The Grammar That Checked Itself

•

Self-published language textbooks are a genre with a deserved reputation: enthusiasm running ahead of evidence, paradigms remembered rather than verified, the air of a classroom handout that escaped. Kaqchikel from the Ground Up arrives with a different constitution. It is an English-language course in Kaqchikel Maya, twenty-six chapters and a set of appendices, delivered as a single offline web app with hover glosses, generated conjugation tables, and spaced-repetition flashcards. Its front matter names its printed authorities, Brown, Maxwell and Little's ¿La ütz awäch? and Salazar Tetzagüic's bilingual anthology of Kaqchikel literature, and behind the book sits something the front matter only hints at: a working corpus of more than a dozen sources, from the sixteenth-century Memorial de Sololá in Maxwell and Hill's interlinear edition to schoolbooks, dialect surveys, dictionaries, and a file of present-day Kaqchikel scraped from public Facebook posts. The author's stated rule is that no Kaqchikel form goes on the page unless the sources attest it or a stated rule derives it, and the project even ships a script whose one job is to flag words the corpus cannot vouch for.

A book like this invites the audit, and this review accepts the invitation. Eight independent passes were made through the text, one per block of chapters, and every checkable claim was taken back to the sources: each paradigm cell against Maxwell's reference grammar, each example sentence against the corpus, each dialect note against Patal Majzul's Variación Dialectal en Kaqchikel, each quoted line of literature against the anthology it cites. Nothing is called an error below unless a source can be quoted against it, with file and line. That discipline matters, because the first and most important finding is the one a hostile reviewer least wants to report.

• •

The engine room holds. The four-way aspect system, both sets of person markers, the rule that drops the r of ru- after an aspect prefix, the bridging vowel after xt- and n-, the harmonic echo vowel of the imperative, the passive and antipassive machinery, the entire positional system with its causatives and statives: all of it was checked cell by cell against Maxwell's grammar summary and against the professional morphological parses in the Kaqchikel Chronicles, and it agrees. Of the hundreds of inflected forms the audits took apart morpheme by morpheme, only a bare handful of outright conjugation errors survived scrutiny across the whole book, and each is correctable in a minute.

Better still, the book's most suspicious-looking claims kept turning out to be true. Ajsamaj glossed "daykeeper" rather than "worker" looked like a howler; the glossary confirms it. The bare intensifier xe "only", the archaic day-name glosses of the Cholq'ij sidebar, the shape-adjective plurals like tz'uyutz'äq, the claim that a polite refusal of thanks is literally "don't say it like that", the obscure anthology author "Marcelino Tavila" who smelled strongly of a typo: attested, attested, attested, attested, attested, usually verbatim. The number appendix, including the alarming high vigesimals up to juchuy 8,000, survives hostile checking cell by cell. When a reviewer sets out to catch a book inventing and instead keeps catching himself doubting things that are true, that is the strongest available evidence that the author worked from the corpus and not from memory. It is worth saying plainly: by the standards of this genre, the linguistic core of this book is unusually trustworthy.

Which is what makes the errors that do exist so interesting. They are not random. They fall into four families, and each family tells the second edition something different.

• • •

Begin with the plain mistakes, because some of them are lovely. Chapter 1, the chapter that teaches exactness, that warns the learner in bold that "the apostrophe is not optional", offers the evening greeting xoq'a without its final glottal stop, and then helpfully supplies a "fuller form", xokoq'a. The source's glossary knows that word. It means rainbow. The attested fuller greeting is xok aq'a'; the book has crossed it with xokoq'a' and produced a learner who bids the family goodnight by announcing meteorological phenomena. In the same chapter the "widespread variant" of rin "I" is given as yün, a form that occurs in no source and appears to be born of a single OCR misreading of the glossary's yïn; the ghost pronoun then walks again in Chapter 2. An exercise advertising minimal pairs offers pairs that differ in two and three segments, and its answer key asserts, of qab'i' and kib'i', that "the vowel is the same". It is not.

The same family of slips recurs at intervals throughout. Chapter 9 prints In nim raqän "I am tall" with a third-person possessor, against its own correct table forty lines later. Chapter 15 twice gives xik'o "I passed through" where the chapter's own paradigm requires xinik'o; as printed it means someone else did the travelling. Chapter 14's exercises conjure nuloq'o', an impossible hybrid that welds the imperative's echo vowel onto an incompletive verb, three times. Chapter 20 glosses Majun b'ey nutz'eton as "He has never seen the lake" when the prefix says I; the book glosses the identical form correctly in its own answer key. Chapter 25 tells you your cousin is your nuch'itichaq', "little younger sibling", collapsing precisely the same-sex, relative-age distinctions the chapter has just insisted are "one of the clearest signs of real competence"; the source lists the full cousin paradigm. And the market chapters gloss xerka as "jacket" when the source describes a man's woolen breechcloth, the rodillera worn over the trousers: a culturally salient garment mislabeled in the chapters that trade hardest on cultural texture.

Two errors in this family rise above the anecdotal, because they live in the reference apparatus, the part of the book "meant to be kept open while you read". Appendix A.4 prints potential-aspect forms, xtqatz'ët, xtkiya', that Chapter 13 explicitly stars as ungrammatical, and the corpus sides with Chapter 13; a learner who trusts the master table will produce forms the book's own exercises grade as wrong. And the dialect cheat sheet inverts the geography of ruma and roma "because of", assigning the majority form, used in Tecpán itself, the book's home dialect, to "older texts and K'iche'-influenced speech". The same sheet sends travellers to say ch'ima' for "chayote" without mentioning what the dialect survey states outright: that in most variants, to adults, the word means "vagina". A cheat sheet exists to prevent exactly that dinner-table incident.

• • • •

The second family is the one the book would least like discussed, because it concerns its own charter. The project's writing brief is admirably strict: linguistic facts may be taken from the sources, but the sources' illustrative sentences, dialogues and wording may not. Against that rule, the audits found a steady undertow of reuse. "Xtzaq qa pa b'ey, He fell down on the road", sentence and gloss, is the glossary's. "Where did you leave your sweater?", sweater and all, is Maxwell's. The bargaining line "Jotöl b'a rajil. Taqasaj jub'a'" is the source dialogue's. Across Chapters 15 to 22 the audits logged more than twenty verbatim or one-word-swapped lifts, and in Chapter 14 the practice caught the disease that proves the diagnosis: the book prints takolo' kan ri atzyaq with a missing ejective, character for character the OCR-mangled form of that sentence in the scanned source, while the book's own tables spell the verb correctly. You do not inherit another book's scanner error unless you copied the sentence.

One instance crosses from habit into something worse. Chapter 23 reproduces the source's conditional "Wi najin ta jäb' chwa'q, man xkojb'e ta", but quietly rewrites the source's translation, "if it rains tomorrow", as "if it should rain tomorrow", in order to support the book's claim that the particle ta adds tentativeness, a nuance the source nowhere teaches and its own parallel examples contradict. That is the single most serious editorial fact this review has to report: not a wrong form but doctored evidence, the one thing a book built on the promise of verification cannot afford. It is, as far as eight audits could determine, unique; it should not survive to a second printing, and the "tentative" doctrine should go with it or find real support.

━━━━━

The third family is subtler: the book has an honesty convention and keeps forgetting to use it. The brief requires a visible flag on any form the author could not verify. Chapter 17 shows the system working beautifully, hedging exactly where the corpus is thin. Elsewhere the discipline slips, and regular-but-unattested extrapolations are printed in the same confident type as corpus fact. Chapter 10's table of score numerals runs serenely from kawinäq 40 to lajwinäq 200 as though all were equal citizens, when the corpora attest only two of the column; the answer key then requires the unattested forms as model answers, while the fully attested k'al series sits one row away. The appendix prints fused spellings of relational nouns, chinuwäch, chiqaxe', that every source writes as two words. Chapter 16 builds plausible frames for "I feel cold" and "he has a fever" and then enters them in the vocabulary as plain fact. Chapter 21 offers a negative tag question no corpus contains. Chapter 22 invents a verb, nqachaq' qi' "we are siblings", by zero-converting a noun in a way Kaqchikel does not license. Meanwhile the flags that do exist have decayed into noise: of the twenty question-marked forms surviving in the book, most are attested in the very corpora on the author's disk, one marks a cross-reference that is correct, and the forms that genuinely deserve flags have none. The apparatus of doubt needs the same maintenance as the paradigm tables.

•
━━━━━

The fourth family is disagreement between the book and itself, and it matters more than it sounds, because in a language course the model dialogues are the learner's primary evidence. Chapter 4 stakes its flagship rule on the starred form *Rin tijonel, then opens with a dialogue in which the cast says Rin aj Tz'olöj Ya', the same construction minus the required marker, four times, without comment. Chapter 6 states the wi fronting rule categorically and then breaks it in its own reading; the corpus shows the rule as stated is too strong, so here the dialogues are right and the grammar box is wrong, which is arguably worse. The spelling of common words drifts silently between chapters, k'ayib'äl and k'ayb'äl, wakamin and wakami, matyox and matyöx, manq'uq' and manq'u'q, in a book that elsewhere flags every variant it touches. The family saga even loses track of its own genealogy: the grandmother who does not exist in Chapter 5 receives a visit in Chapter 14, and a newborn in Chapter 25 is given the name of the book's adult heroine. None of this is fatal. All of it is exactly what a systematic consistency pass, of the kind the author has already automated for attestation, would catch.

• •
━━━━━

What of the book as a course, rather than as a database of forms? Here the news is mostly, and sometimes exceptionally, good. The methods chapter promises retrieval warm-ups, generation-before-explanation boxes, faded worked examples, recycling exercises and answer keys that explain why, and unlike most books that gesture at the learning-science literature, this one actually implements all of it, chapter after chapter. Several set pieces deserve to be named: the "danger zone" drill on the treacherous xki- prefix cluster, which isolates a real segmentation ambiguity no competing textbook flags; the change-versus-state treatment of positionals; the recipient trap of yinaya'; the seven-move decoder of Chapter 26, which turns three hundred pages of grammar into a procedure for reading a colonial text, and whose every worked row checks out. The framing of relational nouns as ordinary possessed nouns, so that one rule from Chapter 5 unlocks the whole system, is the kind of structural insight that justifies the book's subtitle.

The failures of the teaching are failures of appetite. Chapter 24 carries roughly two hundred vocabulary entries against the brief's own ceiling of sixty; Chapter 19 imports a dozen context-free shape adjectives that never recur; Chapter 11 asks fifty-five verbs of a learner in one sitting. Chapter 25 silently abandons the whole pedagogical apparatus, no warm-up, no recycling, no why-notes, as though a different, more tired author wrote it. Chapter 3 demands morphological parsing four chapters before the morphology is taught. And several chapters assert statistics, "well over half the verbs you meet are transitive", that nothing in the corpus is cited to support, a small vice in most books but a live contradiction in this one, whose entire personality is the receipt.

The second edition, then, has a clear docket. Fix the catalogued errors; they are few, findable, and listed below. Run the consistency pass and pick citation forms once. Restore the honesty flags, and mark extrapolation as extrapolation, which costs one typographical convention and buys back the book's whole epistemic warrant. Rewrite or attribute the borrowed sentences; the corpus is large enough now that original examples are cheap. Renovate Chapter 25 to the standard of its siblings, and split the market chapter's vocabulary into a taught core and a browsable pantry. None of this is structural. The structure is sound.

A closing word about what the book is for. Kaqchikel has perhaps half a million speakers and a written record older than most European vernacular grammars, and the learner materials available in English remain scarce, uneven, and mostly out of print. A self-study course that is free, checkable, and honest about its own limits is a real contribution to that small shelf, and the habit it most needs to deepen, saying only what the evidence supports and saying so out loud, is the habit it was founded on. The first edition promised a book built from the ground up. The ground, it turns out, is solid. The second edition's work is in the mortar.