A logical language

version 3.2 (2001.01)
[updated by a more recent version]

by Claudio T. Gnoli


0 : Introduction

1 : Writing system

2 : Phonology

3 : Morphology

4 : Syntax

5 : Vocabulary

0 : Introduction

Liva is a logical language, currently under development. It is created mainly by the volunteer work of its author, Claudio T. Gnoli.

Logical languages are a special kind of artificial languages, namely of speakable languages created from scratch rather than developed spontaneously within a culture: the most famous artificial language is Esperanto. The typical feature of logical languages is that their structure and grammar are intended to express the logical relations between words in clear and effective ways: the most successful logical language is Lojban, while few others are known.

0.1 : Aims

Liva is constructed mainly for fun: at first, it has not any practical aim.

It can be seen as an experiment, trying to satisfy both the requirements of logic and the aesthetic preferences of its author -- two things not unrelated to each other. Just as poets like to express their souls by writing poetry, "glossopoets" like to express their way of seeing the world by coining a language to describe it. There is few hope that such a language will ever be spoken by a wide community, but that does not matter.

A secondary effect of the experiment is learning interesting things about the structure of languages, with special reference to their logical sides. Syntactic and semantic solutions developed in Liva could be applied in future to other information processing tools, for example indexing languages like thesauri and classification schemes.

0.2 : Name

The name "Liva" has no special meaning: I chosen it just because I like its sound. It does not mean anything special in the language itself.

"Liva" is also a name of women: for example, it is the name of a nice young girl living in Slovak mountains, in a novel for young people of the Poly series by Cécile Aubry; and of a former prostitute who meets the leading male character, in film "Mifune" directed by Søren-Kragh Jacobsen of the Dogma group (1999).

0.3 : History

I began thinking about an artificial language when I was very young, perhaps 12. While a teen-ager, I developed the alphabet and grammar of an agglutinative "rational" artificial language, called Aarbane and later Aarbad (stress on the second {a}). Of course it was quite naive; however, some ideas from it survived in Liva.

I did not study linguistics, and did not spend more time with Aarbad. However, after my degree (in zoology), my interest within artificial languages increased again, and was stimulated by reading some linguistics books and discussing my ideas with a couple of friends. At that time I also got in contact with many other people sharing the same hobby, mainly through the mailing list ConLang: such exciting experience has obviously been a major incitement.

So, in 1995, a new language began to develop from the ashes of Aarbad: it was provisionally referred to as "universal logical language" and then Luv, but very soon the name became Liva. I am working on it at sessions separated by long periods of quiescence, depending both on inspiration and time factors.

0.4 : Acknowledgements

I am grateful to all friends and fellow "Langdevils" who have contributed to the development of the language by stimulating discussion and useful information: especially Tommaso R. Donnarumma, Alessandro Emilio, B. Philip Jonsson, And Rosta.

Rick Harrison generously made his Universal Language Dictionary (5.1) publicly available, and Leo J. Moser shared useful additions to it; Michael A. Rouse sent to me his "enormous corpus" of lexical information derived from the Internet, which I have not yet exploited. Allocation of a Unicode Liva character set has been possible thank to John Cowan's initiative.

Other logical languages, especially Loglan and Lojban, are obvious sources of ideas and solutions.

0.5 : Documentation

A first amount of information on the language was originally posted to the "ConLang" mailing list, then hosted in "Mia Soderquist's little conlang page" during 1996: that is considered the version 1 of the language description. This web site do not exist online anymore. A second version has been hosted since 1997 by Nick Summers in his web site, at <>. Recent developments and alternatives on some details, especially concerning phonology, have been discussed in "LangDev" and "Aleppe".

This document is the third and most complete version of the language description. It can be integrated with additions from time to time.

1 : Writing system

An alphabetical system -- as opposed to syllabic, ideographic, etc. -- has been chosen for Liva.

The representation of sounds in Liva is always phonetic, namely each symbol represents only one phoneme, and each phoneme is represented by only one symbol.

1.1 : Liva graphemes

A Liva original alphabet has been planned to represent the phonemes of the language. It is intended to be iconic, namely reflecting by the shape of graphemes the main articulatory features of the phonemes, according to a regular scheme.

C.A. de Moy's alphabet (1787), modern Visual Speech, and to some extent the Korean alphabet have already been based on such idea. Its most complete realization probably is Herman Miller's Lhoerr alphabet, even available as a TTF file. B. Philip Jonsson has also done much research in this direction.

After many previous attempts not completely satisfying, a new Liva alphabet is being studied by the author, with precious suggestions by Tommaso R. Donnarumma and Maurizio M. Gavioli.

1.2 : Roman graphemes

To easily represent Liva texts and communicate them, a standard Roman writing is provided.

Roman graphemes (letters) are commonly written all non-capital. For graphical purposes it is possible to write them all capital, leaving the non-letter characters unchanged. A blank space between words is usually given as a visual help, but it is not strictly needed, as phonotactics make words self-segregating (2.4).

In this document, Liva words and phrases will be written in italic characters. Liva graphemes will be indicated in brackets like {a}, rather than in angled brackets like <a> as used in linguistics; this is because angled brackets can be confused with symbols very common in online documents and email.

As Liva has 32 phonemes, while the Roman alphabet only offers 26 graphemes, some additional symbols are necessary. The use of digraphs (sequences of two graphemes representing one phoneme) has been rejected; instead, some special graphemes have been defined.

1.2.1 : Extended set

The Liva extended Roman alphabet uses, besides the 26 common Roman graphemes, the following ones:

{ð} and {þ} are from the Icelandic alphabet, {ø} from Danish, {µ} from Greek, {ß} from German (it can be replaced by a Greek beta when available), {£} from the British pound symbol (it can be replaced by a Greek lambda when available). The symbol {ç}, from French and Portuguese, can also be used in place of {c}.

1.2.2 : ISO 80646 set

The special letters above are not included in the 7-bit ISO 80646 character set, also known as the ASCII set, which is widely used in exchanging digital information, for example in email.

When not available, the above letters are replaced by the following symbols:

{0} can also be used in place of {&}, making the whole alphabet constituted only by letter and digit symbols.

1.3 : Non-Liva sounds

Non-Liva sounds and words, delimited by special markers (3.2.4), can be represented within Liva text in foreign alphabets. The native alphabet of the foreign language, IPA international phonetic alphabet, or the Lhoerr alphabet are recommended for such purpose.

2 : Phonology

Although human phonatory apparatus is able to produce a continuum of different sounds, languages make distinction between a limited number of relatively fixed sounds, called the phonemes. Phonemes differ from each other by one or more parameters or distinctive features:

2.1 : Phonemes choice

Although it would be possible, by combining the numerous distinctive features, to distinguish between many hundreds of potential phonemes, each natural language only uses a limited number of them, usually some tens, so that misunderstanding is less likely; on the other hand, more phonemes allow to coin more short words. Hence, a number of phonemes between 25 and 40 seems to be a good compromise.

Liva phonemes have to be:

  1. easy to distinguish between each other, both for speakers and for listeners. So, places and manners of articulation will be clearly separated: no semivowels, which could be confused with voiced fricatives or closed vowels; places of articulations will be at quite regular intervals;
  2. simple to articulate. So, sounds formed by overlapping different places, like labiovelars, sounds formed by quick succession of simple sounds, like affricates and vibrants, and sounds produced by unusual air flows will be avoided;
  3. included in an elegant and regular phonological scheme, as an implementation of Liva's characteristic regularity.

Prosodic features, namely intensity, tonality and duration, will not be used to distinguish phonemes. So, it will be possible to pronounce Liva with any accent or intonation without changing the meanings. (For personal aesthetic preferences, I suggest to adopt a fixed stress on the first vowel of words, and primarily on syntagm heads.) Among articulatory parameters, air flow will not be a distinctive feature: all sounds will generally be pulmonar egressive. So, Liva's phonological system will be based only on place and manner of articulation.

By manner, phonemes will be distinguished in:

Ray Brown does not include voiced fricatives in his "briefscript" because they "have a tendency to weaken to approximates and become virtually silent" (LD 1999.01.13); however, Mark Rosenfelder notices that "just about everything has a tendency to weaken (and ultimately drop)" (LD 1999.01.13).

Each of these manners will be realized by four different places; in this respect, phonemes are defined only generically as:

A fifth place of articulation has been longly considered, which in the case of occlusives would have been realised by retroflexed sounds, like in Indian languages; but eventually it has been discarded in order to improve perceptual distinction.

2.2 : Phonemes representation and schema

This is the table of the 32 phonemes of Liva.

manner \ place labial front centre back
occlusives voiced  b  d  ð (6)  g
voiceless  p  t  þ (5)  k
nasal  m  n  µ (4)  q
lateral  ß (8)  l  £ (2)  r
fricatives voiced  v   z  j  h
voiceless  f  s  ç (c)  x 
 closed vowels  y  i   w  u
 open vowels  ø (&)  e  a  o

2.3 : Phonemes description and allophones

Voiced and voiceless occlusives are very common in natural languages, with the possible exception of central ones (palatals), which are found e.g. in Hungarian and Quechua. Front lingual occlusives, included nasal and laterals, can be pronounced as alveolars, as it happens generally in European languages.

Nasals are present in many natural languages: especially [m] and [n] are very common. The velar nasal in natural languages is often at word end; however e.g. Ligurian dialect uses it before vowels, as Liva does: with some training, it is not difficult to pronounce.

A dental/alveolar lateral is very common. A palatal one is found e.g. in Italian and Spanish. Velar lateral is less common, but it can be approximated by the velarized alveolar, as in English "hill". The labial lateral is Liva's most original phoneme: I (and my friends langdevils with me) am not aware of any language having one, neither of an IPA symbol to represent it; the easiest way to realize it seems to be by the tongue's tip between lips, either in the middle or at one side: in this position imagine to pronounce a common [l]; a more rigorous alternative, as it does not involve the tongue and is hence truely "labial", would be to put lips near only in their central part and leaving two spaces on the sides, though it seems a difficult position, or to put them near on a side and leave a space on the other side.

In a phonetic scheme, typical fricatives could appear shifted of one position backwards in relation to the occlusives (e.g. velar occlusive but uvular fricative): this can enhance their distinction; in Liva's phonological scheme, however, occlusives and fricatives can be showed in one column, so to make the representation simpler and considering the possibility of allophones. So, central fricatives are typically prevelar, like in German "ich", but they can also be pronounced, without risk of confusion, as the palatal fricatives common in French (jambon, choisir) and English (azure, sheep). The velar/uvular voiced fricative, which is not common (e.g. Czech and Slovak {h}), can also be realized as a French {r} slightly vibrant.

Of velar vowels, the labialized (rounded) versions are much more common, and they can be used. It is recommended that open vowels {ø, e, o} are pronounced as such. The prevelar closed vowel is common in French and German as "neutral vowel" ("schwa"); it can be thought as the closed or half-closed phoneme corresponding to the open /a/; it must not be labialized.

2.4 : Phonotactics

2.4.1 : Phonotactic criteria

Liva phonotactics, namely the rules according to which sounds can follow each other, is intended to be based on the same principles of easiness both in pronouncing and perceiving, as the definition of phonemes is.

Rick Morneau's essay <_> offers useful ideas for this purpose.

One factor to be considered for this purpose is the sonority of the phonemes. Manners of articulation have different sonority, according to the following scale:
Os < Oz < Fs < Fz < N = L < W < V
Adjacent phonemes should not both have low sonority.

Occlusives (included nasal and lateral ones) can be distinguished from each other mainly for their "solution", namely the final part of their pronounciation changing into the following phoneme: so they should not be in final position in words.

Word boundaries should be distinguished unambiguously, i.e. self-segregating. A very simple way to achieve this is stating that all words begin by phonemes of given classes. (3)

2.4.2 : Phonotactic rules

The following phonotactic rules must be respected:

  1. In a word, each phoneme must be of a different manner class from the preceding one: the pairs OO, FF, WW, VV are forbidden;
  2. O cannot be followed by F: hence they must be followed by either W or V;
  3. WV is forbidden, unless followed by another W: i.e., WVW is allowed
  4. Os, Fs, N, and L always and only occur as the first phoneme of words -- hence they are collectively abbreviated with "B" [?N and L, however, can also occur before homorganic Oz as the first phoneme of primitive fragments in derived words]
  5. Oz and Fz occur always and only as the first phoneme of primitive fragments in derived words

So, the following combinations are possible for primitive words:

of letters


number of

of words

2 BW

   16 x 4
+ 16 x 4
= 128
ku, no, ly, fa
[? BVF]
[? BWF]
   16 x 4 x 4
[+16 x 4 x 4
+ 16 x 4 x 4]
= 256 [+ ...]
qou, sei

2.4.3 : Elision of {w}

Two letter words ending in -w can be shortened, both in writing and in speaking, by dropping the {w}, where this does not make pronounciation and perception too difficult (voiceless occlusives can be especially problematic in this respect). So:

Even the apostrophe marking the elision can be omitted: .

3 : Morphology

Liva is an isolating language: each word has only one form, carrying a fixed meaning. So, words are not inflected for parameters like gender, number, case, tense, aspect; neither affixes (prefixes or suffixes) exist that can be attached to word stems. So each word consists of one morpheme.

3.1 : Primitive and derived words

Words begin by either Os, Fs, N, or L. As those letters can only occur in initial position, word boundaries are marked unambiguously.

Some words can be derived from two or more primitive words, having meanings related in some way to the meaning of the derived words. Derived words still are only single morphemes, as their precise meaning is defined in the dictionary, rather than inferrable univocally by the meanings of the primitive words.

So derived words cannot be said to be compounds. Like And Rosta said about his language Livagian, "there are no rules for forming compounds or for any other kind of derivational morphology. If the meaning of a compound is fully predictable from the meaning of its parts, then the compound is not necessary in the first place. If the meaning is only partly guessable from its parts then it doesn't really matter whether it's clear which parts a compound is made of. In Livagian you create a new word by whatever means you like -- you can toss in some recognizable components if you like, but the only absolute rules are that the creation must be phonologically licit and must not produce homophony." (LD 1997.10.03) A word like English "bluebird" must be defined independently, and the creator of the vocabulary can decide to do it by using the primitive words for "bird" and "blue", but they will just be a mnemonic help; such derived words are equivalent to Lojban's lujvo. As Tommaso R. Donnarumma noticed, a wide use of derived words pushes the language towards the controversial field of pasigraphies... On the other hand, a "blue bird" is a bird which happens to be blue for any reason: the relation between the two concepts is not specified, so the exact meaning can change; Liva will express this by two separate words; this is equivalent to Lojban's tanru.

In derived words, the second and the following primitive fragments begin by voiced O or F homorganic to the first letter of the original primitive words, according to the following scheme:

3.2 : Word classes

Liva words (morphemes) belong to few basic grammatical classes, defined on the base of the logical function of their meaning, and consequently their syntactic behaviour. Most natural languages have more grammatical classes ("parts of speech"), which do not entirely correspond to the logical function of the meaning of their words: for example, the English word "father" is grammatically a noun, although it implies a relation between entities, and hence should be more logically defined as a verb or a preposition.

The grammatical class of each word is marked by the phonological class of its first letter:

3.2.1 : Nouns

Nouns denote sets of things of the world. From the logical point of view, they are considered as constants.

Whether nouns logically are constants or functions is a big question. Both Lojban and Livagian treat them as functions, hence they do not distinguish a class of nouns from the class of predicates; so in Lojban predicates need to be coupled with other words (cmavo) to mean isolate nouns: to mean "dog" it is necessary to say "something-which is-a-dog" (le gerku). In Liva, the possibility of defining nouns as functions instead of constants has been considered: it originates the parallel language Liva C, in which nouns cannot stands alone but have always to be coupled with a specifier, and there is no need for a word meaning "to be"; deictics are the only constant word class, as they refer to objects in the world rather than to other words. In Liva B some "nouns" are constants while some others ("adjectives") are not, like it happens in many natural languages. Liva A, the present version, has been chosen to be developed even because it seems to be a more original experiment. Alternative versions are also different in being modifier-first (Liva A2, C2, etc), like English, or head-first (Liva A1, C1 etc.); head-first version is preferred by the author for deep "library" reasons.

Some common nouns are:

Nouns include proper names, which are specified by the specifier çi:

Variables are nouns derivated from the generic variable pøu and the words for numbers:

3.2.2 : Predicates

Predicates denote relations between things. From the logical point of view, they are functions of constants.

The syntagms (nouns or other predications) connected by a predicate are called its arguments. For example, the predicate "give" provides for three arguments: the giver, the beneficiary, and the thing given. Each predicate provides for a given number of arguments in a given order, depending on its definition in the dictionary. Liva predicates can be:

Provided arguments can be expressed or not. If some arguments are unexpressed, or repeated, ambiguity in the role of each argument is avoided by argument markers (3.2.4): marks argument x0, le marks argument x1, la marks argument x2, le marks argument x3. In symmetric predicates, any argument following the second one is marked by . E.g.:

3.2.3 : Specifiers

Specifiers make the set of things of the world meant by a noun or predicate more precise and limited. So they are sintactically depending on the noun or predicate to which they are appended, and have no meaning as themselves. For example, while ka means generically "some city", ka sew means "that city", ka xa means "two cities", etc.

Specifiers can be appended to other noun-specifier group, e.g.:

When they are appended to predicates, their meaning can apply to the whole predication:

A list of the different kind of specifiers with examples follows.

3.2.4 : Markers

Markers denote logical relations between concepts, by marking the syntactic function of words and syntagms. So they help to reduce the network of syntactic connections to bidimensional strings of words.

4 : Syntax

One or more words can make up meaningful unities called syntagms. Syntagms can be constituted by:

Arguments themselves can be nouns, predicates, predications, or specifications.

Syntagms are connected to each other either by their sequence, or explicitly by markers.

4.1 : Predication

The basic structure of a predication is a predicate followed by arguments:

The markers of the arguments (le, la, lo) and the end of predication (lw) are unexpressed, unless any argument is missing or in non-standard sequence:

4.2 : Specification

Nouns, predicates, or predications can well stand alone, without any specifier: then their meaning is a generic one. E.g.:

When a specifier is added, the meaning becomes more specific. A specifier plus the noun/predicate/predication it specifies make a specification:

A specifier can also be applied to another whole specification, making a more complex specification:

On the other hand, if a second specifier has to be attached to the same syntagm (a noun in our example) as the first one, rather than to the whole specification, it must be marked by ru:

When a specifier has to be attached to a long syntagm, it can be practical to express it before the syntagm, rather than after it; in this case, the specifier must be preceded by the specifier anticipator ri:

Relative specifiers connect a syntagm with a predication of which it is an argument:

If more words follow, the relative syntagm has to be terminated by the specification terminator rw:

5 : Vocabulary

Liva vocabulary is constructed a priori, namely from scratch and independently of any other existing language: unless the vocabulary of international auxiliary languages, Liva's one is not intended to be especially easy to learn.

5.1 : Selection of primitive concepts

A set of primitive concepts, i.e. concepts to which primitive words will be assigned, is individuated to make the basis of Liva vocabulary. The set is based on Rick Harrison's Universal Language Dictionary (ULD), a careful list of 1600 meanings commonly occurring in most languages; few additions and changes are made to the list.

Each entry in the list is then classified into one of the syntactical classes of Liva (nouns, symmetric predicates, diadic predicates, triadic predicates, specifiers), according to its essential meaning.

5.2 : Generation of primitive words

Phonotactic rules and morphology allow for a given number of words for each syntactical class (2.4). Shorter words are generally assigned to the most common or basic meanings. Word generation and assignation is random, with exceptions for the following cases:

5.3 : Derivation

Words for less basic meanings are defined by derivation from primitive words, according to the rules described in section 3.1:

Once they have been defined, both primitive and derived word always keep their precise meanings as reported in the dictionary.

|| Liva : a logical language # 3.2 / Claudio T' Gnoli -- Yahoo!-GeoCities <> = <> (2001.07-)
<< & -- <'liva.htm> (2001.01.05-2001.07)