From an experiment to a language
The story begins in 1955, when the American sociologist James Cooke Brown started work on a language he called Loglan, short for "logical language". He wanted to test the idea that the language people speak shapes how they think (see Does the language you speak shape how you think?). Would people who learned a language built on formal logic start to think more logically? He described the project in Scientific American in 1960, but the experiment itself was never carried out.
Brown regarded the language as his property, and disputes over its control followed. In 1987, a group of Loglan users founded the Logical Language Group and built a new version on the same principles but with a new vocabulary: Lojban, from logji bangu, "logical language". Its full grammar was published in 1997.
Even its vocabulary was designed by a method. For each of its roughly 1,300 root words, a computer program took the words for the same idea in the six most widely spoken languages of the time (Mandarin, English, Hindi, Spanish, Russian and Arabic) and picked the five-letter form that shared the most sounds with them, giving the larger languages more weight. As a result, many roots contain familiar bits for many people, but no one feels quite at home in them.
Words with slots
Lojban does not divide words into nouns and verbs the way English does. Its core words are predicates: relationships with a fixed number of places, each with a fixed role. The word klama means "go" and has five places: someone goes to a destination, from an origin, along a route, by some means. A full sentence fills them in order:
mi klama la .paris. la .london. lo dargu lo karce
"I go to Paris from London along the road by car."
You do not have to fill every place. Places left off at the end are simply unspecified: mi klama la .paris. ("I go to Paris") and even mi klama ("I go") are complete sentences. Only when you want to skip a place and fill a later one do you need the filler word zo'e, "something unspecified": mi klama la .paris. zo'e zo'e lo karce, "I go to Paris by car".
A few small words do the rest of the work. lo turns a predicate into a thing: karce means "is a car", lo karce "a car". la marks a name. And names are set off by short pauses, which Lojban writes as dots, so that a listener can always hear where a name begins and ends.
One structure for every sentence
Take the English sentence "I saw the man with the telescope". Who had the telescope, you or the man? English allows both readings. In Lojban they are two different sentences, because a phrase about how the seeing was done and a phrase about the man are attached in different ways:
- mi pu viska lo nanmu se pi'o lo tutci: I saw the man, using the instrument. The tag se pi'o, "using", belongs to the seeing.
- mi pu viska lo nanmu poi ponse lo tutci: I saw the man who had the instrument. poi begins a description of the man.
Logical words are just as exact. English "or" sometimes means "one or the other, not both" ("coffee or tea?") and sometimes "either, or both" ("do you have a pen or a pencil?"). Lojban has separate words: .a for "and/or", .onai for "one but not both", and .e for "and".
The grammar is written down as a formal grammar, like that of a programming language. A computer can check any Lojban sentence and find its one possible structure.
That does not make every meaning exact. Lojban builds new ideas by putting predicates side by side, and the relationship between them is left open: ladru botpi, "milk bottle", can be a bottle with milk in it or a bottle made for milk, just as in English. A clear structure is not the same as a clear meaning, and in Lojban, too, context fills the gaps.
Feelings and evidence in words
Spoken English shows feelings largely through tone of voice. Lojban gives them words, called attitudinals, which can be added to a sentence without changing what it claims: .ui for happiness, .uu for pity, .oi for complaint. .ui mi viska do means "I see you", with the speaker's pleasure attached.
Another set of words, the evidentials, tells the listener how the speaker knows something:
- za'a: I observe it myself.
- ti'e: I heard it from someone.
- pe'i: in my opinion.
- ru'a: I assume it.
za'a carvi means "It's raining (I can see it)"; ti'e carvi, "It's raining (I've heard)".
This sounds like an invention, but many natural languages do it in their grammar. In Turkish, geldi means "he came", said by someone who knows it, for example because they saw it; gelmiş means "he came, apparently", because the speaker heard it or concluded it. Quechua marks the source of almost every statement with a small ending: -mi for what the speaker knows first-hand, -si for what they were told. In Lojban evidentials are optional; in Turkish, the past tense makes you choose. For anyone learning Turkish, this is one of the new distinctions to notice.
Time is optional
In Lojban, a sentence does not need to say when something happens. If time matters, small words add it: pu (before), ca (at the same time), ba (after). mi pu klama means "I went", mi ba klama "I will go", and mi klama leaves the time to context.
Here too, natural languages got there first. Mandarin verbs do not change for tense: wǒ qù can mean "I go", "I went" or "I will go", and words like zuótiān (yesterday) or the context make the time clear. In a world survey of 222 languages, 88, about two in five, had no grammatical past tense (Dahl and Velupillai). For a learner whose own language marks tense on every verb, such languages can feel vague at first; for a speaker of Mandarin, English verb forms are an extra choice to make in every sentence.
Who speaks it?
Lojban has a small, mostly online community, and only a small number of people speak it fluently. Speaking it in real time, with every place and connective in the right order, takes a lot of practice. Brown's original question, whether a logical language makes people think more logically, has never been tested: too few people have learned Lojban to find out.
What Lojban does show is that a language can have a grammar without structural ambiguity, and that many of its "inventions", from evidence markers to optional tense, already exist in the languages people speak every day.