The Song of Maa

A Philological Edition of Catoid Text 1234, with Grammar and Lexicon

edited from the recovered documents, with the anonymous Grammar (Gr) reassessed

ss ss miao maa mrr "no, no — come: far place, held long. It is good." (O, ed.)

Preface

Catoid reaches us only as a written language. No speaker, no prose record, no bilingual inscription of the ordinary kind survives; what survives is one song in two documents — a text and a companion sheet — and a grammatical treatise that postdates both and depends on them. This monograph edits the song, reconstructs what can be reconstructed of the language from it, and tests the treatise claim by claim against the only direct evidence we have.

Three documents carry the whole discipline.

T, the Text: a complete song, archived under the number 1234 in the series designated mmx. Forty-five lines: seven bracketed rubrics (intro, verse, chorus, verse, chorus, bridge, outro) and twenty-eight lines of writing — 165 word tokens, 91 distinct word types. The script is a plain-letter romanization: lowercase letters only, no diacritics, thirteen distinct letters carrying the whole song (the bracketed rubrics aside).

G, the companion Sheet, bearing the same number 1234. It carries a grammatical preamble (a consonant and vowel schema, word-edge catalogues, a rhyme-scheme note, and the identification of the title word), and then, line by line against T, an English gloss, a syllable count, and occasional performance marks. The Sheet's word sequences collate with T verbatim, line for line, throughout. Its English is a mixed register — archival computing idiom ("record," "cursor," "putRecord") beside contemporary colloquial ("zoomies," "if it fits, sits") — and I make no apology for dwelling on that fact, because it is the only internal evidence we possess for the Sheet's hand.

Gr, the Grammar: an anonymous treatise, self-titled catoid — a language sketch, describing the language as spoken. It is a synchronic grammar with a phonology, a closed class of seventeen function words, an open lexicon of eighteen roots, exemplar sentences, and a doctrine of registers. Crucially, Gr reports its own relative chronology: it remarks, with apparent surprise, that the songs "were generated before the grammar was written," and it quotes the refrain of our Text verbatim. Gr is therefore a secondary witness — later than the material it describes — and I treat it as such throughout: a source to be cited, checked, and where necessary corrected, never a substitute for the Text.

Nothing else is before me. Gr invokes a generative procedure — "the engine" — said to reproduce the root inventory from the dialect key, and asserts that other numbered redactions are "legitimate dialects of the same language." The engine has not been recovered; the other redactions exist, if at all, outside this edition. Everything I say below rests on T and G, with Gr admitted as testimony.

What is attested against what is reconstructed. Every form printed plain in this edition occurs in T. Forms reconstructed or transmitted only by the treatise are asterisked (*mu, *myu, *hrr). Editorial segmentations of attested forms — and I make very few — are marked with a dagger (†). The proportion matters and should be said at the door: of the seventeen closed-class items the treatise reports, ten are attested in T; of its eighteen open-class roots, two are; of its spoken phonology, the song register attests perhaps half the segments and contradicts none. The largest part of this edition is therefore honest annotation of a small text, not grammar-building on air.

The results I would put in a letter to a colleague: the Sheet's scansion is not decorative but exact, and I verify all twenty-six of its syllable counts under two simple rules (§3.2). The two strophes are bound by a sense-for-sense responsion and divided by their line-final particles — the stray strophe negates, the adopted strophe demands and completes (§3.1). The song register reduces the vowel system to three and drops every palatal and guttural hiss the Grammar assigns to the spoken tongue, while liberating percussives the Grammar confines to a single morpheme (§4). And the Grammar itself, where it can be checked, holds: its closed class is quoted exactly, its clause-final particles are confirmed, its reduplication is everywhere. Where it cannot be checked, I say so (§12).

— the editor

Contents

1. The witnesses and their provenance

1.1 T, the Text

T is a single complete song, the only continuous catoid document recovered. Its physical form is unknown to me; it reached the archive as a plain-letter transcription in the mmx series, catalogue number 1234, and I edit it in that romanization. Structure:

28 text lines, 165 tokens, 91 types. The letters used, in order of frequency: m (228), i (143), a (113), o (47), k (43), w (32), s (22), t (9), r (6), d (3), p (1), n (1), g (1). Thirteen letters; no u, e, y, h, or l anywhere in the text (the bracketed Latin rubrics aside). That single fact organizes half of §4.

1.2 G, the companion Sheet

G is titled catoid theme — seed 1234. The word "theme" is the Sheet's own; "seed" appears to be a term of the Glossator's art, attached to the catalogue number, and I confess I do not understand it. The preamble runs, in full:

C[m k t s w l d ng] V[i a o] init[d k m ng s t w] fin[d k l m ng s t] scheme ABAB

title word: maa (hook tail -aa)

Thereafter G gives, for each of the 28 text lines, an English gloss and a syllable count (26 lines carry counts; the intro and outro do not), with six line-final particles set off parenthetically with an exclamative — (ss!)and (miao!) twice each, (mrrp!), (mrr!) — six tokens, four types — a notation T itself does not have. The two verse rubrics are annotated rap · stray and rap · adopted, the chorus sung. I verify the schema, the scansion, and the rhyme scheme below; the glosses I treat as an ancient interpretation, authoritative as interpretation, not as text.

1.3 Gr, the Grammar

The treatise is a competent-looking synchronic grammar of spoken catoid: a segment inventory bounded by "the cat vocal tract: sonorants and hisses," a closed class of seventeen function words, eighteen open roots, exemplar sentences, a register doctrine, and a translator's note. It states that the reference dialect is our number, 1234, and that root forms are "engine-rolled and reproducible" by the procedure it cites. The procedure is not among the recovered materials; the claim of reproducibility cannot be tested, and I mark engine-derived forms accordingly.

Two internal facts fix Gr's position relative to T. First, Gr remarks that the song texts consist of its closed class "to a degree that surprised the analysts, since the songs were generated before the grammar was written" — Gr's authors knew the songs, were surprised by them, and wrote afterwards. Second, Gr's §9 renders the first two refrain lines of our Text verbatim — and in the same breath calls its own reading "consistent with the corpus glosses (stray verse pool → adopted verse pool)." That parenthetical is the Sheet's rubric vocabulary, stray and adopted, verbatim and in the Sheet's direction. Gr is a downstream witness with access not to the bare text but to the glossed corpus, plus something more — native informants, or the engine, or both; which, the documents do not say.

1.4 Provenance and relative chronology

I will not invent a find-spot. What can be said: T and G belong together (the shared number, the verbatim collation, the shared title word); and Gr depends not only on the song tradition but on the glosses. This last point I state plainly because it reverses the order I first assumed. Gr's §9 calls its refrain reading "consistent with the corpus glosses (stray verse pool → adopted verse pool)" — the Sheet's own rubric labels, in the Sheet's own direction. The Grammar cites the Glossator's work; the gloss tradition therefore precedes the Grammar; and the certain order is T, then G or G's tradition, then Gr. What the documents do not settle is whether the Sheet in our hands is the very copy Gr read or a later re-copy. Its colloquial layer ("zoomies," "if it fits, sits") has no parallel in Gr — but Gr's own apparatus runs on the same archival computing idiom (cited engine paths, "engine-rolled"), so register no longer separates the hands the way one would like it to. Everything else about dating is §12's business, not mine.

2. Editorial conventions

Sigla. T = the Text; G = the companion Sheet; Gr = the Grammar (cited by Gr's own section numbers, e.g. Gr §4.1). Text lines are cited as I (intro), V1.1–V1.8, C.1–C.4, V2.1–V2.8, C′ (the refrain repeat, not re-edited), B.1–B.2, O (outro).

Marks. Plain form = attested in T. Asterisk = reconstructed, or transmitted by Gr only, unattested in T. † = editorial segmentation or provisional analysis of an attested form. Square brackets [so] = editorial supply in translations. "…" = sense uncertain.

Glossing. Interlinear glosses follow the Leipzig Glossing Rules as far as the evidence allows, which is not very far: LGR's morpheme-by-morpheme correspondence cannot be met for song forms that no analysis segments, and I refuse to manufacture segmentations to satisfy a convention. Unanalyzed forms carry a bare † in the gloss line; sense lives in the translation line. Full reduplications are hyphenated in the text line (mi-mi, mia-mia-mi, ss-ss, ka-ka) because reduplication is independently established (§7.5). Vowel length is left as T writes it (maa, mii, sikaa); I do not segment length into morphemes in the text line, though I analyze it in the gloss (INT).

Abbreviations. PROX proximal deictic; MED medial; DIST distal; 1.EMPH emphatic first person / reflexive / possessive 'of me' (one item in catoid, Gr §4.1); CREA creature classifier; FOD food classifier; LOC place classifier; INAN thing classifier; NEG negation–refusal; CMP completive–contentment ('done, and it is good'); CMP.PRC the same with percussive release; IMP demand-imperative; SOFT soft irrealis request; HUM hum, no propositional content; ALERT alert particle; UNEV unattainable evidential; RED reduplication; INT intensive (realized as vowel length).

Contour. Gr writes clause contours with arrows (rising = invitation, falling = demand). Neither T nor G encodes contour in any way. The arrows are Gr's reconstruction of a spoken system, and I print them only when quoting Gr, never as commentary on T.

Scansion. I count syllables by maximal vowel runs: a sequence of vowels, however long, is one nucleus (so miao = 1, maa = 1, miadi = 2). This is not my invention; it is the Sheet's practice, demonstrated in §3.2.

3. The character of the corpus

3.1 Structure, responsion, and the particles

The song is a paired-strophe composition with refrain. The two eight-line strophes answer one another sense for sense, though not slot for slot: the stray strophe's flight from "the cursor" is answered by the adopted strophe's chase of it; "hunger ticks up" by "fed, hunger zeroed"; "cold night sidewalk" by "sunny spot claimed"; "skittish stray" by the three-a.m. circuits; "alone under the window" by "tail up, happy"; and the frame is chiastic: the strophe's first line ("no record, no home") is answered by the second strophe's last ("one putRecord, family"), and its last ("watching the warm lights") by the second's fourth ("sleeps on the keyboard"). One pair is weak ("rain on the fur" with "if it fits, sits") and one is strong but crossed in position ("flees the cursor" with "chases the cursor now"); the weak pair is marked so in the table. The Glossator laboured the labels stray and adopted onto the rubrics; the responsion shows the labels are earned, not imposed.

The line-final particles divide the strophes as cleanly as the glosses do. Every particle token in T:

Negation occurs only in the stray strophe and the outro; demand and completive-contentment occur only in the adopted strophe and the outro. The stray strophe closes on refusal; the adopted strophe closes on demand and satisfaction; the outro, uniquely, holds them all in one line. I know of no shorter statement of the song's plot than that distribution.

3.2 Scansion: the Sheet's counts verified

G gives a syllable count for 26 lines. Counting nuclei as maximal vowel runs (§2), all twenty-six verify, provided one rule is added: the particles the Glossator sets off parenthetically — ss, miao, mrr, mrrp — stand outside the metre. The rule does real work on exactly two lines, the two that end in miao (V2.1, V2.4): the letter count is 11 and 12 nuclei, G says 10 and 11. For ss, mrr, and mrrp the rule is idle, since those forms have no vowel at all — they could not have counted. But the parenthetical notation and the extrametricality coincide exactly, on all six marked lines and nowhere else, and that coincidence is not an accident of my analysis: it is the Glossator telling me, in the only notation available, that these particles are performed at the line, not in it. I take the parenthetical-exclamative mark to be a performance instruction — a percussive or vocal outburst at the bar line — and I have printed it as such in the edition.

Metrical profile: strophes run 10–14 nuclei per line, the refrain 6–8, the bridge 6. The refrain is roughly half the length of the strophe line, denser, and monorhymed (§3.3). The whole is consistent with G's performance labels: recitative (rap) for the strophes, sung for the refrain.

3.3 Rhyme

G announces "scheme ABAB," and the strophes bear it out — with the particle outside the rhyme, as it is outside the metre. V1, odd lines: mikaam, ngaam, mam, wam — A = -am/-aam. Even lines: mimim, mamim, mim, mim — B = -im. V2 is the same: odd lines maam, momaam, mikam, mam (A), even lines mimim, mikim, mim, mim (B).

The B rhyme deserves a sentence of its time: four lines of the first strophe end in the reflexive mim or its echo (mimim, mamim) — the stray strophe rhymes, literally, on "myself." The second strophe keeps the rhyme and changes the sense: mimim now closes the line the Glossator renders "fed, hunger zeroed," and mim closes "…family." The same ending, transformed from solitude to kinship, is the responsion carried out in the rhyme itself. I doubt the composers were unaware of it; I am certain the Glossator was not.

The refrain is monorhymed on -aa: every line ends maa or maa-final (mamaa), and G names the title word maa with the "hook tail -aa." The song is named for its rhyme.

3.4 The m-line

22 of 28 lines begin with m; 114 of 165 tokens; 56 of 91 types; 228 of 649 letters (35 per cent). The letter m is to this text what the drone is to the bagpipe: the ground the melody walks on. The deictic paradigm mi/ma/mo/mim supplies most of it, the mam-family much of the rest (§9.4). I flag the obvious circularity myself: the deictics are frequent because they are deictics, and their m-initial shape is a fact of the language, not of this poem. What is a fact of the poem is that the composers set nearly every line on that anacrusis.

3.5 Authorship

Gr reports, at the close of its §9, that "cats wrote none of it, which is the most catoid thing about the corpus." Stripped of the treatise's tone, the claim is that T is a composed text in the language — the work of a tradition that used catoid as its song Latin — not a transcript of native utterance. The corpus itself cooperates with this reading: the high proportion of closed-class tokens (55 of 165, 33 per cent; §6), the formulaic templates (§9.4), and the strophic mechanics of §3.1 are the signature of a composed liturgy, not of conversation. The consequence for method is stated once and obeyed throughout: T evidences the grammar of the song tradition, and whatever of the spoken tongue the tradition preserved. It is not, by itself, evidence for the spoken tongue. For that we have only Gr, which is why Gr is tested rather than trusted.

4. The sound system

A warning to the reader first, because it is the warning I would want. The only direct phonological evidence is a sung text in a special register that G's own preamble says has consonants and vowels disjoint from the spoken inventory. Every statement about spoken catoid phonology below is therefore transmitted testimony (Gr), checked for internal coherence and typological decency, and no more. Every statement about the song register is attested and counted.

4.1 The song register: attested segments

Consonants, by token count (letters; ng and rr are digraphs, §5.2):

Vowels: i, a, o — with length in the nucleus, aa (maa, mikaam, ngaam, maam, mamaa, momaam, sikaa) and ii (mii, twice), which §4.4 argues is intensive morphology rather than contrast. No oo. Vowel sequences attested: ia (e.g. kimmia, siamo), iao (miao), and oa (moa, mikiamoa) — all rising or crossing through a, all single nuclei by the scansion of §3.2. The letter r occurs only in the digraph rr, which Gr analyzes not as a consonant but as the written mark of purr phonation (§4.5); p occurs once, releasing the purr particle (mrrp), which Gr's own doctrine of the song register ("percussive onomatopoeia," Gr §7) licenses.

Three graphemes deserve their singleton status in the record: d appears in exactly two types (miadi ×2, mamida ×1); ng in exactly one word (ngaam); p in exactly one (mrrp). Singleton attestation proves existence, not system — but G's preamble independently declares d and ng (and l) among the schema's consonants, so the Glossator's grammar and the Text agree, and l is the one declared segment with zero attestation (§12).

4.2 The spoken system, per the treatise

Gr §2.1 describes a spoken inventory of eight consonants and five vowels, "bounded by the cat vocal tract: sonorants and hisses":

with vowels i e a o u, three "meow-family" diphthongs (iau, ia, au) as the only heavy nuclei, and two marginal segments: /k/, "the hunting click," surviving "in exactly one grammatical morpheme" (the classifier ki), and the bilabial trill of the alert particle prbt, which "never enters roots."

The system is typologically unremarkable, which in this business is praise. Eight consonants is small but squarely inside the attested human range; five vowel qualities is the cross-linguistic mode; the vowels sit at sensible dispersion (peripheral triangle with mid front and back fill). Whether a cat vocal tract can own such an inventory I cannot know; what matters here is that the treatise's system is coherent, and — the point I return to below — it neither predicts nor forbids what T actually shows.

4.3 Correspondences: what the song did to the spoken system

The two inventories overlap in m, s, w, and (partially) k. The differences run in both directions.

Lost in the song register: n, singleton r, y, h, sh, and the vowels u and e. The losses are not sampled noise — they are total (zero tokens of each in the 649 letters of the text) — and they explain, mechanically, why four of the seven unattested closed-class items are unattested (§6): *mu needs u, *myu needs y and u, *nya needs y, *hrr needs h. Only *na, *wu, and *prbt are absent for no phonological reason, and of those, *prbt carries the bilabial trill that Gr itself confines to interjection — plausibly excluded from song as a class, not as a sound.

Liberated in the song register: k, from one morpheme in Gr to the second most frequent consonant in T (43 tokens); t, absent from Gr's spoken table entirely, with 9; and d and ng, of which Gr says nothing anywhere — its §7 license names k t p for song and is silent on d, ng, and l, so p, being licensed, is not liberated, only spent once. Here the treatise is not wrong but thin: it describes the percussives as ornamental "musical onomatopoeia, the way scat singing uses sounds the language doesn't have," and the Text contradicts the decoration's scale — in the song register the percussives are structural, carried by the rhyme (A = -am on percussive-heavy words) and by the classifiers (ki, 6 tokens).

On u: the only place the question becomes concrete is the demand particle. Gr's diphthong doctrine writes the meow as iau, and its intensive exemplars spell miiau; the same treatise's own function-word table spells the particle miao, and T spells it miao (3 tokens). So Gr itself wavers, T does not, and T has no u at all. Whether spoken u merged with o in the sung register, or the song tradition merely avoided u-bearing words, one redaction cannot say (§12). I mark nothing with an asterisk that T does not force me to.

4.4 Length

Vowel length is attested only as aa (8 tokens of maa alone) and ii (mii, 2). Gr §2.2 analyzes length not as contrast but as intensive morphology("miao 'summon' → miiau 'summon, urgently, with the full chest'"), and the Text supports the analysis in the one place it can be tested: ma and maaboth occur free (4 and 8 tokens), in complementary work — the short form weaves through the strophes as a medial deictic, the long form is the refrain's final word, the title word, the hook. The far place, held long. The Grammar's phonology, here, reads the poem correctly.

4.5 Phonation registers

Gr §2.3 makes phonation grammatical: purr (written rr), growl (h-proclitic), chatter (the unattainable evidential hrr), modal (unmarked). T attests the purr marking exactly, in the particles mrr and mrrp, and nothing else: no growl, no chatter, no rr inside any root. The doctrine is coherent with the Text and untested by it — one register mark in two function words. The outro's final mrr, which the Glossator renders "purr fading out," is as close as this corpus comes to primary phonetic evidence, and it is a performance note, not a phonology.

4.6 Contour

Gr writes clause contours with rising and falling arrows and says of them "it is not optional in speech." Perhaps. The archive's romanization is contour-blind; nothing in T or G encodes pitch. The entire apparatus — including the attractive reading of the refrain as rising request against falling settlement — is untestable from the corpus and stands or falls with Gr's credibility (§13).

5. Phonotactics and orthography

5.1 Phonotactics of the song register

No vowel-initial words. Of 165 tokens, zero begin with a vowel. Gr claims the same constraint for the spoken language ("no vowel-initial words," Gr §3), and the song register honours it without exception.

Onsets. Single consonants m, k, t, s, w, n(g). Complex onsets attested: mw-, kw-, tw- (mw-: 8 types, 11 tokens; kw-: kwi ×2; tw-: sitwa, twomamo). Every attested cluster takes the glide w as its second member; every cluster therefore rises in sonority toward the nucleus, exactly the law Gr states for the spoken language (Gr §3: onset clusters "always rising in sonority") — with this difference: Gr licenses nasal+glide and r+glide; the song register, having lost r and y, realizes the same law as anything+glide, stops included (kw-, tw-). The law holds; the inventory under it shifts.

Codas and word-ends. The declared final set in G is [d k l m ng s t]; the attested final set is: vowels (133 tokens: i 64, a 44, o 25), m (25), s (4, all ss), r (2, in mrr), p (1, in mrrp). Strip the particles out and the picture is clean: song words end in a vowel or in -m. Final d, k, l, ng, t are declared but never once exemplified — whether the Glossator's catalogue describes the scribal system and this copy is simply short, or the catalogue overstates, cannot be decided. Under the onset law above, singleton internal codas are forced in five words — kimta, kamso, komsiki, mimimkam, mimimki — all m; the mm-words (timmam, kommwiki, mommwi) I take as ambisyllabic geminates, not clusters.

Syllable shape. (C)(C)V(V)(C) — glide-second onsets, heavy nuclei only in the a-crossing diphthongs and long vowels, codas m. The Grammar's spoken template is the same with n, r codas and no stops in onset clusters. Gr's summary of its own phonotactics — "every segment is sonorant or hiss, so every syllable is hummable and singable" — is as good a one-line description of the song register as I can write, minus the percussives Gr forgot to except.

5.2 Orthography

The romanization is engineered for a plain-letter channel: twenty-six lowercase ASCII graphs, no diacritics, length written as doubling (aa, ii), the palatal fricative as sh (Gr's spelling, unattested in T). Two digraphs occur in T: ng and rr. Both are uniquely decodable in this registerin the strong sense the reading machine needs: plain n and plain g never occur apart from ngaam (one and one tokens, exactly the one word), and plain r never occurs apart from rr. A longest-match tokenizer over T cannot mis-segment. In Gr's spoken system the corresponding statement has to be checked rather than assumed — sh against s+h is the dangerous case, since Gr's inventory has both s and h — and Gr asserts it ("sh the only consonant digraph") without demonstrating it. I flag, and cannot settle, the point.

G's preamble, read as an orthographic statement, adds the edge catalogues (§5.1) and "scheme ABAB" (§3.3). It is a compact, checkable, and almost entirely accurate piece of grammatology — all but l, which it declares and the Text never spends.

6. The closed class

Gr §4.1 reports a closed class of seventeen function words and remarks that the songs consist of them to a surprising degree. The Text confirms the remark and most of the list:

That is 10 of 17 attested. The ten account for 46 of 165 tokens. Two further attested forms are derived from items in the table and counted under them: maa (ma + INT, 8 tokens — I, the refrain finals C.1, C.3, C.4 and their repeat, and O) and mrrp (mrr with percussive release, V2.5). With the derived forms the closed class carries 55 of 165 tokens (33 per cent); crediting the reduplications mimi and mimim to their bases raises it to 61 (37 per cent).

The skeptic's question — how do you know the class is closed from one song? — has a short answer: I do not. Closure is Gr's claim about the language; the Text can only show that a small set of short forms does disproportionate work, which it does, Zipfianly: items of the closed class in any form hold the five most frequent types outright (mi, maa, ki, mo, mim) and eight of the ten most frequent. What the Text adds over Gr is the distributional fact of §3.1 — the particles are not merely frequent, they are deployed (ss to the stray strophe, mrr/mrrp/miao to the adopted) — and the absence pattern of §4.3: of the seven unattested items, four are missing for exactly the phonological reason the song register predicts (*mu, *myu, *nya, *hrr contain u, y, h), and only *na, *wu, *prbt are absent without cause. The closed class, in other words, behaves like a real paradigm under a real register filter. That is the most a closed class can be asked to do in a corpus this size.

One member needs a comment. Gr's table gives myu as "soft request… the mew," and its §6 builds on it a request scale myu↑ < miao↓ < miiau↓ < miao miao. The scale's top two rungs are attested in T (miao ×3; reduplication everywhere). The bottom rung is not. The Treatise translates myu as "please" nowhere — its translator's note explicitly warns against importing deference ("soft the way a kitten is soft — effective, not deferential") — and I pass that warning on, since English glosses of myu as "please" were already current in the secondary literature before this edition.

7. Morphosyntax from the texts

7.1 The particle closes the line

Gr §5: particles "close the predicate"; the utterance template is (TOPIC) (ss) PREDICATE (CLASSIFIER) (ASPECT/MOOD) (CONTOUR). The Text confirms the closure claim in the register it can: of the 16 strophe lines, 6 end in a particle (V1: ss ×2; V2: miao ×2, mrrp, mrr) and the remaining 10 close in nominal material — seven of them on the reflexive rhyme -im(§3.3), the deictic and reflexive forms the Grammar calls topical; the other three on -am — the root mam, wam, and momaam — forms its topic taxonomy (deictics, classifier phrases, *mu) does not cover. The refrain lines all close in maa, the intensive medial deictic. Nothing in T contradicts the template, and the closure position is where the rhyme and the metre both put their weight. In the sung register, grammar, rhyme, and metre are the same fact observed three ways.

7.2 A sentence made only of grammar

The outro is the corpus' grammar specimen, and I give it here because every word of it is closed-class and fully analyzable — the longest such line in T (the opening mmm maa is the other):

"No — no. Come. Far place, held long: it is good."

Reduplicated negation (Gr: particle reduplicated = strengthened), the demand particle, the intensive deictic, the completive. Four morphological processes and four function forms, one complete utterance, zero lexical roots. Gr §6 says "silence is a well-formed utterance"; the outro shows the composers could write a well-formed utterance out of nothing but the grammar. I know of no better evidence that in catoid the function words are the language and the lexicon is upholstery.

7.3 Topic-initial deixis

22 of 28 lines begin with m; the six exceptions are komwo, ka, kwi, ki (k-initial) and somima, ss (s-initial) — and two of those six are still closed-class forms (ki the classifier, ss the negation). Eight lines open with a bare closed-class or closed-reduplicated form (mmm, ma, mim, mimi, ki, mi, mo, ss), and the refrain's two miamiami openings reduplicate the open root mia. Gr's topic-first template, with deictics as the canonical topic, is again consistent — and again I will not stretch a sung register's formulaic openings into a syntax of the spoken tongue. What the Text shows is a poem that begins where the Grammar says an utterance begins.

7.4 The addressee who is never conceded

mu "you" — the one pronominal form that would mark a non-speaker — occurs zero times in T. The entire song is addressed: "adopt me," "feed me," "stay, stay." Gr §1 remarks that in catoid the unmarked subject is always "I," the unmarked possessor always the speaker, and that marking mu on anything "is a concession, made grudgingly." The Text takes the doctrine to its limit: twenty-eight lines of address to an addressee the grammar never once acknowledges by name. Consistent, and — as evidence — merely consistent; but it is the kind of consistency that makes one trust a treatise on the points one cannot check.

7.5 Reduplication

Attested: mimi (4), mimim (2), mia-mia-mi (2), ka-ka, ss-ss, and the echo-chains of the lexicon (§9.4). Gr §4.3 gives three readings by category — noun: plural/mass; verb: habitual; particle: strengthened — and the Text supports the third fully (ss-ss against ss in the same song's outro) and the general force of "more of the same" everywhere. The refrain's second line is the showcase: mia-mia-mi mia mi — a tripled root, then the bare root, then the bare deictic, in one breath: up, up, up — high — here. The scale Gr describes between miao and reduplicated miao miao is exhibited here for mia inside a single line.

7.6 Classifiers at large

ki, the creature classifier, stands free 6 times, in both strophes, in line-initial (V2.5), line-medial (V1.6, V2.4) and immediately pre-nominal (V1.1 ma ki mikaam) positions — free enough that Gr's "classifier phrase" is credible, not free enough to fix its syntax. si, the food classifier, appears once, free (V2.8 mi si wima — "here [is] the food… set down," on which see the commentary ad loc.). The other two classifiers (*na, *wu) are unattested. One usable syntactic fact: ki never follows its noun in T; Gr's exemplars (mo ki hrr) put it after the deictic, before the predication. The order deictic–classifier–rest, cleanly attested twice (V1.1 ma ki mikaam; V2.3 mi-mi ki timo), is the one piece of noun-phrase syntax I would defend from this corpus.

7.7 Possession: one candidate, honestly labeled

Gr §4.4: possession is juxtaposition, POSSESSOR–POSSESSED, unmarked possessor = speaker. The only Text candidate is kimmia (C.1), which Gr's own §9 renders "creature-of-mine." The rendering implies a segmentation ki + mim (+ia?) that Gr nowhere prints; the cut is mine, and I mark it †(ed., after Gr). The refrain's most important word is thus, on the treatise's reading, a possessive construction: the classifier plus the reflexive, and I note that the reading — adoption framed as possession of the creature by the speaker of the possessive — inverts the Glossator's "adopt me" in a way the commentary (§10, C) takes up. It is a beautiful analysis. It is also unprovable from one form, and if the reader takes nothing else from this section, let it be that I know the difference.

7.8 Negation placement

Gr §5 says negation "floats early." In T, negation is line-final twice and nothing else. Either the sung register fixes the particle in the closure slot for rhyme (where it is extrametrical, §3.2, and outside the rhyme, §3.3 — final in the line, first in the ear), or Gr's statement is wrong for the tradition that composed T. I lean to the first; §12 records the doubt.

7.9 What the Text cannot show

Absence of passives, tense, number agreement, honorifics — all claims in Gr's typological profile — is not evidence in a 28-line hymn, and I refuse to enter any of them in the scorecard. The Grammar argues some of its sociology from what I can only call theology ("there is no passive, because nothing happens to a cat that a cat has not, at some level, permitted," Gr §5). The sentiment is charmingly in character and distributionally empty. The profile stands or falls with Gr's informants, not with T.

8. Registers

Gr §7 lists four: spoken catoid; purr-talk (flattened contours, close range, "glosses as assent to everything"); chatter (hrr-clusters, "grammatically an evidential noun-phrase, pragmatically a whole sentence about birds"); and song, "the register of the attested corpus," licensing percussives as "musical onomatopoeia" plus long rhyme units and held vowels.

The Text attests three registers, none of them named quite as Gr names them:

Spoken catoid, purr-talk, and chatter are unattested by any document. The concentration is worth stating plainly: every connected catoid text we possess is sung or recited. The prose of the language, if it ever had prose, did not survive — or was never written; which of the two, §12.10 declines to decide. The situation is the familiar one of song Latin in a vernacular-less archive, and this edition is written under its discipline: the register described in Gr §7 is not a stylistic appendix to the grammar. It is the only grammar we can check.

9. Lexicon

9.1 The closed class

Tabulated with attestations in §6.

9.2 The Grammar's open roots against the Text

Gr §8 lists eighteen roots for the reference dialect, engine-rolled. Against T:

Two of eighteen: mam free (4 tokens); mia free (×2) and in the refrain's compound. The obvious explanation is register-filtering again — every one of the sixteen missing roots contains at least one segment the song register lacks (e, u, n, r, y, or sh) or describes the spoken world rather than the song's — and the obvious explanation is also not demonstrable. What is demonstrable is that the Grammar's lexicon and the Text's lexicon are nearly disjoint, and any lexicography of catoid that cites Gr's roots as "attested" without qualification is misreporting the record.

9.3 Index verborum of the Text

Every type not in §6 or §9.2 — 77 forms. Counts, loci, and such notes as the evidence carries. The senses are line-senses, not word-senses: G glosses by the line, and a line gloss cannot be distributed over the words of the line; I have not tried. Analyses marked † are editorial; "Gr" marks a reading I take from the treatise.

A. m-deictic and reflexive extensions (closed-class nearest; miamiami reduplicates the open root mia, §7.3):

B. the mam-family and hunt forms (contains Gr's root mam; see the caution below):

Caution: given a medial deictic ma and a root mam, any mam sequence in running text is ambiguous — mam + i, ma + mi, ma + ma… — and the reduplicative aesthetic compounds the ambiguity deliberately. I segment none of these. The family reads as a texture (the hunt pervading the song's diction) rather than as a derivation, and the texture is exactly what the paired strophes want: every line of the stray strophe and most of the adopted one is built from hunt-shaped material.

C. the mw-series (glide-onset family, §5.1):

D. remaining types (single tokens unless noted), with loci:

9.4 Word-formation in the Text

Three productive-looking patterns, honestly graded. (i) Full reduplication — established, §7.5. (ii) Echo-pairs — miki ~ moko (V1.7), timmam ~ timo ~ tommi, kamso ~ kammamo, kommwiki ~ kommoma ~ komsiki ~ komwo: rhyming variants sharing an onset template, whose semantics (if any) the corpus cannot reach; I call them melodic formulae rather than derivations. (iii) The m-template — 56 of 91 types begin with m and most fill a (m)V(m)V frame; this is register aesthetics, not morphology. Gr's own compounding doctrine (head-final: mau-na sun-place, ri-shua warm-hand, Gr §4.5) is unattested in T and may well be true of the spoken tongue; the song builds its long words by echo and rhyme instead.

10. The Text: transcription, interlinear edition, commentary

10.1 Diplomatic transcription

Reproduced exactly, rubrics included:

In the interlinear edition, reduplications are hyphenated; the parenthetical particles are printed with G's exclamative mark restored as (—) after the line; G's gloss is quoted first, my rendering after a slash.

10.2 I. Opening

G: "greeting purr" / ed. A hum. Then, held long: the far place.

The hum is the only token of mmm in T, and Gr §6 says of it that it carries "no propositional content" — true of the lexeme, and yet the composers put it first, before the title word. The song opens on pure voice and closes on purr; between those two non-words, everything else happens.

10.3 V1. The stray strophe (recitative)

G: "no record, no home" / ed. There, here — a stray's nowhere: no wall that is mine, no ledger with my name in it. No.

G: "flees the cursor" / ed. It runs from the moving point; the small self runs.

G: "cold night sidewalk" / ed. Cold outside, hard ground; here in the chest-wall, hunger. No. — The sole ŋ of the language lands here, in the coldest word in the text; a philologist may enjoy that, and should not build on it.

G: "hunger ticks up" / ed. The outside-place: the belly's count climbs, hunt on hunt, nothing fills it.

G: "skittish stray" / ed. Here-there the stray starts at every sound, hunting what is not there.

G: "alone under the window" / ed. Beat, beat — yonder, here, here: one small creature under the lit wall. Myself. — Note the deixis: distal, then proximal doubled. The whole strophe's ache is in that sequence.

G: "rain on the fur" / ed. Dim here, dim there: rain in the fur to the skin; here to yonder, water.

G: "watching the warm lights" / ed. Far houses, far kin, warm windows — watched from outside by none but me.

Commentary. The strophe's grammar is its grief: four of eight lines close on the reflexive rhyme (§3.3), two close on refusal, and the deictics run medially and distally (ma, mo) with the proximal doubled only in the alone-line. The Glossator's "no record, no home" I take up in the commentary to V2.8, where the record returns.

10.4 C. The refrain (sung)

G: "adopt me" / ed. Here, here and here — this creature of mine, toward the far one made near: home, home. — Gr renders: "here, here-and-here — creature-of-mine, toward the far warm one: home, home."

G: "feed me" / ed. Up, up, up to the high here: the far hunt become the full belly — home.

G: "warm home" / ed. My hunt turned yonder-warm, food and hearth: the far place, held long.

G: "stay, stay" / ed. High-here, yonder-warm, food kept long in the far place: stay, stay.

Commentary. The refrain has a lexicon of its own — miadi, kimmia, moa, siamo, miaki, sikaa, mamaa, miamiami occur nowhere else in T — and it dislikes percussives (the strophes' k's and t's nearly vanish; maa carries the melody instead). On the two witness-readings: the Glossator gives the refrain as four petitions and settlements ("adopt me / feed me / warm home / stay, stay"); the treatise reads the first line as a possessive — the creature of mine — and the whole as motion, "the far place become the near place" (Gr §9). The readings are not rivals. G translates what the line does to the addressee; Gr translates what it says about the speaker; and the song, characteristically, declines to mark the addressee at all (§7.4). C′.1–4 repeat verbatim.

10.5 V2. The adopted strophe (recitative)

G: "chases the cursor now" / ed. Myself, my own small one: now it is I who chase the moving light. Come!

G: "fed, hunger zeroed" / ed. The far wall's warmth is mine; the belly's count has run down to nothing, nothing.

G: "sunny spot claimed" / ed. Here, here the creature sleeps: the warm place is taken and held.

G: "sleeps on the keyboard" / ed. The small one pours itself out over the warm ledge of the writing-hands. Come!

G: "three a.m. zoomies" / ed. The creature takes the dark house at a run, round and round: done, and well — and a snap at the end of it.

G: "if it fits, sits" / ed. Hunt of mine, small vessel of mine, whatever holds: if it fits, I sit — mine, myself.

G: "tail up, happy" / ed. The small self runs upright, tail aloft, round the warm house: the hunt is done, and it is good.

G: "one putRecord, family" / ed. Here the food is set down and entered — one entry in the ledger — the far ones mine, myself theirs: kin.

Commentary. The strophe answers V1 sense for sense (§3.1) and rhymes on the same endings (§3.3) — with the reflexive rhyme now closing "fed" and "family" instead of "alone." The particles have changed sides entirely: no negation anywhere; two demands, two completives. On the Glossator's "putRecord": the computing register again, matching V1.1's "no record." The two glosses, read together, give the song's frame in the Glossator's own idiom — the stray is in no ledger; the adopted is entered in one, and the ledger is called family — and whatever we make of the Glossator's English, the catoid pair (V1.1 …ss / V2.8 …mim) supports the antithesis: refusal against reflexive-kinship. This is the one place I would defend a reading built substantially on G: it is confirmed by the particle distribution, the responsion, and the rhyme.

10.6 B. The bridge

G: "deep sleep" / ed. The house goes soft and slow: the long rest.

G: "dreaming of birds" / ed. Yonder, the air-creatures cross through the sleep: the long rest.

Commentary. The smallest strophe rhymes on mii — on rest itself. The distal deictic opens the second line, and the dreamed-of creatures of the air are the treatise's birds, the paradigm case of hrr, the unattainable (Gr §6: "window-birds, moths on the far side of glass, the Outside generally"). The particle does not occur in T at all. Inside sleep, it seems, the barrier is gone — but I flag the sentence just written as the commentary it is: nothing in the catoid marks it, only the absence of a mark.

10.7 O. Close

Treated as the grammar specimen in §7.2:

G: "purr fading out" / ed. No — no. Come: far place, held long. It is good.

The last event of the song is a purr that outlasts the words. Gr §6 again: "silence is a well-formed utterance, usually affirmative." The Text ends by demonstrating the doctrine.

11. A continuous rendering

Interpretive throughout; G's line-senses respected, my renderings responsible for everything between them. The refrain is printed once where first sung and marked at its repeat.

Opening. A hum. Then, held long: the far place.

Strophe one — outside.There, here — a stray's nowhere: no wall that is mine, no ledger with my name in it. No. It runs from the moving point; the small self runs. Cold outside, hard ground; hunger in the chest-wall. No. The outside-place: the belly's count climbs, hunt on hunt, nothing fills it. Here-there the stray starts at every sound, hunting what is not there. Beat, beat — yonder, here, here: one small creature under the lit wall. Myself. Dim here, dim there: rain in the fur to the skin; here to yonder, water. Far houses, far kin, warm windows — watched from outside by none but me.

Refrain — sung.Here, here and here — this creature of mine, toward the far one made near: home, home. Up, up, up to the high here: the far hunt become the full belly — home. My hunt turned yonder-warm, food and hearth: the far place, held long. High-here, yonder-warm, food kept long in the far place: stay, stay.

Strophe two — inside.Myself, my own small one: now it is I who chase the moving light. Come! The far wall's warmth is mine; the belly's count has run down to nothing, nothing. Here, here the creature sleeps: the warm place is taken and held. The small one pours itself out over the warm ledge of the writing-hands. Come! The creature takes the dark house at a run, round and round: done, and well — and a snap at the end of it. Hunt of mine, small vessel of mine, whatever holds: if it fits, I sit — mine, myself. The small self runs upright, tail aloft, round the warm house: the hunt is done, and it is good. Here the food is set down and entered — one entry in the ledger — the far ones mine, myself theirs: kin.

Refrain.

Middle. The house goes soft and slow: the long rest. Yonder, the air-creatures cross through the sleep: the long rest.

Close. No — no. Come: far place, held long. It is good. (a purr, fading)

12. What the evidence cannot decide

The honest list, with what would decide each item.

13. The Grammar reassessed

My verdict, for what a single witness is worth: the Grammar is a good faithful downstream witness to the closed class, the derivational morphology, and the phonotactic laws — everything that shows through the song register intact. It is thin on the song register itself, silent where it should have described (d, ng, l, the three-vowel reduction), and its lexicon belongs to a world the Text never enters. Where Gr and the Text disagree (negation placement), the disagreement is small and register-borne; where Gr and the Glossator disagree (the refrain's reading), they are answering different questions. I would keep the treatise on the shelf beside this edition, consulted for the spoken tongue, corrected by the Text for the sung.

The song's own thesis, in the treatise's words, remains the best thing ever written about it: the far place has become the near place. maa — held long — says so at the end of every refrain; mim, rhymed from solitude into kinship, says so at the end of both strophes; and the last audible event of the document is a purr. Of the speakers of meu, "the talking," we know nothing else. This is their entire surviving literature, and it is enough.

Edited from T (catoid_1234, mmx series) and G (catoid_1234 sheet), with Gr (catalogue: catoid — a language sketch) cited by section. The original sketch remains untouched in the archive. Forms asterisked above are transmitted by Gr only; forms daggered are this edition's analyses.