SUL: A Technical Reference Grammar

This page is a formal description of Simple Universal Language (SUL) for linguists, conlangers, and the professionally curious. The lessons teach the language gently - this page states the rules exactly, in plain English with the standard linguistics terms alongside. Every rule of the language is on this page - there are no others.

1. Typological profile

SUL is an engineered international auxiliary language. Typologically it is strictly agglutinative: every word is built by chaining morphemes - the smallest meaning-carrying pieces, roots and grammar bits - with an audible linker letter between them, and the pieces never fuse and never change shape. There is no inflection of any kind - words never bend for grammar: no agreement, no case, no verb conjugation, no irregularity. Word order is rigidly SVO (subject, then verb, then object). The design goal that drives everything below is a small, closed rule set with literally one exception class (foreign names), plus a lexicon engineered so that every utterance is segmentable and parseable without ambiguity - by ear, by eye, or by touch.

2. Phonology

2.1 Segment inventory

SUL has exactly 12 phonemes (distinct sounds): 10 root phonemes (5 vowels, 5 consonants) and 2 linker consonants that occur only at morpheme boundaries. One letter, one sound - no digraphs, no silent letters, no sounds that shift depending on their neighbors (allophony) for a learner to memorize. IPA values are targets - wide phonetic tolerance is intended, and the guard-slot system (5.5) engineers the most dangerous fine-detail confusions out of the grammar.

No other segment exists: b, r, m, j, l, x, h and the rest are not SUL letters. The linkers n and w never occur inside a root. This is constitutional, not incidental: because no root contains a linker, every n or w in a word is a morpheme boundary, so any word splits into its pieces in exactly one way - by ear, by eye, or by machine. Fusion of morphemes is never legal.

2.2 Phonotactics

• Every root has the shape (VC)+: a chain of vowel-consonant syllables. Roots begin with a vowel, end with a consonant, and have an even letter count (2, 4, 6...). Consonant-initial and vowel-final words do not exist.
• The word-slot inventory is generated exhaustively at lengths VC, VCVC, and VCVCVC: 25 + 525 + 11,025 forms. Generated forms never change - only meaning assignments move between slots.
• Generator constraint: adjacent syllables may not repeat the same vowel, unless the earlier syllable's consonant is g (asaf and asag are excluded - agas and agat are kept).
• Roots contain no consonant clusters and no vowel sequences - the (VC)+ template makes them impossible. The only consonant-consonant contact in the language is a glue letter landing after a root's final consonant, and that contact is itself the audible morpheme boundary.

2.3 Prosody

Pitch never changes a word's meaning (no tone, unlike Mandarin) and stress never distinguishes two words (unlike English REcord vs reCORD). All rules of rhythm, stress, and pitch exist for one purpose: making morpheme and word boundaries recoverable by ear.

• Stress the first syllable of every morpheme. Each stress beat announces a new morpheme.
• Linkers are pronounced short and soft, never stressed - an audible "joint" between stressed morphemes.
• Slightly lengthen the vowel of a word's final syllable. Since all words end in a consonant, this is the vowel before the last consonant - the lengthening is the word-boundary cue.
• Question intonation rise is optional - the question morpheme already marks interrogatives.
• Digit strings are spoken on a steady beat with a small pause every three digits.
• Pausing anywhere inside a word never changes meaning - prosody carries no semantics.
• Sentence end: the final syllable falls to low pitch and its vowel is stretched a touch longer than an ordinary word-final stretch - the spoken mirror of the written rail break. A following pause is natural but optional - a pause alone never marks anything, so hesitating mid-sentence is always safe.

SUL text takes no punctuation, ever. Exclamation lives in the voice - interrogation lives in the question morpheme. The word and sentence separators below are script glyphs - a baseline rail and its absence - not marks attached to the text. (Punctuation you see in lesson notation belongs to the English, not the SUL.)

3. Morphology

3.1 The two linkers

All morphology is concatenation with one of two linkers. In lesson notation they are written - and + - in the language itself they are the letters n and w.

• n (notated -), the flat chain. Left-scope rule: an n-chained morpheme takes scope over everything before it in the chain. eat-no-want is want(not(eat)) - "to want to not eat".
• w (notated +), the nested chain. A w-chained morpheme modifies only the single morpheme immediately before it, and takes exactly one morpheme itself - it cannot attach a multi-morpheme chain. eat-want+no is "to not want to eat". Its uses are one-word tweaks (fast+very) and scope disambiguation.
• + may stack. A run of stacked + follows the same left-scope rule as -, restarted mid-word: each +-morpheme applies to everything before it since the last - (or the word start), and the run ends at the next - or at word end. story-bird+big+angry = "story of the big angry bird" (big→bird, angry→big bird). There is no fallback and no reattachment: if a +-morpheme is an illegal fit for its scope (a degree word on a non-gradable, negation in a thing-chain), the word is simply ill-formed - rephrase as separate words.

Left-scope is the law - different linear orders are different meanings, never free variants. The following minimal pairs are fully productive:

3.2 The word

One word = one chain = one root = one sentence role. A chain is a single word - spoken as one unit, acting as one unit in the sentence - built on exactly one root. The common-case morpheme order is a slot template - but the template is not an independent rule - it is simply the order the left-scope rule produces for the most frequent meanings, kept as the default:

When the intended meaning needs a different scope, the chain follows the meaning, as in the scope table above. When two orders would produce the identical meaning under left-scope, the slot-template order is the only legal spelling - order variation exists solely to mean something different. Two constraints partition the template: ownership (slot 6) is a chained pronoun owner and appears only on thing-chains (dog-me = "my dog" - a noun owner is an ordinary description-slot modifier, dog-teacher = "the teacher's dog") - subject, tense, negation, and mood appear only on action-chains. No chain uses both sets. An unbroken run of chained digit morphemes forms ONE number for all scope purposes: dog-1-2+ordinal = "the 12th dog", with the ordinal covering the whole number.

3.3 Chaining restrictions

• Only pronoun subjects chain onto verbs. Noun subjects, objects, and adverbial phrases are always separate words. The principle: chaining is legal exactly where it is unambiguous. A pronoun is already fully specific - there is no pile of me's for a chained morpheme to narrow down - so me-fast can only be a claim ("I am fast"). A noun's chain is already spoken for by naming (dog-fast = which dog), so a noun subject gluing onto a verb would collide with it.
• Connectors never chain - if, because, until and the rest always stand alone between clauses.
• A modifier with its own modifier normally stays in the chain with + (xhe-sing-bird+similar). When the meanings are identical the in-chain form is THE spelling (one way to say it) - the split form (xhe-sing bird-similar) is used only when no legal in-chain form exists - + takes exactly one morpheme and must be a legal fit.
• Degree words grade only gradable morphemes (actions and descriptions) - a degree word aimed at a thing is ill-formed with either linker. There is no in-word way to intensify a noun compound itself - give it a gradable modifier and grade that: child-dog-bad+very = "an utterly BAD son of a dog".
• Stacked moods disambiguate by left-scope alone: me-eat-want-should = should(want(eat)). No extra rule.
• Reduplication does not exist: repeating a morpheme is never a device of the language. Emphasis is expressed with the degree word very (good-very, no-very).

3.4 No derivational apparatus

SUL has no derivational morphology distinct from its vocabulary. The antonym, plural, tense, agent-nominalizer ("doer"), and mood morphemes are ordinary dictionary words that chain like any other word. They are chain-only: a bare grammar morpheme never stands alone as a word, and never starts a chain - it attaches to a word before it. The two exceptions are no (two standalone uses, see Syntax) and the mention marker (section 6). The core grammatical morphemes, straight from the live dictionary:

4. Syntax

• Order: SVO, fixed. Describing chains sit next to what they describe. A bare verb may stand in object position and names the activity itself ("we-go eat" - eating is what we are going to) - this is plain SVO, not a serial-verb mechanism.
• Zero copula (no "is" before descriptions): a description after the subject acts as the verb itself, with no is-word (the Indonesian/Mandarin pattern): a pronoun subject chains onto it (you-smart = "you are smart"), a noun subject stands before it as its own word (dog fast = "the dog is fast"), and it takes tense, negation, and mood like any action chain (dog fast-past = "the dog was fast"). The is-word ("to equal") is an ordinary verb, not a connector: it is reserved for equality between thing-chains, never appears with a description predicate, and behaves like any verb - noun subjects stand alone, pronoun subjects chain onto it (it-is), and tense, negation, and mood chain on (world is-no light-sun = "the world is not sunshine"). Disambiguation is structural: chained dog-fast is the noun phrase "the fast dog" - separate words dog fast are a sentence.
• Tense: the verb never changes shape. Present is the unmarked default. Exactly two tense markers exist, past and future, chained onto the end of the verb (me-go-past, never me-past-go). There is no aspect system - to stress "right now", chain the ordinary word now onto the verb.
• Tense marking is obligatory for non-present, even when a time word is present: the verb marker is grammar (which direction in time), the time word is vocabulary (when exactly). me-eat-past day-1-past is not redundant.
• Time words are composed, not stocked: yesterday = day-1-past, tomorrow = day-1-future, standing as their own words - the count scales (day-3-past = 3 days ago) and the unit swaps (year-1-past = last year). A bare unit-past (day-past) is only the vague "a past day". The ordinal is wrong for offsets: day-1+ordinal = "the 1st day" of some sequence, never tomorrow. Aspectual nuance is carried by time words, not markers.
• Negation: chain the ordinary word no - it takes left scope over everything before it. Narrow negation nests with +.
• Questions: the question morpheme chains as the very last morpheme of the sentence's final chain. Question words (who, what, where...) sit in the sentence as ordinary words, in the exact position where the answer would go - English moves them to the front ("what you-see"), SUL never does (you-see what-question).
• Mood: want, should, command, can and the like are ordinary words in the mood slot.
• Embedded clauses: a dedicated complementizer opens an embedded clause - the whole clause then acts as one part of the outer sentence (xhe-know that me-think that you-smart). Its only meaning is "an embedded clause starts here" - it is never a demonstrative or a relativizer (English fuses those three jobs into one word - SUL does not), and it is never dropped.
• Relative clauses: a dedicated relativizer opens a description clause immediately after the thing it describes: dog which eat-past run-past = "the dog that ate ran". It covers English who, which, and relative that - like all connectors it stands alone, never chains, and is never dropped. Inside the clause the described thing is omitted (a gap), and the gap's role is determined, not guessed: a clause with no subject gaps the subject (dog which eat = "the dog that eats") - a clause with a subject gaps the object (dog which me-see-past = "the dog I saw") - location gaps ride on exist (building which me-exist = "the building I am at"). Complementizer and relativizer are distinct words for distinct jobs: me-tell person that xhe-go = "tell the person THAT xhe goes" - me-tell person which go = "tell the person WHO goes".
• Demonstratives: this (near speaker) and that (away from speaker) are ordinary roots in the description slot - dog-this, dog-yon - and stand alone as pronouns. Same for the place pair here/there. Both members of each pair are roots per the pointing-word (deictic) exemption to positive-pole-only (5.2). (The far demonstrative is glossed 'yon' in the dictionary and in notation, since 'that' is reserved for the complementizer.)
• Number: bare nouns are number-neutral (the Indonesian/Mandarin pattern) - dog covers one dog, several dogs, or dogs in general, and context carries the count. Chain the plural morpheme only when more-than-one-ness is itself the point (you-plural = "you all") - chain a number when the exact count matters (dog-3), never both. Pronouns keep their inherent number. There are no separate plural pronouns, no gender or formality distinctions, and no inclusive/exclusive distinction - one we covers both.
• No articles: no word for a/the exists. Bare nouns are definiteness-neutral - the demonstratives, numbers, and context do the work.
• Prepositions: in, on, to, with, about and the like are ordinary words for precise relations. A preposition stands alone before its noun, never chains, and the phrase sits at the end of its clause. There is no word for at: plain location goes through exist (xhe-exist home = "xhe is at home") or a bare place word at clause end - a preposition appears only when the exact relation (inside, on-top, together-with) is itself the point.
• Adverbial placement: adverbial words and phrases (time, place, prepositional) sit at the end of their clause, after any object - when several stack, place precedes time (me-eat-past food home day-1-past = "I ate food at home yesterday"). Positions are fixed - placement is never stylistic.
• Comparison: the comparative markers more/less chain onto the description (fast-more = "faster") - the standard follows the connector as: you-smart-more as me = "you are smarter than me". Bare as with no marker is the equative (the "as fast as" construction): you-fast as me = "you are as fast as me".
• No passive voice: the doer stays the subject. With an unknown or unimportant doer, use they generically, or make the patient the subject of get with the bare activity verb as its object (window get hit = "the window got hit") - plain SVO, no added mechanism.
• Ditransitives (give-type verbs): exactly one object may follow the verb - the recipient rides in a clause-final to-phrase: me-give book to you. Never two bare objects.
• Coordination joins whole units: a connector joins two complete clauses or two thing-words (dog and cat run). A shared subject pronoun is repeated in each clause: me-eat and me-drink, never me-eat and drink.
• Standalone no is legal in exactly two places: as the one-word answer, and directly after a connector, where it negates the whole clause that follows ("if no X" = unless X, "but no X" = except X).
• Answers state the truth of the proposition as asked: yes means the statement - including any negation in it - is true - no means it is false. you-go-no-question answered yes = "correct, I do not go". There is no English-style polarity flip.
• Tense is absolute in every clause: markers always point from the speech moment, never from the outer clause's time - xhe-say-past that xhe-go places the going at speech time.
• Reflexive and reciprocal: the self word stands alone in object position and points back at the subject (you-ask self = "you ask yourself") - with a plural subject it also covers "each other" (they-see self) - context carries the difference.
• Zero quantity: negation never sits in a thing-chain - "no dog" is the digit 0 in the quantity slot - dog-0 come-past = "no dog came" - person-0 = nobody, thing-0 = nothing.
• Counterfactuals (if-sentences about unreal situations) take no special marking: if + literal tense covers real and unreal conditions alike - context carries the unreality. No subjunctive (special unreal-verb form) exists.
• No irregularity by design: no irregular verbs, plurals, spellings, or numbers exist. The one real exception class is foreign names (section 6).

The pronoun inventory in full:

Connectors always stand alone between clauses. Per the bijection rule (one word per meaning, 5.1), several English connectors map onto one SUL word - an English lexical split is not treated as a sense split (see 5.4). There are no words for then, unless ("if no"), or except ("but no") - they compose.

5. Lexicon design

The dictionary is not a list that grew - it is a designed artifact with hard rules. The live dictionary (browsable here) is the sole source of truth for all vocabulary.

5.1 One word per meaning, one way to say it (the bijection rule)

• Every word has exactly one meaning, and every meaning has exactly one word. Two words never mean the same thing (English has synonyms like big/large - SUL never does), and one word never carries two unrelated meanings (English bank is a river edge and a money building - SUL bans this).
• A word's single meaning is allowed to be broad, as long as it stays one idea: big can describe a house, a number, or a problem. When a finer shade of meaning is needed, modifier words chain onto a word you already know - a new word is never coined for nuance.
• The language also never offers two equally correct phrasings of the same thing - for any meaning there is one right way to say it. This is deliberate: every time you hear a meaning it is worded the same way, so everything you hear is a repetition you can reuse.

5.2 Positive-pole-only antonyms

For opposite pairs of descriptive words, the dictionary keeps only one side ("pole") of the pair as its own root - the other side is said by chaining the antonym morpheme. True is a root, false = true-antonym - full is a root, empty = full-antonym. Grammar and answer particles are exempt: both poles stay roots there (past/future, yes/no), because they are grammatical machinery, not descriptions. Top-of-speech poles are exempt on frequency grounds: a pole inside roughly the top ~400 words of everyday speech gets its own root even though a chain covers it, because at that frequency the chain would cost an extra morpheme in nearly every sentence and audibly lengthen the spoken language - bad, death (covering dead and die), less, out, down, off, old, big, back, right, wrong, hard, high, late, under, end, last, and never are all roots. The antonym chain stays legal grammar (good-antonym is understood), but the root is the word. Color pairs are exempt as well - white and black are both roots, because colors name points on a scale, not poles of a description. The temperature pair is exempt the same way: hot and cold are both roots - they name regions of a scale (lukewarm is neither), and cold is not a mere negation of hot. Pointing-word (deictic) pairs are exempt too: this/that and here/there all stay roots - they are high-frequency, and the far pole positively points somewhere rather than merely negating nearness, so X-antonym does not capture it. Safe and danger are both roots: danger is a positively present threat, not the mere absence of safety, and a warning you shout must be word-sized. Taste words are exempt like colors: sweet, sour, and bitter are all roots - they name distinct taste qualities, not poles of one scale, and bitter is not the opposite of sour. Colors never moonlight as light-levels - darkness is light-antonym, never black, because black is a color. The bare antonym morpheme can never stand alone as a word, and an antonym chain is legal only where the opposite is unique: a root with more than one opposite rejects it entirely - old opposes both new and young, so old-antonym is ill-formed - say new or young directly.

5.3 Tiers and the 500-root ceiling

• Base 1 is ~100 words (survival vocabulary, grammar particles, UI words) - Base 2 brings the base to ~300 - the hard ceiling is 500 base roots - root #501 requires evicting one.
• Tier membership is derivable from position - no tier column exists in the dictionary: Base 1 = the assigned words in the first 100 generated slots, Base 2 = every other assigned root, extended = the n-chained compound rows.
• A small dictionary is not a goal: SUL is a full language and its lexicon grows like one - the tiers cap what lessons teach, not the language. Simplicity comes from frequency: the most common meanings get the shortest forms.
• An unbounded extended tier holds curated transparent compounds (spelled with the n linker - w is also legal where the meaning needs nesting - and every component a live root), with no completeness goal. Lessons and site UI use base words only.
• A dictionary compound owns its string: a chain spelled identically to a curated compound IS that compound and means what the dictionary says (water-salt = "the sea", never ad-hoc "salty water"). Bijection holds for compounds exactly as for roots.
• Every word is kept as short as its frequency merits, and during development no assignment is permanent: when a more popular meaning - or one better suited to a short form - arrives, the incumbent is shuffled out to a longer form and the short slot reassigned. The shortest forms belong to grammar morphemes and the highest-frequency words. Length tracks commonness in every tier: everyday meanings get roots, fairly common meanings cap at a two-root compound, and only genuinely rare meanings wear longer chains. Once the language has real speakers, assignments freeze.
• No root exists for what a chain already says - if a meaning is cleanly expressible with existing morphemes, it gets no root (false is just true-antonym). The frequency exception: everyday high-frequency meanings get short roots even when a chain could say them, because daily vocabulary must be word-sized. Rule of thumb: a meaning inside roughly the top 2000 most common English word families deserves a root unless an antonym chain or stretchy near-synonym already covers it - except the top ~400 of everyday speech, which get roots regardless - outside the top 2000 it stays a chain or becomes a curated extended-tier compound (rainbow = color-rain was the first). The top-2000 gate is applied strictly only to verbs, adjectives, adverbs and abstract or functional meanings - concrete object nouns (coffee, key, cup...) may stay extended-tier compounds even inside the top 2000: a picturable thing tolerates a longer transparent spelling - grammar-adjacent and abstract vocabulary does not.
• Compounds put the head root first, then modifiers: a learning place is place-learn, not learn-place. A compound's semantic head should itself be a root - compounds headed by compounds stack length silently (beer as alcohol-bread came out 14 letters because alcohol is itself a compound).

5.4 Stretchy connector roots

When English splits a meaning that SUL does not, one root stretches across the near-synonyms: when covers while, because covers causal for, from covers since and after, and as covers both the comparison standard (X fast-more as Y = "X is faster than Y") and the equative (X fast as Y). The exist root stretches the same way (Indonesian ada, Korean itda): it covers existence, "there is", and location - water exist = "there is water", xhe-exist home = "xhe is at home". A near-synonym root is only added if a real sentence shows the readings diverge in a way no chain can fix.

5.5 Guard slots: engineered mishearing resistance

k and g differ only in whether the vocal cords hum (voicing), and f and s are both quiet hissing sounds (fricatives), so two words one such substitution apart are easily misheard as each other - and in a dense generated slot inventory, the misheard form is usually another real word, so the error goes undetected instead of self-correcting. SUL engineers around this:

• The 2-letter tier is thinned to 15 of its 25 forms - per vowel: one of Vg/Vk, one of Vf/Vs, plus Vt. The other 10 forms (af, ag, es, eg, if, ik, of, og, uf, ug) are permanent guard slots that must never be assigned - grammar particles and one-word answers are exactly where context cannot rescue a misheard consonant.
• The 2-letter tier is reserved for grammatical morphemes (pronouns, tense, plural, negation, yes/no, connectives, derivational affixes) - a content word never outranks an affix there.
• New assignments prefer guard-safe slots: forms with no assigned k/g or f/s one-substitution neighbor. Where a confusable neighbor is unavoidable, the two meanings must be cross-category (different sentence roles), never two words plausible in the same position.
• Content-word confusable pairs are otherwise accepted - context disambiguates them, as in natural languages.

5.6 Numbers and measurement

• Base 10. Digits 0-9 are audibly distinct short words - larger numbers are spoken digit by digit, with the small pause grouping digits in threes from the right (1 234 567).
• An unbroken digit run is one number morpheme for all scope purposes (see 3.2).
• Ordinals: chain the number onto the thing with n, then the ordinal morpheme onto the number with w (dog-3+ordinal = "the third dog" - dog-1-2+ordinal = "the 12th dog").
• Decimals and negatives compose from ordinary words: the decimal separator is the word point between digit runs (3.14 = "3 point 1 4") - a negative number chains the antonym morpheme (5-antonym = -5, a number's unique opposite). No percent word exists - proportions are decimals - half is the everyday fraction root and other fractions are said as decimals.
• Quantities are unit-chains: the unit root is the head and the number fills its quantity slot (meter-3 = "3 meters"). Dates and times are the same pattern in ISO order - year-2026 month-7 day-30, hour-14 minute-30 - each unit its own word, no leading zeros. The "2026-07-30" digit notation is international notation, not SUL text.
• Metric only - SI units are native roots, not loans. 24-hour time only. Temperature is Celsius only, with the degree as a native unit root.

5.7 Slang

Slang never enters the dictionary. It lives only on the slang lesson page, composed at render time from existing roots. Slang bends register, never grammar - with one tolerated rule-break: the grandparent's welcome eat-eat, a doubling. Doubling remains meaningless in the grammar (it marks the plural in many natural languages, so it is too ambiguous to be a device) - eat-eat survives as knowingly broken grammar, exactly like "eat, eat!" or "mangia mangia" - incorrect in English and Italian too, but grandmothers say it and nobody corrects a grandmother. It licenses no other doubling.

5.8 The language's own name

The language's own name is the compound language-all-simple - by left-scope, simple(all(language)) = "Simple Universal Language". It is the wordmark in the site header, and like all SUL text it renders from the live dictionary:

okuslanguagenasokallnuteksimpleokusnasoknutekookkuussnnaassookknnuutteekkHear

5.9 Sound-feel guideline

Consonant texture should not fight the meaning. f and s are the soft consonants, t mild, g harder, k hardest - and the final consonant colors a word most. Warm and social words (greetings, gratitude, affection) prefer f/s - harsh or dangerous meanings may take k/g - and percussive k/g is deliberately kept on energetic words (laugh, party, dance, kiss): a burst of sound for a burst of energy. This is a guideline steering new assignments, not a law.

6. The one exception: foreign names and quoted words

SUL has no loan words. It admits exactly one exception class: proper nouns - people, places, brands, pets, including names SUL speakers coin themselves - and quoted words (metalinguistic mentions - how do you say "dog" in English). A proper name is never an SUL word, so the bijection rule can never collide with a name. These are quotations, not SUL words:

• They are never translated and may break phonotactics freely.
• A dedicated marker word is said as its own word before every mention, not just the first. Its entire meaning is "the next word is mentioned, not used - it is not an SUL word of this sentence" - this covers foreign words, all names, and metalinguistic mention of an SUL word itself.
• In writing, the foreign word stays in its own native script, never SUL symbols.
• In speech, it is pronounced the way its owner pronounces it, or the closest approximation one can manage - approximation is explicitly permitted.
• A name never chains: nothing attaches to it, and it attaches to nothing. In a grammatical role it stands as its own word next to what it relates to - an owner name follows the head chain (dog [marker] Maria = "Maria's dog"), and a subject or object name sits in its SVO position.
• Chained grammar skips names: the question morpheme lands on the sentence's final SUL chain, and a sentence-final name is transparent to it (you-see-question [marker] Maria = "do you see Maria?").
• No direct-quotation mechanism exists: speech is reported with the complementizer clause - exact words are mentioned one by one with the marker.
• Accepted edge case: a foreign name that happens to sound like a valid SUL word is ambiguous in speech - writing disambiguates via script. Rare, accepted.

7. Writing systems

SUL has three parallel scripts - print, braille, and sign - sharing the same 12 letters and identical grammar. The written language and sign and braille lessons show them in action.

7.1 Print

• Two forms: expanded (one glyph per letter) and condensed (each VC syllable collapses into a hand-designed ligature square - one letter is sometimes scaled or nested for legibility). In a condensed square the vowel is always read first - every root starts with a vowel, so no reading-order marker is needed.
• Because every root is (VC)+, running condensed text is squares only - standalone letter glyphs appear solely in teaching material.
• The ten digits 0-9 have their own square glyphs - the script's only logograms: each writes the whole digit word, reads aloud as that word, chains like any morpheme, and shares the square grid, so numbers keep the monospace rhythm of condensed text. They are identical in expanded and condensed form. No other numeric symbol exists - the decimal point is the point word, a negative is the antonym chain, and dates are unit-chains.
• Linker marks sit above the text, and only linker marks may appear there: n is a single dot above the start of the following symbol (the one its sound leads into) - w is a short horizontal line reaching back to the left, pointing at the symbol its morpheme modifies.
• The rail system: the word separator is a line along the baseline bridging the gap, so a sentence is one unbroken rail from first word to last. The sentence separator is a true blank - the rail's break IS the sentence end. A line that ends mid-sentence ends with the rail line (keep reading below) - a line that ends bare has finished its sentence. These are the script's only two non-letter glyphs.
• Layout convention: one sentence per horizontal line, so line and rail coincide - the line's end is the period. Text runs on and wraps only where space is at a premium (novels, dense print, narrow columns).

7.2 Braille

• Derivation rule: each stroke of the printed letter becomes three raised dots - both endpoints plus the midpoint - on the stroke centerline, on the same grid. The linkers follow suit (n is one dot inside its printed dot - w is three dots along its printed dash), and the digit glyphs derive by the same stroke rule.
• Braille metrics are identical to the print font, so braille is embossed directly on printed text - dots on strokes. One printing serves sighted and blind readers - every book is its own braille edition.
• The word-separator rail becomes three raised dots like any other stroke, so a fingertip never loses the thread mid-sentence: dots in the gap mean the sentence is still running, and a smooth blank is the sentence end - sentence length is knowable by touch.
• Collapsed braille squares are hand-designed dot patterns tracing the printed square's strokes, with nothing above them - only linker marks may appear above the text.
• Design policy: public text - signage, menus, packaging - carries the dots molded in by default. Tactile is the normal product, not a special edition, and touchable copies sit at hand height.

7.3 Sign

• Vowel handshapes encode the vowel by number of extended fingers (a = 1 ... u = 5) - all letters and digits have handshapes (fingerspelling).
• Fingerspelling is the fallback: every word is to get its own dedicated sign motion, recorded in the dictionary alongside its other forms - the motions themselves are not yet designed, so fingerspelling carries everything today.

8. Worked examples

Each row gives the English, an interlinear gloss (SUL morphemes with English underneath, linkers shown), the condensed written form with a letter-by-letter breakdown, and audio.

That is the whole language. Everything else is vocabulary - browse the dictionary to see it.