another, thus misinforming his or her machine. Also, as we shall see in the next chapter, Loglan words resolve uniquely from the speech-stream; no 'I scream'/'Ice cream' phenomena exist in it. So even spoken instructions are unequivocal in Loglan. This is true of no other language. Being able to speak freely composed instructions spontaneously would add immeasurably to the speed and comfort of the interaction for humans, and yet, because it's Loglan, its being spoken would not diminish its precision for machines.
What would guide such a spontaneous--one might even say unstructured--interaction? From the computer's point of view, its sole requirement is eventually to understand exactly what you want it to compute. So it would first parse your sentences, look up the meanings of your words in a shared dictionary, make inferences from its understanding of your message, check them out against a shared knowledge-base, and finally check them out with you. When satisfied from your answers that it had understood your problem, it would translate the set of Loglan instructions it had received from you, and thus verified in this way, into some programming language or languages that it judged suitable to your task. In other words, it would program your problem. Perhaps it would have a set of computer languages, and libraries of stock programs in these languages, from which to choose. The computer would in effect program itself to execute whatever instructions it had elicited from you in the interface language.
What do we human partners in this high-powered interaction require? That we be permitted to express our thoughts fully, freely and spontaneously without the risk of seriously misinforming our machines. That we be able to understand most of the machine's word-choices and all its utterance-forms immediately, and be able to clarify by interrogation whatever part of the computer's responses to us we do not immediately understand. That we be comforted by the knowledge that the machine is also attempting to understand us, and is using similar inquiries into our meanings and usages in order to do so. That all relevant knowledge be accessible to us as well as to the computer during the problem-defining session. And finally that we be able to make new bodies of relevant knowledge accessible to the computer when we discover that there is something it does not know that we think relevant to our joint task.
Loglan is peculiarly well-suited to the dual role that would have to be played by the interface language in so rich an interaction. There appears to be no practical limit to the human thoughts or kinds of human knowledge that can be expressed in it despite its very small formal size. Yet it is that small size coupled with formal completeness that, as we have seen, makes it understandable to machines. Although its grammar is small and simple, the semantic domain of Loglan is immense. Unlike ordinary programming languages, which have severely limited vocabularies, Loglan will soon have, and in principle has already, a full-sized human lexicon. So from the human's point of view, anything that comes to mind can be said. Moreover, and unlike the subset natural languages now in use at many interfaces--for example, Subset English--the Loglan-speaking computer user will be able to use the full resources of his or her second language. Whatever the human says grammatically the machine with access to that lexicon will understand.21 At the same time, the predicate calculus on which the basic grammar of Loglan has been built will provide a way of both storing human knowledge and making it useful to both machines and humans after being stored.
Despite its formal suitedness, there are at least two commercial barriers to Loglan's being used at the machine-man interface for some time. One is the financial risk of building an interface around a language of which there were at the outset very few--even a few thousand--speakers. The other is the cost to the non-Loglan-speaker of preparing him- or herself to use such an interface, namely learning a whole second language. Machines can learn Loglan instantly; humans require time. So until the number of humans who have already invested time and money in learning Loglan has grown substantially--and presumably they will have done so for some other reason--it seems unlikely that any but a public agency would risk an undertaking that required such large investments on the part of potential users.
It is true that the dream of using Loglan to talk with computers may itself be a powerful motive for learning it. But it is a dream that is not likely to be realized until a great many others have also been motivated to learn it. So it is good to consider that there may be other motives for learning Loglan. Once a population of Loglan-speakers exists, however, and whatever their reasons were for learning it, there seems to be little doubt that the formal suitedness of Loglan to enabling machine-man interaction--a property which it may well possess uniquely among the spoken languages of the world--will eventually commend it to those concerned with the integration of computers into human life.
Let me mention briefly yet another possible use of Loglan, one for which it also seems peculiarly well-suited, and for some of the same reasons: that is as a medium of translation among natural tongues. Such a medium would be useful to an international agency charged, for example, with disseminating scientific or technological information which had originated in one language into many others, the United Nations, for example.
Consider the problem. An original document, say a French article on galactic evolution, is to be translated into a dozen other languages, from Chinese to Swahili. As this project would be implemented now, it would turn into a dozen separate translation tasks, each performed by its own bilingual expert, or team of experts, if as many as a dozen could be found. But with Loglan as the translation medium, the project would be transformed into essentially one task: translation of the French document into Loglan. Admittedly this would require human effort aided by whatever computer algorithms the agency had developed for this purpose. But the resulting Loglan document could then be more or less instantly retranslated into almost any number of other natural tongues, and this second step could in principle be performed, and so eventually in practice, by machines.
The reason for the asymmetry is plain. Loglan is syntactically unambiguous; the other languages are not. Therefore, once the sense of any document has been satisfactorily rendered into Loglan--once its metaphors and idioms ("idiotismos" as the Spanish wryly call them) have been transformed into literal Loglan by human workers who know both languages well, and once the most plausible interpretations of its syntax have been settled for--then the resulting document will be unambiguous in every sense of that word. Not only will the Loglan word-meanings be literally translatable into other languages--as is unlikely to have been true in the source language--the document will now be expressed in a language that is syntactically unambiguous as well. So it will now be ready for machine-translation into other tongues.22 No further human work will be necessary. Any computer with access to a Loglan grammar and lexicon will know what the Loglan document is saying; so given bilingual dictionaries and even partial, translation grammars of other languages, saying it again in one of those other languages will be a task that computers probably can perform. If they can, then translations of the Loglan document could go out on as many different natural language wavelengths as we had bilingual dictionaries and translation grammars for.23
Note that none of the ultimate consumers of this international translation service need know a word of Loglan. What Loglan has made possible, and what the human translator into Loglan has supplied, is a clear statement to the Loglan translating machines of what a certain source message probably means. In a certain sense Loglan has enabled the human translator of the original document to disambiguate that message, or it has forced him or her to do so by its requirement that, in Loglan at least, he or she be syntactically clear. Once this has been done, it is easy for the second order machines to re-express the now-unequivocal Loglan message in some relatively unequivocal words and phrases chosen from the target languages. Such unequivocal expressions are not always pretty. The English, for example, that results from this sort of automatic translation out of Loglan is peculiar English, bristling with curiously logical phrases like 'the mass of all ...' or 'the event, state or condition of ...', but it gets the job done. What is most striking about this kind of English prose is that it is crystal-clear. It is in fact "Loglanized" English. Reading it not only teaches one a lot about the Loglan way of seeing the world, just as Trobriandized English would teach one about Trobriand metaphysics, but Loglanized English also teaches one about English metaphysics. For it makes explicit what we English-speakers normally take for granted about the world. Now to make any metaphysics explicit is in some curious way to make it non-metaphysical; it is to deprive it of the protective inattention that metaphysics normally enjoys. Assumptions that have been made explicit are no longer quite assumptions. They are propositions to be rolled over in the mind.
Thus the covert process of translation into and then out of Loglan may have a small but interesting side-effect on those who read the "loglanized texts" that issue from it. They may become aware, perhaps quite incidentally, of what was formerly metaphysical in both their own language and the language of the source document.
Another not quite so incidental by-product of using Loglan as a translation medium would be that the Loglan texts so created would be well-adapted for the machine storage and retrieval of the information they contained. For one of the same reasons that Loglan Is suitable at the interface, namely that knowledge stored in the predicate notation is apparently usable by both machines and humans, texts translated into Loglan and stored on some electronic medium could later be searched and even studied by machines. The studying Machines would be computers "trained", i.e., programmed in the AI style, in the human art of scholarly reading. Although key words and Phrases can be searched for now, and in texts written in any language, natural language texts cannot yet be understood by computers in this way.
Once again Loglan yields a special benefit because its grammar is transparent and its meanings clear.
There is yet a seventh way of viewing Loglan. In fact, a substantial minority of those who are already interested in Loglan24 see it as a lively candidate for the universal second language. Everybody who has thought about the matter seems to be convinced we're eventually going to need one on this planet, and a constructed language with the promise of political and cultural neutrality, as well as Loglan's apparent aptness for computer talk, seems a particularly attractive prospect for such a role. People in the "auxiliary language", movement, as it is often called, insist that the new second language would not replace any existing tongue. Indeed, its very existence might help now-threatened minor languages to survive once a powerful second language were available to its speakers for communication with other folk. At first the international auxiliary would serve as everybody's communication channel with travelers or visitors, i.e., with everyone with whom one did not share a native tongue. Later it might be used preferentially even in groups that did share a native tongue if the topic being discussed--such as travel, technology or science--was more advantageously addressed in Loglan than in their common natural language.25 Although Loglan was not designed for this bright future, it may nevertheless have attributes that fit it for the job.
Some might say that the still-divided state of the world militates strongly against the adoption of an international second language even if we had a number of good ones to propose. Quite so; but the world is a shrinking place. It seems likely that within the next fifty years or so the desirability, even the practical necessity, of an international second tongue will become apparent to all those charged with the education of the world's children. This is not the place to discuss the merits of the many plans that have been proposed to meet this contingency should it arise. Not even the basic division between those who favor the teaching of a constructed language, such as Esperanto or Loglan, and those who favor the use of a widespread natural one, like English, may merit more than passing attention at this time. Even so, it is quite possible that, by the time this contingency does arise, research with Loglan, or with other Loglan-like laboratory tongues, will have disclosed certain large, unrealized functional potentials in the human language faculty that will settle the argument decisively in favor of a deliberately constructed international tongue. It is also possible, of course, that it will not.
Even supposing that the most detailed confirmation of the Sapir-Whorf hypothesis had taken place by the time the moment of decision arrived, I at one time doubted that a language of Loglan's severely ratiocinative cast, as I then thought of it, would prove the best choice. An international language, above all languages, it seemed to me, would be called upon to perform a multiplex function. Loglan was, I reasoned, a laboratory instrument: designed to optimize linguistic performance along a single, if important, functional dimension: rational thought. It was, in short, an uncompromising tool. it was even possible that the uncompromisingly logical character of the language might have some interestingly negative effects on literary productions in the language; see Section 1.3. In Loglan, as I was fond of putting it then, one is forced to be clear.
I now believe this view to be mistaken, and that the fear it leads to is quite groundless. In the last ten years I have made some discoveries about Loglan, perhaps also about the language arts in general, that persuade me that Loglan's "severely ratiocinative cast" would not unsuit it to be an international auxiliary tongue for everyman. In the first place, Loglan does not wear its logical dress incessantly; it displays the propositional calculus only when its speakers wish to display it. In the second place, it turns out that the soul of Loglan is not its logicality but its optionality. It is as optionally illogical as it is logical. It is, in short, whatever its speakers wish it to be.
What has happened in the last ten years is that I have spoken a language that was designed from the ground up and heard it spoken; I have observed the struggles--quite common in second language learners--to convey old messages over new channels, to pour old wine into new bottles without spilling it. I have examined the sometimes poetic and often frolicsome literary inventions of my fellow loglanists...especially when they were not decanting old semantic wine but being outrageously inventive and loglandical in their new language. Above all I have come to realize that while a new language may liberate inventiveness in surprising ways, no language forces anything but its obligatory grammatical arrangements on its speakers.26
Not even clarity. Clarity is there in Loglan if you wish it. And clarity is genuinely more available to the speaker of Loglan than of other languages. If you want to be clear in Loglan, you will not, for example, be tripped up by the massive, and largely unnoticed, ambiguity of your mother tongue, which I presume is English. But in Loglan, as in other languages, one is not obliged to be clear.
Can one, in Loglan, be unclear? Of course. One can use the same pronoun twice in the same utterance with different referents. One can use an argot word with listeners who do not know the argot. (Who could guess, for example, what modern particle physicists mean by 'charm'? Translating it into Loglan will not suddenly make it clear.) Or one can inadvertently use a bad metaphor, one that has little chance of getting one's message across. It is true that the grammatical structure of one's bad metaphor will be clear to one's auditors in Loglan; that is, the order in which its parts were joined by you will be known to them. And this will help those auditors try to decipher your bad metaphor. But it does not guarantee that any of them will ever succeed.
In these last ten years of engineering optionalities into Loglan I have formed a view of language which I suspect is very different from the one with which I began. I see language now as something very like a sign-painter's kit: a box of brushes, paints, templates and other tools which each of us carries about. With one's own personal sign-painter's kit--one's personal collection of English sign-parts, say--one constructs one's signs...in the air. One speaks English. Others similarly equipped with English sign-making tools, and so familiar with their use, listen to you and attempt to decode your signs. If you have labored well, if you have constructed your sign intelligently and taken carefully into account how those particular others are likely to interpret your efforts, your sign may very well succeed in doing what you wanted it to do: it may sing of your intentions to them. If not, it won't; and you may never know that you have failed.
If this view of language as a collection of sign-making instruments--acquired in bulk during one's childhood but which one never tires of refurbishing in detail--is generally correct, languages do not force anything on anyone. Particular brushes or colors or templates do not leap out of the box into your hand. You pick them up, use them in your often quite idiosyncratic ways, and, when either satisfied with or defeated in your efforts, you put them back down again. Others may "look at" your finished air-paintings, your utterances; and, if they choose, they may comply with or respond to whatever message you have succeeded in conveying to them. They may answer your question (if they understand it), or some other question that they thought you asked (if they didn't); but they will do all this as voluntary agents. They will not be responding to irresistable "forces" any more than you were. The language act, in both directions, is a totally voluntary, creative affair. It is a painting of signs in the air. And the watchers are free to invent for themselves what you mean, and act on it or not, as they choose.
Where does the Whorfian sense of our being limited by particular languages come from, then? It comes partly from the simple matter of conforming to their obligatory grammatical arrangements if we want to be understood. But it comes also, I suspect, from deprivation: from not having in your native language kit what you might, in some vague way, be ready or yearning to use. If you don't have mauve in your sign kit, you can't use mauve to paint a sign. The musical genius born on the ice floes off Northwestern Greenland will never make the music he or she was, as we Europeans might say, "destined to make", because there are no grand pianos or other musical implements for miles around. In a precisely parallel way "illogical" languages lead to illogical speech because they don't make the machinery of logic easily available to their speakers, and so they don't direct the attention of their speakers to the logical dimension of their ideas. If this is so, then "logical" languages are those which do make the full machinery of logical analysis easily available to their speakers, and so do direct the attention of their speakers to the validity-conserving dimension of their own thinking. But the point is that no part of that machinery ever has to be used. We all know people who have never said--and in the remainder of their lives will never say--that something is A "if and only if" it is also B. That particular shade of logical English lies unused in their sign kits. It is there. Each of the words is there. Even the usage is there in the kits of other modern English speakers to be understood and copied if they choose to. But they don't choose to. Nothing except a stern schoolmaster could force them to.
Apparently no piece of the linguistic apparatus has to be used. English and the other Indo-European languages suffer, in the Whorfian sense, from exactly this kind of limitation of their tense systems (to take just one example). We do not have a "tenseless tense" in these languages. Nor do we have an "epistemic" one, like Hopi. So we make do with the present tense...or with our 'Probably's and 'Certainly's. We say 'Socrates is a Greek philosopher'. We pick from the sign kit a sign that is usually used to mean something else, and with the help of our infinitely cooperative auditors, we make the present tense serve as the eternal tense that we do not have in English. In the same way that we can use screwdrivers as tent-pegs and jackknives to drive screws, so we are free to use any part of our languages in any ways that work to convey our messages.
What is the bearing of this rough, tool-kit theory of language on the possible aptness of Loglan as an international auxiliary tongue? The bearing is this. Loglan takes limits off; it has not applied them in now directions...not even logical ones. All the logical apparatus of Loglan is optional, just as its tense system is. It also has an optional case system, which, paradoxically, means that we may also speak without cases if we choose. Its standard (unmarked) word-order is Subject-Verb-Object (S-V-0). But every other possible order of the main ingredients of a sentence may also be used. Such a boundless language, full of optionalities, is precisely what will be wanted in everybody's second tongue. For every user of it will have a native language which will have put, if Sapir and Whorf are right, the first impress on his or her mind. If that native language has cases, he or she will want at first to use cases in his or her second language, too; if it uses a Subject-Object-Verb order in subordinate clauses as German does, then in such places he or she will want to use that order in the auxiliary tongue as well. Loglan will allow all these variations. And it will make them mutually understandable to persons who come from very different language backgrounds, precisely through the agency of their invitingly boundless second tongue. Loglan, in short, is a kind of linguistic no-man's land. It therefore belongs potentially to everyman. While I once had some doubts about the utility of Loglan as an international auxiliary, I would argue now that it is precisely the boundless optionality of Loglan that makes it especially well-suited to be everybody's second tongue.
Loglan will have one doughty competitor for that role when the occasion for choosing a universal second language comes around, namely Esperanto. Much older than Loglan, and the only constructed human language that has ever successfully spread beyond the borders of Europe, Esperanto has already captured the hearts of many of the internationally inclined. Loglan, as the newcomer, will have to acquire its adherents in a different metier, probably in the laboratories of science. So it is likely to have a very different kind of early adherent. It may well continue to draw them first from the world community of computer professionals, the programmers and engineers, or from artificial intelligence workers, or from the linguistics labs and anthropology departments where Whorf-talk is still spoken, and, were Loglan to prove useful as a medium of scientific translation, it may ultimately draw adherents from the international community of science itself.
But Loglan must not let itself be confined to such communities of intellectuals, or let them dominate any more of its affairs than are directly relevant to their concerns. If the language is to grow and spread, to make good its promise as a potential world auxiliary, it must have the support of plain people as well, and of the politicians and educators who will ultimately come to speak for them.
I have been making the tacit assumption all along in this section that Sapir and Whorf will turn out to be in some sense right, that Loglan will have a special mind-expanding contribution to make as the international auxiliary. But suppose Sapir and Whorf are not right? Suppose particular languages do not set limits on human thought? Then it hardly matters what language our children choose for our grandchildren to talk in when they travel. Thought is; and thought will be. For the commanding alternative hypothesis is that the forms of reason are, as Kant surmised, among the biological givens of the human animal.27 In that case, such incidental features of a language as whether it is easily learned or not, or speakable by and to computers, or useful in translation, might well turn out to be decisive. So, ironically, Loglan may turn out to be the preferred alternative after all...even if the hypothesis that led to its construction is refuted.
There is a final way in which Loglan may be viewed which I wish to acknowledge briefly but candidly. This is the perspective from which Loglan is seen by many individuals, not as a research tool, not as contribution to the machine-man interface, not as a candidate for the international auxiliary, but as a delightful and very human toy. Perhaps nothing is so playful in human behavior as language play. And one of the language games we play is building languages. Logicians do it; mathematicians do it; poets and children do it. Indeed the whole, long, outwardly solemn history of international language building may be in no small measure an expression of language play.
Moreover it gives me no pain to confess that building Loglan was also fun. And now that it is done, or nearly done, I am pleased to say that its smallness of scale makes it as apt for recreation in the living room as for manipulation in the laboratory. Learning any language is fun, and apparently the odder it is the better. Loglan is odd, no doubt about that; and many of the effects of learning Loglan are subjectively immediate and compelling. Whether those effects prove to be permanent or not--which is to say, whether Whorf is right or not--it is apparently fun to lose your English-speaking mind. There is, of course, nothing either strange or illegitimate about this playful use of Loglan. A telescope is a source of delight as well as a tool for collecting scientific facts. It is not surprising that putting on a new pair of linguistic lenses should also furnish a measure of delight.
This book is quite openly addressed in part to those who are prepared to regard learning Loglan as pure fun--perhaps even useless fun--as well as to my more utilitarian scientific colleagues and to my fellow partisans of the international language movement. Many if not most of those who have written me about Loglan in the years since its earliest publication have quite plainly placed themselves in the first attractive category. To them, and to those others who may yet find joy in it, I confess my own delight--no matter what may be the outcome of my scientific labors--at providing them with an engaging toy. Besides, the first speakers of the language have come from just these playful learners. They and their motives are therefore no less important for the eventual capture of the Whorfian phenomenon--if Loglan proves to be its instrument--than the sober experimenters who may one day seek them out.
In the rest of this book I propose to treat Loglan as if it were a living, spoken tongue. Strictly speaking this is not quite yet the case. While sustained speech in the language has occurred, and while many of the 3,000 buyers of the 1975 books also learned to use the language in some way, that was some years ago. The engineering interest has been the dominant one in the Loglan community since the early 80's. But now, with the engineering phase of the project finally over, the language is ready to be learned and used...in some cases, as in my own, relearned and used again.
Learning Loglan was, when it first happened, a bootstrap operation. Its early speakers had to simulate its speechways, and these simulated bits of fluent speech then served as stimuli for the free response of others. Little by little the necessity for simulation receded. When it was completely gone, when veridical Loglan speech began passing spontaneously from speaker to speaker, as it did in the rooms of The Loglan Institute in 1977 and '78, the first recorded demonstration of a wholly synthetic linguistic phenomenon had taken place.28 I had hoped that thereafter there would always be genuine speakers from whom others could learn the language in ways more natural to the race. But engineering soon replaced speaking among those early speakers, including myself; our skill died back from want of use; and that first generation of loglanists are now very largely incompetent again. This is because the successful re-engineering of Loglan augmented the language hugely, especially its morphology; and there is now much that is new to be learned.
I have no doubt that new masters of the language will quickly emerge with the publication of these 1989 books and their accompanying cassettes and teaching programs. Already groups exist that are waiting impatiently for these new materials to be issued. By the time you read this book, a new generation of Loglan speakers are likely to have leapt into the world. Still, in writing now about their "speechways" I confess I am engaging in a genial deception. The behavior and insights I will attribute to "speakers of Loglan" in the remainder of this book are partly based on what I learned about the Loglan-speaking process in the period 1975-80 when the Loglan world was new. But partly they have been derived from the Sapir-Whorf hypothesis, which as everybody knows is still untested; the crucial experiments have not yet been run. We do not yet know what really happens to or in the heads of those who speak Loglan. Yet I have found the pretense of foreknowledge a useful, even an indispensable expository device. So what I will describe is the working out of at least one set of consequences of the Whorf hypothesis as it applies to Loglan. They will be tentative ones, to be sure. But one cannot, after all, entertain incessantly the possibility that Sapir and Whorf were wrong. Yet obviously they may be. To say so one more time is probably enough; the reader is forewarned.
The rest of the book is straightforward exposition. I shall first describe the sounds of the language. This will take only a few pages, for there are not many sounds. Then I shall enumerate its word-forms, and these, too, are very few. We turn next to the three departments of Loglan grammar: its "predicate" forms, its "argument" forms, and the forms of its "utterances"...all terms which you will presently understand. These three grammatical chapters will comprise the bulk of the book. Then I shall describe the provisions that have been built into Loglan for its future growth. Then a summary chapter on the Whorf hypothesis and some illustrative appendices will bring us to the end of the book. The reader will find among these appendices a glossary of the primitive predicates of the language, a list of its little words and affixes, and one or two grammatical paradigms.
This, obviously, is not all there is to a language. It is not even sufficient to learn a small one well. But we plan to accompany this 1989 edition of Loglan 1 with a set of teaching programs as well as audio casettes. The former will teach word recognition, construction and decipherment as well as the formation of simple utterances; and the latter will exhibit the speech sounds and utterance rhythms of the new language. A second volume, Loglan 6: Formal Structures, is planned for publication in 1990. This volume will give the technical user of Loglan a detailed formal exposition of all the lists, rules and algorithms that constitute the structure of the language. Then, if funding permits, we will staff the revision of the 1975 dictionaries and publish the Third Edition of Loglan 4 & 5: A Loglan-English/English-Loglan Dictionary.
A word about how we plan to finance these and further publications of The Loglan Institute. A certain proportion of the dues of those who elect to become members of The Institute--currently a large proportion, but it will be smaller as membership grows--will, be used to maintain the essential services which must stand behind any new language: the continuous update of its grammar, and, in our case, the maintenance of the unambiguity of its grammar; lexicon expansion and the continuous update of our master dictionary files; the building of a library of Loglan texts as these are generated by Loglan writers and translators; and the periodic publication for the Loglan-using public of new or updated printed materials such as primers, readers, dictionaries and reference manuals. The Institute invites readers to help provide these services to the Loglan-using by becoming members. Members will also enjoy, of course, the more usual benefits of membership such as interaction with fellow loglanists on topics of shared interest, as well as being regularly apprised of new publications and of the growth and spread of the language.
1 See Sapir (1921) and Whorf (1956), the latter a posthumous publication. References throughout this book will be made by author's name and parenthetic date of publication and under this heading complete particulars will be found in the Bibliography.
2 It should be pointed out that not all linguists agree with Sapir and Whorf's analyses; see, for example, C. F. Hockett, "Chinese versus English: An Exploration of the Whorfian Theses," (1954). The exceptions taken by Prof. Hockett were fairly typical. From his examination of Chinese and English he argued that their structures affect, not what can be said at all in these languages, but what can be said easily. He therefore disagreed with Whorf's thesis of the existence of a fundamental linguistic constraint on thought. But like most early criticisms of Whorf, Hockett replaced the rather simple, deterministic mechanisms Whorf supposed to be at work with more sophisticated, probabilistic ones which are, of course, equally hypothetical. What is wanted is experimental isolation of these mechanisms; for as Hockett himself points out, the Whorfian phenomena, if they exist at all, are bound to be particularly difficult ones to observe in situ.
3 Summaries of this early and generally confirmatory work are to be found in Brown and Lenneberg (1958) and Carroll and Casagrande (1958). In general, it seemed to show that intra-cultural perception, at least, tends to follow grammatical, morphological and even phonological lines. The most important apparent disconfirmation of the hypothesis to date is probably Osgood's (1960) finding of a large cross-cultural similarity in semantic response despite large linguistic differences. None of this work confronts Whorf with his largest claim, of course--that language furnishes both the mechanisms and the boundaries of thought--and therefore none of it can claim to be decisive.
4 A symposium on the scientific strategies available to linguists for dealing with the Sapir-Whorf hypothesis was held in the early 1950's and reported in Hoijer (1954). None of the members of this symposium, however, entertained the direct, experimental approach prepared for here. One gathers that for most anthropologically trained linguists languages, like mountains, are inimitable: a part of the extra-human macrocosm that one must be content to observe. Yet, questions of mass aside, a large, if as yet undetermined, part of every language is obviously artifactual, if not quite artificial, and hence as alterable as a city street. The interesting modern question, of course, is exactly which parts of language are artifactual, and which are not.
5 This is not quite true any longer. Semantic structures are currently being given a good deal of attention by linguists. See, for example, the "deep semantics" hypothesis of Geoffrey Leech (1981).
6 By 'second' I really mean 'second or subsequent', that is, second, third or fourth, etc. But this long phrase is awkward to reuse. It's a pity we're not writing and reading Loglan, in which the word for this important and surprisingly frequently used notion is sutori, a word which breaks up as su ('at least') + to ('two') + ri ('-th'). So what we are talking about in this section are, in loglanized English, "at-least-second" languages.
7 We shall have more to say about positive versions of the Whorf hypothesis--various ways in which languages and written notations may be suspected of facilitating, or even enabling, certain kinds of thought--in Chapter 7.
8 For example, Bishop Wilkins' 17th Century project to construct "a philosophical language." Wilkins' efforts were preceded by the Ars Signorum of the Scot, Dalgarno, and were shortly followed by Leibniz' own efforts to construct an "algebra of thought." These, and other early essays in language planning, are described in Bodmer(1944).
9 Or at least the justifying arguments that accompany the results of thought, for certainly no logician would claim these days that he is engaged in the naturalistic modeling of human thinking. Even so, the conduct of a justifying argument is a kind of thought, even a very important kind, and for this the logician's formal reconstruction of the most direct routes over which thought might proceed provides a very interesting, and functionally unequivocal, paralinguistic model.
10 At least in English. Sometimes calling themselves the "ordinary language philosophers," this group is largely British and largely centered at Oxford; among representative works see Strawson (1952) and Ryle (1962).
11 For example, 30 days of an hour a day of vocabulary drill with flashcards plus somewhat less time spent in passive listening to cassettes seems to have been sufficient to prepare the four participants in the 1977-78 apprenticeship program for a month of face-to-face instruction in speaking the language. By the fifteenth or twentieth day of such instruction, three of these apprentices were sustaining Loglan-only conversation for 45-minute periods daily. These are the only hard data we yet have on the speed with which Loglan can be learned; and they are obviously strongly biased by the unusually high motivation of these early post-1975 subjects. Many other loglanists have, of course, learned to use the language in less intensive ways since then; but we do not have similarly quantitative records of their experiences. Data on this and other learning questions will soon be supplied, however, by our present, computer-based methods of recording the Loglan learning experience. These new methods will allow us to monitor the learning experiences of volunteer subjects even at a distance.
In connection with learning speed, I have made one observation that shows that the speed with which a language-like system is learned cannot be a simple function of its size. My earliest efforts to teach Loglan, in 1957, gave me definite indications that the language was then too small to be learned. Its then very small size--400 words and approximately 100 rules--seemed to retard, rather than facilitate, its learning. (Like a pea in a shoebox it seemed to rattle around in my subjects' heads.) This observation supports the current view that the organizational structure of language is innate, that whatever is learned as a language must first of all be a language. No such difficulties have arisen in recent teaching trials. Evidently something happened to the structure of Loglan between 1957 and, say, 1977, when the first sustained speech in Loglan was observed, that made the 1977 language better suited to whatever may be the "hard-wired" features of the language acquisition device in human heads.
12 For example, the Harvard translation grammar of English had well in excess of 6,500 sub-rules in the last report I received: Kuno and Oettinger (1963). The fewer than 200 sub-rules cited for Loglan are from the current (June 1988) formal grammar described in Notebook 3 (Brown 1987), a book that will be replaced by Loglan 6: Formal Structures (Brown and McIvor, forthcoming).
13 The number of syntactic categories in the Harvard grammar; Kuno and Oettinger (1963).
14 1 am using here Whorf's sense of the word 'metaphysical': the sense in which a particular view of the structure of reality is "forced" on the speakers of a language by its obligatory grammatical arrangements. Of course Whorf cannot mean literally forced; since we escape these metaphysical "obligations" by several routes, among them philosophical analysis and learning second languages.
15 The word 'primitive' has been adopted from formal logic where it means one of the undefined terms of a deductive system. In contrast, a "primitive predicate" of a given language is one that is underived within it, that is, a word that is derivationally simple in that language. The point here is that there is a crucial difference between definition and derivation, in that defined terms are susceptible to systematic logical transformation and derived terms are not. The use of the word 'primitive' in this book may, therefore, tempt the sophisticated reader to expect more of them than he will get.
16 These procedures are described from the standpoint of the word-maker in Section 6.5.
17 One reason for this may be that languages, unlike the more consciously acquired parts of human culture, evolve mostly by accretion, only occasionally by sloughing off. Thus a bewildering variety of grammatical arrangements co-exist for engaging the same point in the semantical field; and this is so despite the fact that each once-new grammatical arrangement was presumably devised by some brave speaker to engage a new point in that field. But it inevitably does so by redundantly covering old points as well. Loglan, having been constructed all at once, can have less grammatical overlap in covering the same field.
18 The Installation of this property in the language was a two-stage process. Unambiguity was closely approached by a heuristic procedure in the years 1962-64. During those years I used systematic computer searches of the utterance domain of the language to locate and remove ambiguities. I carried on this work, experiencing lower and lower discovery rates, until the rate at which ambiguities turned UP on these searches fell to, and finally remained at, zero. Later, Aho, Johnson and Ullman's (1975) discovery of an algorithm for demonstrating conflict-freeness in certain classes of formal grammars allowed me and several co-workers to demonstrate that a formal analog of Loglan grammar was unambiguous in February 1982. In the course of obtaining this proof we discovered and of course removed the remaining ambiguities in the language (Brown 1982).
19 A possible second reason why natural grammars are so much larger than they apparently need to be (see Note 17) has been suggested by our work in disambiguating Loglan. There may have been, and may still be, a conflicting, and so possibly alternating, pair of behavioral pressures driving the evolution of human grammars; see Brown and Greenhood (1985, and in press). One of these pressures is what we there call the burden of "incommunicable images": the body of complex images which the speaker can create in his or her own mind but which he or she cannot yet say to others with any hope of being understood. The other pressure is the "disambiguation burden" on the hearer: the number of possible, i.e., "grammatical", interpretations of a speaker's utterances among which his or her hearers are obliged to choose. The first of these pressures would lead human grammars to become ever more intricate, for they would do so as ideas which were once only thinkable became sayable. This would lead, in our model, to a temporary increase in the disambiguation burden; for any increase in the elaborateness of the grammatical apparatus is at first very likely to increase the number of legitimate interpretations elsewhere, largely as a consequence of the grammatical overlap phenomenon mentioned in Note 17. When, however, the disambiguation burden grows so large through these accretions as to become intolerable, this would lead once again to its reduction through either the further refinement of the grammatical grain of the lexicon ("lexemification") and/or to the addition of still more clarifying particles ("particlization"). These, we argue on the linguistic evidence, are the two kinds of structural devices by which human grammars have become more powerfully disambiguating. Indeed, their application to the reduction of the disambiguation burden is what may have created human grammars in the first place. Loglan, having been deliberately engineered, is syntactically unambiguous; so its speakers place no disambiguation burden whatever on its hearers. In effect, the hearer's burden has been reduced to zero. If zero-disambiguation can be maintained in Loglan despite vigorous and widespread street-use--and this is a question to which we must, as scientists, be very attentive--and our model of the grammar-evolution process turns out to be approximately correct, then the further development of Loglan may well be driven by the pressure on speakers alone: by the self-renewing burden of their incommunicable ideas.
20 We do have some recorded Loglan speech output, namely the recordings of the daily Loglan-only sustained conversation sessions held between me and my apprentices in 1977-78; see Note 11. But we were not yet fluent. So most of our pause-time was spent searching for words rather than in grammatical processing of complex sentences. Moreover, the grammar itself has changed in minor but relevant ways since that time. Still, it is a first step toward a recorded corpus of spontaneous Loglan speech.
21 Parsability does not of course exhaust the problem of meaning. What is left over is the meaning of individual predicates and the references of the designations in the parsed sentence. Where the Loglan lexicon is indecisive--and it is inevitable in a growing language that occasionally it will be--then the machine will be able to conduct inquiries: 'What kind of nucleus do you mean? Atomic? Or cellular?'...all carried out in Loglan, of course. The point is that such inquiries will take place within an implicitly agreed-upon grammatical structure. That is, the role of nukli in the Loglan sentence will be plain to both machine and human.
22 It was once thought that even ambiguous documents could be translated by machine-executed algorithms into other ambiguous documents, mainly by teaching the machine to thread its way through the maze of possibilities by using a kind of "plausibility calculus" as humans appear to do. But this effort, despite having been well-funded in the U.S.A in the 50's and 60's, was in the end abandoned. Natural language turns out to be too riddled with ambiguity, its multiword predicate expressions (often called "idioms") too metaphorical, and those metaphors too culture-specific, for purely machine-translation from one natural language into another to succeed. The human translator of natural language text can be much aided by machines; but apparently he or she cannot yet be replaced by one.
23 1 do not mean to suggest that the writing of such one-way "translation programs" out of Loglan into other languages would be trivial exercises. The writing of each program would be a major undertaking by Loglan-speaking scholars who were adepts in the given target language and in the sciences in which translations were to be produced. What I do mean to suggest, however, is that such undertakings, unlike the machine-translation projects of the recent past, are very likely to be feasible; since nothing stands in the way of expressing meanings more or less unequivocally--even in a normally ambiguous tongue--once the meanings to be so-expressed are known. For example, a Loglan-into-English machine-translation program already exists in skeletal form; it needs only to be fleshed out lexically for the various scientific vocabularies with which it might be used. The point is that once such one-way programs were available to the translating agency, their execution in a given case would be virtually instantaneous. In that sense their use would be a trivial step in the total information distribution task in which they were employed.
24 The prolegomena to Loglan were first published in the Scientific American (Brown 1960). This 16-page article generated a gratifyingly large and unexpectedly varied response, a substantial portion of which came from readers vitally interested in the international language movement.
25 The preferential use of one language over another according to topic often happens in the speech of genuine bilinguals when they are in the company of others known by them to be bilingual in the same languages. Usually speakers in such bilingual circles are quite unaware of having switched from one of their languages to the other; and their bilingual auditors are often similarly unaware that the switch has taken place, although both can easily remember such episodes in retrospect. If made aware, by an observer, that such language-switching has just been going on, both speakers and auditors report a general awareness that language-switching often happens in their circle; and they tend to explain the phenomenon as caused by what they believe to be the "functional superiority" of one of their two languages over the other for some topics of discussion, while the other language will be regarded as "superior" for other topics. This might be regarded as the banal sense of the Whorf hypothesis, the sense in which every bilingual traveller "knows" that it is true.
26 Let us distinguish between the way a language might be said to "force" something on a speaker, in the sense of compelling him or her to make a certain (therefore "obligatory") grammatical or lexical choice, and "inducing" something in him, in the sense of directing a speaker's attention toward, or creating awareness in him of, certain aspects of the speaking/thinking process, or of the world which is its target, and in that way cause some language options to become more frequently chosen than others are. What I am saying here is that Loglan (like all languages) will compel very few choices, but may induce a great many.
27 Curiously enough this view has become more, rather than less, plausible as modern research discloses larger and larger dimensions of the structure of language. For example, one argument for the existence of a large, innate component in the language structure is that the unconscious mass of orderly linguistic behavior is now known to be so large that no general learning theory can account for its acquisition. Besides, no society troubles itself to provide efficient instruction for its children in matters of speech behavior, though speech is the single most complex behavior that children evidently "learn." This is not to say that some people do not spend a good deal of time talking with their children and even correcting their speech behavior. But this is not instruction. It is providing data about the local language from which their children, like all human children, draw their own conclusions. Human children are apparently language-acquisition experts. What they require, and apparently all they require to learn a language, is a copious stream of the language itself. These modern observations have opened the door--so long-thought firmly closed--to the Kantian presumption that there are innate "forms" of "pure" and "practical reason"; and they do so by a curious twist of the Whorfian argument that deprives it of its relativity. We will return to this argument in Chapter 7.
28 There have, of course, been other sudden emergents. Over a hundred thousand people once spoke Esperanto--the number has diminished now to around thirty thousand, according to Forster (1982)--and there must have been a time when learning Esperanto was just such a "boot-strap" phenomenon as I have described. Secular Hebrew is also the result of a boot-strap operation in that apparently one family, among the early Zionist settlers in Palestine, resolved to use Hebrew and nothing else for all their daily needs; and from that center of necessitous invention the Israeli language as it is spoken today apparently spread. Speaking Loglan will differ from these other deliberate efforts to promulgate "new" languages in two ways: (1) its focus is scientific, and it is very likely, therefore, to be more closely observed; and (2) its newness is total. The unnewness of modern Hebrew is a complement of the fact that a sacred tongue existed. The vocabulary of that sacred language had to be rapidly expanded, but very little else. And Esperanto has a grammar that in all important particulars is "Standard Average European," as Whorf would say, and is therefore housed in the minds of its learners before they begin. Thus Esperanto was not cut from whole cloth and Loglan is. In this lies one, perhaps the greatest, of its scientific values.