The following text has only been published in the working papers PERILUS XII (1991), pp. 25-47 (Department of linguistics, Stockholm university).
Striking similarities can be observed between Grice's (1967) conversational maxims and Tauli's (1968) principles of language planning: In order to function well, a language must be such that it makes a well behaved conversation possible. Nevertheless, many ethnic languages as well as constructed interlanguages possess features which are incompatible with these principles. The paper contains an analysis of such cases: The compulsory expression of number and gender, which is in conflict with the principle of facultative precision; ambiguity and lack of distinctiveness in the names of the Latin letters; excessive length of certain morphemes; and restricted freedom to place sentence constituents in a pragmatically adequate order. The compulsory nature of certain distinctions, which cannot be understood on functional grounds, is ascribed to an excessive activity of "universal grammar", which is seen as one of the instincts of human beings. It is further shown that some constructed interlanguages (Volapük and Interlingua) clash with some of Greenberg's (1963) universals concerning the order of meaningful elements. While Esperanto is free from that type of deficiency, it is in conflict with the principle of facultative precision in the same way as most ethnic languages of Europe, as distinct from those of East Asia.
Johann Martin Schleyer's Volapük of 1879 and Ludwik Lazar Zamenhof's Lingvo Internacia of 1887, known as Esperanto, were the first artificial interlanguages that succeeded in attracting a considerable number of users. When these languages were constructed, the question how a language should be constituted in order to serve as an efficient means of communication can hardly be said to have been treated satisfactorily by linguists. In the 17th century, several philosophers such as Descartes, Leibniz, and Komensky had been concerned with the deliberate construction of a "perfected" language. These attempts resulted in the creation of several artificial languages, among which the philosophical language of bishop John Wilkins (1661) is the most elaborated example. The concern of those scholars was, however, not so much the ordinary communicative function of language, but rather its pedagogical use.
Discussions of the above-mentioned question, at length and/or in depth, started only subsequently to the spread of Volapük and Esperanto. Among the participants in these discussions philosophers, e.g., Louis Couturat, who had studied the works of Leibniz, were not absent, but there were also professional linguists, such as Otto Jespersen. In 1908, a reformed version of Esperanto, known as Ido, was launched by Louis de Beaufront and Louis Couturat. Although some of the changes that had been made improved the communicative function of the language, some other essential changes had been motivated on philosophical rather than on practical grounds. This applies mainly to the rules that govern word formation. In Ido, derivation has been made to follow more "logical" rules. Subsequently, René de Saussure (1915) investigated how it comes about that word formation in Esperanto works satisfactorily despite its alleged lack of logicalness and he formulated his principles of necessity and sufficiency, which say that what is clear anyway does not need to be expressed. This appears to have been the first explicit recognition of a difference between formal logic and the kind of logic that governs language use. A deeper understanding of the logic in human communication has been arrived at more recently by H.P. Grice (1967) in his by now classical paper on "Logic and Conversation".
Grice observed that conversations, like other human interactions, are governed by a cooperative principle, telling that you should "make your conversational contribution such as is required, at the stage at which it occurs, by the accepted purpose or direction of the talk exchange in which you are engaged". This implies, i.a., that you need not and should not supply information which you can assume that your audience already has - just as suggested by the principles of necessity and sufficiency.
From his cooperative principle, Grice derived a set of maxims concerning what should be said in a conversation and how it should be said. These maxims are not always fulfilled. A speaker may quietly violate a maxim (and mislead his audience), he may explicitly opt out, he may be faced with a clash between different maxims, or he may flout a maxim in such a way that the listener can be assumed to understand that this is being done. The latter case is especially interesting since it gives rise to a "conversational implicature" that is at variance with the literal meaning of the utterance. In such a case, the speaker is said to "exploit" a maxim. For a discussion of examples, the reader is referred to Grice's (1967) paper.
It would appear necessary to possess an understanding of the points exposed by Grice, before one can treat the question of how a language should be constituted in order to serve as an efficient means of communication. The most erudite general treatments of this topic are those by the terminologist Eugen Wüster (1931) and by the fenno-ugrist Valter Tauli (1968).
In the present context, we are not directly concerned with the question of what a constructed language should be like in order to make it easy to learn. That topic is only marginally related to Grice's conversational principles. It is, however, an important and fundamental question, which cannot be neglected totally even in the present context. It is generally agreed that in order to make a constructed language easy to learn, it has to be simple and regular in its grammatical structure and its lexical items should have a high mnemonic support value (Wüster's "Merkwert") for the prospective learners .
Tauli summarized the principles that should be obeyed when introducing linguistic innovations into ethnic languages as well as when constructing a language from scratch. Those of Tauli's principles which are relevant in the present context are arranged in Table 1, where they are set out alongside Grice's conversational principles.
Table 1: Some of Tauli's principles of language planning and Grice's conversational maxims. Tauli's abbreviations: C clarity, E economy. Analogous abbreviations have been added to Grice's principles: N quantity, Q quality, R relation, M manner.
WHAT SHOULD BE SAID
HOW IT SHOULD BE SAID
E2. The expression must be the shortest possible.
E3. The more frequent the expression the shorter it must be.
As can be seen in Table 1, Tauli's principles correspond in quite a striking way with Grice's. Although the two sets of principles have been arrived at from quite different starting points, we should expect them to agree: In order to function well, a language must be such that it makes a well behaved conversation possible. Therefore, Grice's way of reasoning can be said to confirm the correctness of Tauli's. As for "how it should be said", Tauli's formulations can even be seen as elaborations of Grice's.
Some of Grice's principles lack a counterpart among Tauli's principles of language planning. As for the maxim to be orderly, this appears to have been missed by Tauli. This point will be taken up in section 3.3, at the end of this paper.
As for the maxims concerning the truth value and the relevance of what is said, corresponding principles of language planning appear to be unnecessary since in these matters the responsibility would seem to rest entirely on the speaker rather than on the language used. This is, however, an erroneous belief. Some languages compel their speakers quite frequently to express something irrelevant or something for which they lack adequate evidence. Occasionally, languages even induce their speakers to say something which they know is not true - if taken literally. In such cases, there is a clash between the pragmatic rules of conversation and certain rules of grammar. In some languages, such clashes are much more frequent than in others. In a consciously planned interlanguage, it should be possible to minimize their occurrence. This problem has not been overlooked by Tauli, but he considered it covered by his principle E5. In all these cases, the expression contains more meaning than is necessary, i.e., some excessive, irrelevant meaning that may even be at variance with truth.
We are going to take a look at cases in which the grammars of some ethnic languages and some constructed interlanguages interfere with the conversational maxims. Among such cases, one can distinguish degrees of severity. In the less severe cases, the process of communication is merely slowed down. In the more severe cases, the meaning to be communicated risks distortion when the message is produced or the message risks misperception.
If, in a conversation, you want to fulfill the maxims of appropriate quantity, quality, and relation, you have to say no less and no more than what is required. In order for this to be fulfilled, the language must give you the freedom to choose a more or a less specific expression, according to the circumstances. The corresponding principle of language planning is known as the principle of facultative precision.
There is a basic restriction in the freedom of expression. Language operates with a finite set of lexical items, each of which has a meaning that covers a certain delimited part of reality. In most instances, the limits of the meaning of lexical items agree with what is felt by the users of the language to be a "natural" subdivision of reality.
In most instances, we can elaborate our utterances in such a way as to narrow down the meaning of sufficiently general lexical items as much as is required for the situation at hand. Thus, e.g., instead of just a noun, we may use a noun phrase with an appropriate adjective or demonstrative. When the intended meaning transcends a "natural" subdivision of reality but is more narrow than what can be expressed by a more general lexeme that covers all of the intended meaning, we need to coordinate several expressions. The need for increased elaboration in these cases is basically in agreement with Grice's maxims. We have to be more elaborate if we want to express an unusual idea. However, in some cases the grammar of a language - or its semantic structure - compels the user to express certain semantic distinctions in concepts that appear to form a natural unit. In many such cases, the distinction cannot be avoided by choosing a lower level of precision.
Strictly speaking, the existence of any compulsory (grammaticalized) semantic distinctions in languages is in opposition to the principle of facultative precision. In some cases, the compulsory distinctions would have to be made explicit anyway. In other cases, their grammaticalized expression is redundant, producing a less severe, "superficial" conflict (with E2, M3). In such cases, the quantity of meaning (E5, N2) is not affected but only the quantity of expression. In still other cases, grammar compels the user of the language to communicate some information that would not have been communicated otherwise. This produces "fundamental" conflicts. Even such fundamental conflicts are not particularly severe given that the additional information is accessible to the speaker and that its communication is not counter to his intentions. A severe conflict arises, if the language compels the speaker to express something for which he lacks adequate evidence, or which he does not want to communicate.
Most of the various grammaticalized distinctions in languages lead only rarely or never to a severe conflict with the rules of conversation. This appears to hold for the grammaticalized distinctions serving the expression of tense and aspect. As for tense, nearly all languages have a default category (Dahl, 1985) that is used in a time-indifferent sense, like the "present tense" in English. As for aspect, there is usually one alternative which is "true" in all cases in which the other one is "true" (but not vice versa). With these categories, severe conflicts arise only in the process of translation between languages that express different categorical distinctions. Translation, however, typically involves some amount of "severe conflicts" in any case.
The clearest and most widespread case to exemplify a compulsory semantic distinction that occasionally leads to a severe conflict with the rules of conversation is the number distinction made in many languages. This distinction affects unitary concepts and it cannot be avoided since it cuts across all levels of precision in expression. About 62% of humanity is compelled to be explicit about plural vs. singular.
While all languages offer some ways of undoing the number distinction explicitly, this does not really resolve the conflict. In English, e.g., it is possible to speak about an X or several Xes or at least one X. Such solutions are sometimes used in texts with legal force, such as patent specifications. However, when number is irrelevant, it should not be expressed at all. Using a paraphrase like those above, you avoid a commitment as to number only by being even more explicit about that irrelevant piece of information. In informal conversations, such paraphrases can only be used appropriately in cases in which the irrelevancy of number is assumed to be counter to the default assumption of the audience.
In languages with compulsory number expression, the plural is sometimes used as a default choice in cases in which neither a plural nor a singular interpretation should be excluded. Thus, we ask Do you have any children? rather than Do you have a child?. There are, however, other cases in which the singular is preferred, as in Do you have a car?, rather than Do you have any cars? or the like. The choice appears to depend on quite subtle pragmatic factors, e.g., the consideration that even if you own several cars, you are not likely to operate more than one at a time, unless you direct some enterprise in transportation. Only in the latter case the plural would be the default choice in a question directed to you. These kinds of solution can be said to reflect attempts to choose the form that is most likely to be correct or else least adverse if incorrect.
While it is clear that the singular/plural distinction is often relevant in utterances, it is a disadvantage having to express this even where it happens to be irrelevant or unknown. The system of Turkish, with a facultative affix used to express plurality in nouns when number is relevant and not indicated by a numeral or by any other means, appears to be functionally much more adequate.
Another case of a compulsory semantic distinction that affects unitary concepts across several levels of precision concerns natural gender. Its compulsory expression is somewhat less widespread than that of number. In Table 2, the most populous language families are listed. These represent more than 90% of humanity. Among the languages in that table it is true that all those which have compulsory gender distinction also have compulsory number distinction, in agreement with one of Joseph Greenberg's (1963) universals of grammar (no. 36). The compulsion to express gender is not completely general in scope. Probably all languages do have some expressions that are indifferent in gender for meanings in which gender might be distinguished.
Table 2: Major language families and languages grouped according to their expression of number in nouns, and gender in pronouns or via grammatical gender.
Compulsory expression of both number and gender: Indo-European (except Iranian and Eastern Indoarian), Drawida, Semitic, Hausa.
Compulsory expression of number but not of gender: Eastern Indoarian, Iranian, Bantu.
No compulsory expression of number and gender: Sino-Tibetan, Vietnamese, Khmer. Malayo-Polynesian, Japanese, Korean, Turkic, Yoruba.
The compulsory expression of gender is, i.a., reflected in the pronoun systems of many languages. It is most commonly found among third person pronouns (38%) and much less often in second person (6%). If expressed in first person (3%), it has a function similar to that of politeness distinctions. The given percentages refer to the corresponding share of humanity. There are no common third person pronouns in Korean and in Hindi-Urdu and Punjabi, where demonstratives are used.
In some languages, the distinction of gender among animate nouns, usually signaled in their ending, has evolved into a distinction of "grammatical gender" even in inanimate nouns. In the present context, we need not occupy ourselves with that phenomenon. Since it does not express any semantic content at all, it cannot be in conflict with pragmatics. However, when there is a female/male distinction in grammatical gender, this is usually employed to communicate the distinction in sex. In languages with grammatical gender, that distinction affects in general also some word classes in addition to nouns, such as articles, adjectives, and verbs. Grammatical gender is known by 43% of humanity. Among the Indo-European languages, it exists in all, except Eastern Indoarian (Oriya, Bengali, Assamese), Iranian, extra-continental Western Germanic (English, Afrikaans), and the Pidgins and Creoles. Either in pronouns or via grammatical gender, the distinction is imposed on 48% of humanity.
Most but perhaps not all of the languages that compel their speakers to be explicit about gender can be characterized as patriarchal. A patriarchal language presupposes male gender unless otherwise marked. It has at least one affix or other morphological means to mark animate nouns as female. Typically, there is no such means to indicate male gender unambiguously, and if gender is distinguished in the plural, it is marked for groups containing only female individuals as distinct from groups containing at least one male individual. It has been claimed that a patriarchal way of thinking is part of human nature and that this is reflected in these languages. It is interesting, however, that new languages that have been created spontaneously, viz., Creoles and stabilized Pidgins, like Haitian Creole and Melanesian Tok Pisin, do not reflect that patriarchal structure.
In languages without compulsory gender distinction (see Table 2), there is usually one form that contains no information about gender, and there are morphological markers as well for male as for female gender. In plural sense, these affixes may be used for pure groups. While traces of patriarchism can be attested even in these languages, these are isolated cases rather than a basic trait of their grammar.
In some European languages, such as Swedish, there has been a development away from expressing gender in nouns designating professions and performers of activities. In Swedish, the meaning of previously established lexical items has been reformed by disregarding any gender information that may be present in the constituent morphemes. Thus, idrottsman (idrott 'sport', man 'adult male human being') can be applied to female activists, and sjuksköterska, lit.: 'a female who takes care of the sick', is the term for 'nurse', including male professionals. The Nordic languages appear here to make a probabilistic choice rather than a patriarchal one.
During the past decades, the patriarchal structure of various European languages has been subjected to criticism by activists opposing "sexism". In most of the industrialized countries within the western cultural sphere, the traditional, overtly patriarchal structure of society has been largely transformed into a more egalitarian one. As a consequence, the patriarchal structure of language seems no longer to be adequate. In English, this has had the effect that for professions and performers new expressions are created that are neutral in gender. Thus, to admit ladies as well as gentlemen, chairman is now quite often replaced by chairperson. It is interesting to note that with their semantic reinterpretations, the Nordic languages have settled on an alternative means to achieve the same goal. Reinterpretation is not unfamiliar in English either. The American English use of guys, in referring to a group of women is an example.
In English, graphic strings like s/he and Ms., are sometimes used to avoid the specification of sex and marital status, the latter being another example of a compulsory distinction found in several languages. These solutions are, however, markedly expedient and exceptional in character. Being unspeakable, these symbol strings remain outside the natural language. Conversely, this is not the case with that written form of the Chinese third person pronoun ta, that is marked as female and has been created to mimic patriarchal pronoun systems. Although that form cannot be distinguished from the ordinary ta in speech, it can at least be spoken.
It needs hardly to be stressed that languages which do not compel their speakers to be specific about gender, are much more in line with the needs that exist in contemporary industrialized societies. Moreover, it is unlikely that a patriarchal language structure would really offer any overall advantage even in a stable patriarchal society.
The singular/plural distinction is compulsory in a vast region including nearly all of Europe, most of Africa, as well as Arabia, Persia, and India. Therefore, the possibility of making this distinction facultative has apparently never been considered by the creators of Volapük, Esperanto, Ido, Occidental, Interlingua, and nearly all the "naturalist" interlanguages, which use a predominantly Romance vocabulary. An exception was the mathematician Giuseppe Peano, who skipped it in his radical Latino sine flexione.
Jespersen was made aware of the inconvenience of compulsory expression of number by a Japanese Idist (Jespersen, 1928, p. 63). Therefore he provided for the possibility of being neutral as to number in his Novial. In Novial, nearly all concrete nouns end in -e in the singular, but in animate nouns that ending may be replaced by -o and -a for the facultative expression of male and female gender. In the plural, an -s is appended. The forms which are neutral in number as well as in gender are obtained by dropping these endings. In some cases, however, this results in final consonant clusters that do not fit into the phonotactic system of the rest of the language. The neutral noun forms are, thus, clearly exceptional in character. They are not likely to be chosen by default. Although Novial does not compel you to express number, it still does suggest that you should whenever you can.
As for the expression of gender in pronouns, Novial is similar to Ido. These languages have gender neutral third person pronouns alongside gender specific ones. For second and first person they have no gender specific forms.
In the construction of Interlingua (Gode, 1951), the objective was to deviate as little as possible from common Romance. This was given much higher weight than any considerations of functional adequacy. As in all the major Romance languages, in the pronoun system of Interlingua there is a compulsory gender distinction in the plural as well as in the singular and the usage is patriarchal. The grammar of Interlingua deviates from common Romance in one important detail: The phenomenon of agreement between adjective and noun has been banned. Plural is marked with an -s or -es appended to the noun. Attributive adjectives are ordinarily placed after the nouns they qualify, and they remain unmarked for gender and number. This, however, is not only different from Romance, but it is also incompatible with one of Greenberg's (1963) universals (no. 40) which says that "When the adjective follows the noun, the adjective expresses all the inflectional categories of the noun. In such cases the noun may lack overt expression of one or all of these categories" (my italics). The same incompatiblility is also present in some of the other naturalist interlanguages as well as in Volapük, but it is absent in those constructed languages that have agreement and/or adjectives preceding nouns, like Esperanto, Ido, and Novial.
Esperanto and Occidental are representatives of patriarchal languages of the same type as English. Their systems of pronouns are semantically similar to that of English, deviating from the Romance type in that there is no distinction of gender in the plural.
As for the formation of male and female nouns, Volapük was symmetric. It used the prefixes hi- [hi] and ji- [Si] (from the English pronouns) to supplement neutral animate nouns with gender information. In Esperanto, however, the morpheme -in- is suffixed to derive nouns with female gender from unmarked nouns that are interpreted as male. There are (practically) no basic nouns that are interpreted as female by default, thus all female animate nouns have the suffix -in-. This includes even such words as patrino 'mother', derived from patro 'father'. The generality of this process makes Esperanto a very strictly patriarchal language. Criticism of the inconvenience of this asymmetric system is by no means recent (cf. Baudouin de Courtenay, 1907). [Correction added in 1996: What is said about Volapük in the preceding passage is true only for reformed versions of the language that appeared later than Esperanto. In the original version of Volapük, gender was also expressed in a strictly patriarchal way.]
Esperanto has two additional affixes for the expression of gender. Quite in line with the patriarchal nature of the language, these are only used when sex is particularly relevant. In such cases, the prefix ge- is used in the sense of 'group including representatives of both sexes', as in gepatroj 'parents', and the morpheme vir- originally and properly: 'adult male human being' is prefixed when the maleness of an individual needs to be stressed, as in virbovo 'bull'. A mere bovo might be an ox.
While in number and gender none of the mentioned interlanguages satisfies the principle of facultative precision, James Cooke Brown's (1975) Loglan, which can be said to represent a contemporary philosophical language, does. This creation, however, has a variety of properties which place it distinctly outside the range of natural human languages, as it was meant to be an atypical language.
The compulsory specification of number and gender belongs to those features of grammar for which no satisfactory functional explanation can be given. Although information about number and gender is often important, communication does not benefit from its compulsory specification, which entails a severe handicap whenever these distinctions are not wanted to be made. How, then, can it come about that many languages do compel their speakers to be explicit about them? If the evolution of language was controlled by functional factors, such tyrannical traits should only rarely come into existence and they should disappear in the long run. Otherwise, this must be driven by a powerful process that overrules the demand of pragmatic adequacy.
If there are traits in language which run counter to what would be optimal from a pragmatic point of view, these traits may tell us something about human predisposition to language as distinct from general rational capacity. The question why languages are the way they are has been treated extensively from a genealogical point of view. These studies explain the relations between earlier and later forms of a language or, in general, the relations between languages. The question has been treated by far less extensively from the point of view of our pragmatic needs in communication and our physical and mental predispositions (faculté de langue). Within phonetics, however, one or the other aspect of this question has been touched upon quite frequently, and there are also studies focusing directly on it (Lindblom, 1990).
As for the distinctions we are discussing here, there is a paper by Larry Hyman (1984) treating the conflict between the interests of pragmatics and these of Grammar. Hyman ascribes certain autonomous "interests" to universal grammar, referred to as Grammar with capital G. In some cases such as those discussed above, the interests of Grammar are in conflict with those of pragmatics and they overrule the latter.
Essentially, Grammar wants expressions to obey syntactical, morphological and phonological rules irrespective of pragmatic considerations. While, historically, these rules arise as a result of frequent pragmatic coincidences, they become detached from pragmatics in the process of grammaticalization (Hyman, 1984; Heine et. al., in print).
A closer look at the cases of number and gender shows us, however, that the roots of the phenomenon sit deeper than at the level of linguistic form. In many languages there are a variety of morphological means of expressing plurality and/or gender. While the grammar prescribes one of these means in most of the particular cases, fundamentally, it compels us to make a semantic distinction, whatever the means of expressing it. Apparently it does so even in those exceptional cases in which there is no difference in form. Thus, in the mind of a native speaker of English, even a form like sheep is interpreted as either singular or plural in sense or it raises the question whether one or several are meant. The form trees does not as automatically evoke the question whether two or more than two are meant.
In order to understand how it comes about that some languages structure reality in more detail than what is optimal from a functional point of view, we must first realize that as representatives of the speaking animal we are born with an instinctive urge to acquire language, i.e., to construct a language for ourselves. While an instinct to acquire complex vocalizations is present also in various other species, such as song birds, the instinct to structure experiences semantically and to associate the resulting minimal meanings with vocalizations and, further, to recombine such elements (morphemes, lexemes) into complex utterances with complex meanings is unique to our species. "Universal grammar" is nothing else but the linguist's conception of that instinct of ours.
While "grammar" describes how utterances are associated with meaning in a particular language, "universal grammar" describes how grammars are constituted in a particular species. A complete generative grammar of a particular language would have to generate all the utterances, with their associated meanings, that are possible within the language in question. A generative universal grammar would have to generate all the grammars that are possible within the species in question. I am not aware of any attempt to strictly formulate universal grammar as a whole, but since Greenberg (1963) many of its features have been described in the form of implicational rules.
Like most other instincts, the human urge to acquire the grammar of a language is strong enough to guarantee the achievement of that goal in almost all the members of the species, even in situations with impoverished input, but in a favorable environment, instincts tend to produce some excessive activity, such as overeating. Cases in which languages compel their speakers to structure their expressions in more detail than what is functionally useful appear to be the result of an excessive activity of Grammar, in a historical perspective. As we know, this can go to such extremes as assigning gender to inanimate nouns. Excessive categorization would probably not have arisen if Grammar had not been so strong, but this might also have resulted in a suboptimal degree of structuring in some areas of grammar and in deficient language acquisition in some members of our species. Considering the severity of these disadvantages, we have reason to believe that evolution has led to a grossly optimal strength of Grammar.
Even if we choose our words in the most economic way, most of our utterances still contain a lot of redundancy. Some of it originates in our speech organs due to physical restrictions (coarticulation) and some of it is due to conventional phonotactic restrictions or to empty slots in the lexicon, to syntactic restrictions, and to semantic and pragmatic improbabilities in the given context. A certain amount of redundancy is necessary in order to communicate successfully in various non-optimal situations, such as in a noisy environment, or when the listener cannot see the speaker, when he is at a large distance from him, when he has a hearing loss, etc. Redundancy is also required in order for language to be learnt without formal instruction. Thus, some amount of redundancy is quite useful and it is often necessary to avoid obscurity of expression. Of course, even more fundamental for perspicuity is the requirement to avoid sheer ambiguity.
In a simplistic approach, the distinctive difference between two utterances with different meanings can be expressed in terms of the number of speech sounds (dn) by which they differ. For ambiguous utterances, dn = 0. If dn > 1, there is redundancy. In many cases, a semantic difference signaled by just one phoneme (dn = 1) is distinctive enough, but certain pairs of speech sounds, such as [m] and [n] or [d] and [t] are much more likely to be confused than others, say, [m] and [t]. Thus, in a more elaborated approach, it is necessary to quantify the distinctiveness of each pair of speech sounds. In certain cases even two or more phonemes may be insufficient to signal a semantic distinction with the necessary clarity. Consider, e.g., the German numerals [tsvaI] and [draI]. For the former, there is an alternative [tsvo] that is used when distinctiveness needs to be improved. This is the only motive for its use in the standard language.
There is an antagonism between the maxims of clarity and of brevity. Grice's maxim "Be brief" corresponds to two of Tauli's four principles of economy telling us that an expression should be as short as possible and, especially, that "The more frequent the expression the shorter it must be". Ideally, expressions should be no longer than they need to be in order to satisfy the requirement to be sufficiently distinctive.
The application of this principle in language planning is somewhat problematic since the necessary distinctiveness of an expression depends to a large extent on the context in which it is uttered. If the principle is applied to sentences in context, it will be found that most of them contain a large amount of excessive redundancy distributed erratically over their constituents. In isolated one word utterances, the amount of redundancy is usually much smaller. Redundancy can be reduced by shortening the lexical forms, but if this were done in a way that was optimal in a given context, the distinctiveness would be deficient if the same forms were used in a less informative context.
As for the distinctiveness between lexical items as such, there are very large differences between languages. In the lexicon of some languages with predominantly monosyllabic morphemes, such as Chinese and other members of the Sino-Tibetan group, nearly all entries have several homonyms and for each, there is a large number of entries from which they differ by just one distinctive feature. The degree of ambiguity in the lexicon is much lower in all the major languages spoken outside East Asia. While in a given context, most Chinese utterances have just one plausible meaning, the maxim "Avoid ambiguity" has to be taken into consideration quite frequently. Often, ambiguity is resolved by using a compound instead of a monosyllabic morpheme.
In sharp contrast with spoken Chinese, written Chinese lacks lexical ambiguity almost totally. Since for some purposes, the degree of distinctiveness in spoken Chinese is felt to be suboptimal, the absence of lexical ambiguity in the logographic writing system is highly appreciated. Among European languages, a similar although much less drastic discrepancy between homophony and homography exists in French.
From a rational point of view, any unintentional ambiguity of utterances is unwanted, and it appears quite reasonable to require this from a planned interlanguage. However, this can only be achieved at a certain cost. Lexical ambiguity can, in general, only be avoided at the cost of a loss in Merkwert, while avoidance of syntactical ambiguity is likely to increase the length of utterances.
As for ambiguity, the interlanguages Esperanto and Interlingua differ quite radically. In Esperanto, the morpheme is the basic invariant unit of grammar, and, allowing for a moderate range of polysemy and all-European metaphoric usage, there is, essentially, no ambiguity at the level of morphemes. Users of Esperanto are quite sensitive about this and they will, typically, not allow you to attach an additional meaning to a pre-existing morpheme.
The verb komputi is a case in point (Golden, 1987). It used to have the meaning 'to measure by integration' (e.g., in measuring electric energy or water consumption etc.). When electronic computers began to become more widespread, the English noun computer was borrowed into many languages. If the verb komputi had not existed previously, it would probably have been adopted at that stage and given the meaning 'to process data' (as by computer), with the regular derivatives komputa 'computational', komputo 'computation', komputado 'process or activity of computation', komputilo 'computer', komputisto 'computation specialist', etc. Due to the previous definition, however, it took thirty years of struggle for this new meaning to be accepted, replacing the old one, by what at present appears to be a majority of members of the Esperanto community.
The meaning of the lexical units of Interlingua, on the other hand, is much less restrained. The words of Interlingua are allowed to have as wide a range of polysemy as in its source languages, and no attempt is made to eliminate accidental ambiguity from the vocabulary. While in Esperanto polysemy is kept within limits set by other existing lexemes, Interlingua is characterized by overlapping meanings. Synonymous morphemes exist in both languages, but they are more numerous in Interlingua.
If a language were to be planned from scratch, or constructed on the basis of a wider sample of ethnic source languages, it would be inadequate to allow as much homonymy and synonymy as in Interlingua. In Interlingua, planning has been restricted to a minimum, as its aim was to achieve word forms which can be recognized easily by speakers of languages with a Romance vocabulary.
In his avoidance of ambiguity, Zamenhof focused mainly on morphemes, and it is fine that these are unambiguous, but what really matters is the meaning of utterances - of one-word utterances and of any longer utterances. Considering one-word utterances out of context, even Esperanto has a considerable amount of ambiguity. This is due to the freedom to compound morphemes without asking whether the word so coined has been used previously. Quite frequently, such compounds agree in form with primitive morphemes, e.g., kat+aro 'a pack of cats' and kataro 'catarrh'. Zamenhof accepted a considerable amount of loss in recognizability in his attempts to avoid this happening too often.
A radical elimination of ambiguity has been attempted, successfully as it appears, by Brown (1975) in his Loglan. The costs, however, appear to have been very high. Loglan imposes on its vocabulary severe morpho-phonotactic restrictions such as are not found in any ethnic language. It has, however, not been shown that such restrictions are necessary in order to achieve that goal. Brown had deliberately chosen to distinguish the parts of speech and signal their boundaries on a phonotactic basis, to the exclusion of all distinctive prosody.
The ambiguity of utterances is nearly always resolved by context and passes without notice by the partners in a conversation. On the other hand, there are cases in which distinctiveness is crucial in most contexts and it is not sufficient just to avoid ambiguity. Violations of Tauli's principle C4: "The greater the possibility for semantic confusion the greater must be the difference in expression" are more often a problem than plain ambiguity.
The philosophical languages created in the 17th century failed to satisfy Tauli's principle C4. Their deficiency in distinctiveness could not have gone unnoticed if these languages had come into use in practical life. The vocabulary of those languages was structured in a way directly opposite to C4, such that there is only a minimal difference between expressions for which the risk of semantic confusion can be expected to be largest. In Wilkins' proposal, e.g., [ti] means 'sensible quality'. Different kinds of sensible quality are distinguished by appending one consonant, e.g., [tid] 'colour', [tib] 'other visible quality', [tig] 'sound'. Colours, in turn, are distinguished by an appended vowel, such as [tida] 'redness', [tide] 'greenness', etc. The latter distinctions might work very well if stress were to be placed on the last vowel, but this is not the rule prescribed by Wilkins. The confusions most likely to occur are those between words with different final consonants. The system is characterized by a complete lack of lexical redundancy, and it guarantees that serious misunderstandings will arise, if it were to be used.
The philosophical languages illustrate what is likely to happen when Grammar is allowed to work without being checked by pragmatics. Some deficiencies of this kind can also be found in Esperanto. The personal pronouns mi 'I', ni 'we', vi 'you', li 'he' are acoustically not distinct enough from each other. Quite a low level of environmental noise makes them indistinguishable and there are no additional syntactic cues. On many more occasions than the speakers of the language realize, the successful comprehension of these pronouns is likely to rest entirely on contextual information or on the distinctiveness in their visual reflections. When the participants in a conversation can see each other's faces, lip-reading is known to contribute considerably to comprehension, even for listeners with normal hearing.
In all ethnic languages, the phonetic differences between the pronouns are larger than in Esperanto. While some pronoun systems include a subset whose members share a common ending and which are distinguished only by one initial consonant, as in Spanish me - te - se - le, or German mein - dein - sein, the consonants that appear in these forms represent an acoustically more adequate choice than those in the Esperanto pronouns. Although a lack of distinctiveness similar to that in Esperanto characterizes first vs. second person plural pronouns in most Slavic and Romance languages, e.g., my - vy in Russian and nous - vous in French, in these languages the distinction is often supported by verbal inflexion.
In Esperanto, the pronouns have been provided with a unique ending in analogy with the larger classes of words. However, unlike -o, -a, and -e, the endings of nouns, adjectives, and derived adverbs, the pronominal ending -i is not used for derivation. It does not constitute a morpheme. The endings serve also the purpose of facilitating the syntactic analysis of what is said. For pronouns, however, such a measure is unnecessary. Due to their small number and high frequency of use, the pronouns are recognized easily and quickly, whatever their form.
Criticism of this kind was raised at an early stage and corresponding changes were already included, by Zamenhof, in a proposal of modifications subjected to a poll by the readers of the newspaper La Esperantisto in 1894. The proposal was, however, rejected. While each of the constructed languages that have been published by well informed authors in this century (Ido, Novial, Occidental, Interlingua, Neo, etc.) has more distinctive pronouns than Esperanto, the marginal distinctiveness between first and second person plural in the Romance languages has remained in Ido, Novial and Interlingua, where there is no support by verbal inflection. Only Occidental and Neo pass without any critical comment.
Although insufficient distinctiveness within classes of semantically related and syntactically equivalent lexical items is a typical trait of the early artificial languages, similarly drastic deficiencies can also be attested in ethnic languages. Letter names constitute the clearest case in point.
The names used in most European languages for the letters of the Latin alphabet are considered to have originated in Etruscan (Hammarström, 1930). From that language they were borrowed into Latin, through which they later spread to most of the various languages whose speakers adopted that alphabet. It appears likely that the Etruscans created their letter names by attempting to sound the phones they represent without adding anything else. This is, of course, in no way problematic for the vowels. At the point in time when the Etruscan letter names were introduced, the consonants m - n - l - r - s appear also to have been pronounced syllabically. The initial [e] in the Latin names for these consonants is due to a later development that occurred probably already in Etruscan. For the stop consonants, a schwa vowel intrudes itself automatically. This schwa was subsequently interpreted as an [e], except after velar consonants, where it sounds more like an [a], unless the consonant is produced with lip-rounding, when it sounds like [u].
While the Latin letter names appear to be the result of a spontaneous development, there are also some examples of more "Grammar driven" inception, where the names of the consonants and the letters representing them are formed invariably by appending the same, more or less arbitrarily chosen vowel. Such cases are at hand in Bulgarian (Cyrillic) and Turkish (Latin), both appending an [e] and in Pilipino (= Tagalog), where an [a] is appended. In Esperanto, they are formed by appending an [o].
The consonant letter names used in Bulgarian, Turkish, Pilipino and Esperanto are, of course, patently inadequate in any acoustically non-ideal situation. The letter names of Russian and the Latin-based letter names used in most European languages such as English are only slightly less inadequate, just consider the lack of distinctiveness within the groups b - d - p - t - v and n - m. The insufficiency of these letter names is attested by the existence of various more or less standardized improvisations, called "spelling alphabets", which are used in situations when the ordinary letter names are likely to fail their purpose. The letter names which spread together with the art of alphabetic writing from its Semitic origin, and which are used in Arabic, Persian, Hebrew, Armenian and Greek, are much more adequate for communication. In these languages, there is no need for any auxiliary spelling alphabets.
At the time of their creation, the ending -o of the Esperanto letter names was apparently thought of as being identical with the morpheme -o that is appended to all nouns. Ordinary nouns add a -j [j] for plural and an -n for object case. In derivatives, the -o is dropped unless there is some particular reason for not doing so, and an -a can be substituted for -o to create an adjective. In actual use, none of this is valid for letter names. In these, the o is treated as part of the morpheme designating the letter in question. Thus, instead of *boj - *bon - *bojn the forms used are booj - boon - boojn, and the adjective is boa and not *ba. In this case, Grammar has overruled Zamenhof and nullified the motive for having the same vowel in all consonant names. Universal grammar assigns the vowel in the letter names to the root morphemes, since all other roots have at least one vowel.
A more satisfactory set of letter names is obtained by placing a vowel in front of the consonant if it is a sonorant or an unvoiced fricative, just as in the Latin system, and by choosing a different vowel for each place of articulation, [o] for labials, [e] for alveolars, [i] for palatals, and [a] for velars. These are the vowels that would be obtained by excessive coarticulation if a hypothetical [unrounded half open back vowel] were taken as the origin. The consonants n - l - r can be dissimilated as [en] - [ul] - [ar].
While cases of insufficient distinctiveness must be seen as a severe deficiency, its opposite, too long word forms are also inconvenient. Ideally, the length of morphemes should be inversely related to their frequency of use. Thus, morphemes whose function is mainly grammatical should be short. This is also required of pronouns, whose essential function is to serve as abbreviations for proper names and noun phrases. If too long, such pronouns are avoided by speakers.
Clear cases of too long pronoun forms are at hand in Japanese. The pronouns for first and second person singular are watashi 'I' and anata 'you'. They have to be followed by a case postposition. The length of these forms contrasts sharply with that of the many monosyllabic Chinese words which have been borrowed into the language. These pronouns are not shorter, but longer than an average noun. No wonder, then, that they are hardly ever used in spontaneous speech. This holds also for the Swedish third person pronoun vederbörande, which would otherwise be very convenient since it is indifferent for number and gender.
Some of the derivational suffixes that have spread from Latin into the modern European languages are also excessive in length. Consider the following causative suffixes:
Interlingua -ificare -ification French -ificier -ification German -ifizieren -ifikation Russian -ificirovat' -ifikacija English -ify -ification German -igen -igung Esperanto -igi -igo
Although the Latin suffix was excessive in length from the beginning, it has absorbed further redundant material on its way through French and German into Russian. The short verbal form used in contemporary English is adequate in length. In Esperanto, neither the verb nor the noun is too long. It might have been more appropriate, however, to shorten the general European (Interlingua) -ifica-, to -if- , as in Ido, instead of using the exclusively German -ig-, whose use together with morphemes of Latin origin is unfamiliar even in German.
Esperanto does not possess any morphemes that are felt as clearly too long. The copula esti 'to be', estas 'is', estis 'was' may be a marginal case in question. The copula is treated like a regular verb, while in most Indo-European languages it is irregular, and this is partly due to its high frequency of use and its usually being unstressed. In contemporary usage, many speakers of Esperanto tend to avoid the copula in front of verbal participles by just appending the verbal mode- and tense-suffixes to the participle, thus treating it as a verb. They say li jam skribintus ^gin, instead of li jam estus skribinta ^gin 'he would already have written it'; mi fieras mencioti, instead of mi fieras esti menciota 'I am proud I shall be mentioned'. This has become usual for predication of any "vivid" qualities including those which would otherwise be expressed by an adjective preceded by the copula, as in ^cu vi pretas? instead of ^cu vi estas preta? 'are you ready?'; ru^gas la oriento instead of la oriento estas ru^ga 'The East is red, viz., communist'. The spread of this usage is due to its attractive shortness. For purely static predications, it is not accepted.
In this connection, it is interesting to note that a proposal of linguistic and orthographic reforms submitted by Zamenhof to Emile Boirac, the president of the esperantists' Lingva Komitato, in 1906, included the admission to use the mere inflexional endings to fulfill the function of the copula (Courtinat, 1964). The proposal included also a small list of alternatives for morphemes which might be mistaken as compounded. Zamenhof withdrew his proposal due to the opposition by Boirac. 
The last one of Grice's maxims, "Be orderly", shall probably be understood to mean that one should present the information in an order that is pragmatically adequate for the conversation. This concerns mainly the larger constituents of speech. Within constituents such as simple noun phrases which do not contain complicated attributes (relative phrases), word order is in most instances given by the grammatical rules of a language. These have, however, also arisen on the basis of pragmatics, which is reflected in the universals of word order, described by Greenberg (1963).
At the sentence level, the basic requirement of order is that the speaker should first announce his topic (thema) and then make his comment (rhema). This requirement might be in conflict with the syntactic rules of those languages which prescribe a certain order of sentence constituents on the basis of the formal grammatical relations between verbs and their arguments (subject, object, instrument, etc.). However, Most languages allow a topic to be placed at the beginning of a sentence, whatever its grammatical relation to the main verb. As in other languages that lack case markers for both S and O, in English there are some restrictions in this freedom, and a finite main verb can never be placed in topical position. Beyond the basic requirement, the possibilities of observing a pragmatic order of constituents are quite restricted in most languages.
In most languages, the various relations between a verb and its arguments are indicated with preposed or postposed case markers (appositions or inflexional affixes). For indefinite objects, the verb itself often has the function of a case marker. In such cases, the object has to be placed in immediate proximity to the verb (V O in prepositional languages and O V in postpositional ones). One case is often unmarked. This is usually the one that includes among its functions indicating the subject of an intransitive verb. In English statements, however, the subject of a sentence is consistently marked by its position immediately in front of the verb, while the object, with its default place after the verb (S V O) or after an intervening indirect object (S V D O) can also be moved to the beginning of a sentence (O, S V) without being marked as an object.
The role of position in marking the grammatical function of sentence constituents should not be overstated. Although in most languages there is a standard order in which these constituents are placed, that order is often not compulsory. Thus, the position of the object tends to follow the conversational rules whenever there is a case marker at disposal. This may be inflexional endings, as in Russian, or a preposition such as ba used in Chinese for anaphoric objects, or a, used in Spanish for animate objects. Occasionally, the functions of subject and object are not marked at all. This is, e.g., the case in German, where these functions are marked only for nouns that are male in grammatical gender. If neither S nor O is male, the distinction of their functions rests entirely on semantic probability within the given context, the word order being given by a T v R rule, where T stands for topic, v auxiliary or verb, and R for 'rest of sentence in S D O V order (if the constituent in question is left over)'.
In Loglan, all case relations are expressed by position only. This is one of the features which distinguish that language from natural languages, and which make it difficult to handle for humans . Although to a much lesser degree, the same inadequacy is characteristic of most codes designed for restricted man-machine communication, such as, e.g., computer operating systems. These rely, e.g., on position for specifying source and target of a transaction. While this poses no problem for the machine, for its human partner it would be much more natural to mark the target with an apposed particle. In English-based codes, the preposition to would be well suited for this purpose.
The positional grammars have not been constructed in analogy with natural language. They are, instead, analogous to the way of using numeric signs in the Indo-Arabic system of writing numbers. Although in this particular semantic field a positional grammar appears to have certain advantages, in spoken language, position is not relied upon even for numbers. The grammar for numbers is analogous to that for sentences in that the function of each figure, except at most one, is marked by appositions (like n-hundred, n-thousand), sometimes with inflexion (like n-ty). While in all languages, there is a standard order in which the figures are told, it is not this order, but the function markers which are decisive for how a figure has to be interpreted. Occasionally, two standard orders coexist, as in Norwegian, where fem og seksti, lit.: 'five and sixty' = sekstifem, lit.: 'sixty five'. The markers may be skipped only when the digits merely serve the purpose of identification rather than that of telling a natural number.
It is a reasonable demand on a planned language intended primarily for human communication that it should allow sentence constituents to be placed in a pragmatically adequate order, while their function should also be made clear. As for the distinction between subject and object, in Interlingua this requirement is only fulfilled when pronouns serve the function of an object. It was fulfilled in Volapük and it is with some exceptions (numerals, non-assimilated proper names) in Esperanto and Ido, which mark the object with a suffix -n. This suffix is, however, an exceptional feature in these languages, since all the other case relations are indicated only by means of prepositions. It appears as even more exceptional in Ido, in which it is used only for objects in preverbal position. The structure of these languages would have been simplified in an essential point if even object case had been indicated with a preposition. In contrast with the consonantal suffix, a preposition would not pose any problem with proper names.
As for the order of meaningful elements in sentences, agreement with the universals of human language is the minimum one should require of any language that is proposed to serve as an interlanguage on a global scale. This is, of course, not a sufficient selection criterion, since it is satisfied by almost any ethnic language. It is, therefore, quite embarrassing to find that some of those constructed interlanguages that have found a number of adherents contain features which are incompatible with universals, as mentioned in section 2.2 for the expression of plural in Volapük and Interlingua.
The grammar of Volapük contains one additional incompatibility with a universal. It concerns the order of the suffixes indicating case and plural. Greenberg's universal no. 39 says that "Where morphemes of both number and case are present and both follow or both precede the noun base, the expression of number almost always comes between the noun base and the expression of case." In Volapük, on the contrary, the plural morpheme is placed in final position.
The grammar of Esperanto appears to be free from incompatibilities with universals. However, the actual usage of Esperanto, as distinct from any prescriptive grammar, contradicts Greenberg's universal no. 19, which says that "When the general rule is that the descriptive adjective follows [the noun], there may be a minority of adjectives which usually precede, but when the general rule is that descriptive adjectives precede, there are no exceptions." The English expressions in which the order noun - adjective can be attested, as in secretary general, Charles the great, appear to remain outside the scope of Greenberg's formulation, since they are either compounds or they contain an epithetical, not merely descriptive adjective. In Esperanto, the general rule is that the adjective precedes the noun, but it is also permissible to place it after the noun. The latter alternative is commonly chosen, in humble conversations as well as in poetry, when there is emphasis on the adjective, especially in contrastive contexts, as in Ili havas la rimedojn HOMAJN, sed ne la MATERIAJN 'They have got the HUMAN resources, but not the MATERIAL ones'. Since this occurs although it has not been explicitly prescribed, it shows us that Greenberg's formulation of that universal is exaggerated rather than that there would be anything wrong with the few prescriptive rules of Esperanto grammar.
: As for this requirement, all the better known constructed interlanguages, like Esperanto, Ido, Novial, Occidental, Interlingua, and Neo appear to be distinctly biased if proposed for global use. The western cultural sphere, whose common international vocabulary these constructed languages utilize, embraces only about one quarter of humanity. There are, in addition, three other large cultural spheres with their respective common international vocabularies (Traunmüller, 1980) which should not be neglected.
: Even in Loglan any one argument of a verb can be brought into topical position, but this is done by a readjustment of the positional code. The feeling one gets when trying to use such a system is perhaps akin to that a chimpanzee gets when trying to use human language.
: I am obliged to Christiaan B. Vos for having drawn my attention to this proposal by Zamenhof.
For literature on constructed languages, the reader is referred to the bibliography in Blanke (1985).
Baudouin de Courtenay, J. (1907). Zur Kritik der künstlichen Weltsprachen. Annalen der Naturphilosophie 6, 385-433. Reprinted in R. Haupenthal (ed.) Plansprachen. Darmstadt: Wissenschaftliche Buchgesellschaft, 1974, 59-110.
Blanke, D. (1985). Internationale Plansprachen. Berlin: Akademie-Verlag.
Brown, J. C. (1975). Loglan 1: A Logical Language. 3rd edn. Gainesville & Palm Springs, USA: The Loglan Institute, inc.
Courtinat, L. (1964). Historio de Esperanto. Vol. I. 1887-1913. Bellerive-sur-Allier. 185-201.
Dahl, Ö. (1985). Tense and Aspect Systems. Oxford: Basil Blackwell.
Gode, A. (1951). Interlingua-English Dictionary. New York: Storm Publishers.
Golden, B. (1987). Terminologia kaoso: komputilo, komputero, komputoro. In Centjara Esperanto, Chapecó-SC, Brasil: Fonto. 105-142.
Greenberg, J. H. (1963). Some universals of grammar with particular reference to the order of meaningful elements. In J. H. Greenberg (ed.) Universals of Language. Cambridge, Mass., USA: MIT Press. 73-113.
Grice, H. P. (1975). Logic and conversation. In P. Cole (ed.) Syntax and Semantics. Vol. 3. New York: Academic Press. 41-58.
Hammarström, M. (1930). Die antiken Buchstabennamen. Arctos 1, 3-40.
Heine, B., Claudi, C., Hünnemeyer, F. (in print). Grammaticalization, A Conceptual Framework, Chicago: Chicago University Press.
Hyman, L. M. (1984). Form and substance in language universals. In B. Butterworth, B. Comrie, Ö. Dahl (eds.) Explanations for Language Universals, Berlin/NewYork/Amsterdam: Mouton. 67-85.
Jespersen, O. (1928). Eine internationale Sprache. Heidelberg: Carl Winter. (Translation from the English original.)
Lindblom, B. (1990). Models of phonetic variation and selection. PERILUS XI, 65-100, Institute of Linguistics, Stockholm University.
Saussure, R. de (1915). Fundamentaj Reguloj de la Vort-teorio en Esperanto. Bern. Reproduced in 1969. Saarbrücken: Arthur Iltis.
Tauli, V. (1968). Introduction to a Theory of Language Planning. Uppsala: Almqvist & Wiksell.
Traunmüller, H. (1980). Socia bazo kaj komunika efikeco de interlingvoj. In I. Szerdahelyi (ed.) Miscellanea Interlinguistica. Budapest: Tankönyvkiadó. 205-215. [This was based on a term paper (1975) "A universal interlanguage: Some basic considerations".]
Wilkins, J. (1668). An Essay towards a Real Character and a Philosophical Language. Reproduced in 1968. English Linguistics 1500-1800, no 119. Menston: Scolar.
Wüster, E. (1966). Internationale Sprachnormung in der Technik. 2. Aufl. Bonn: Bouvier.