Maxwell vs. Maxwell

•

The first review ended with a docket: a textbook of Kaqchikel Maya, audited line by line against its own corpora, handed a catalogue of errata for its second edition. The docket has since been worked through. The rainbow greeting is a greeting again, the ghost pronoun has been exorcised, the potential-aspect tables agree with the chapters, and the honesty flags were re-audited one by one. That would ordinarily be the end of the story, a review and its remedy, and this column would have moved on to some other book.

Then the evidence changed. The textbook's primary source is Brown, Maxwell and Little's ¿La ütz awäch?, the standard English-language introduction to Kaqchikel; and its authors, it turns out, wrote a sequel that never reached a press. Kik'aslemal ri Kaqchikela' / Kaqchikel Maya Life, by Little, Maxwell, Cuma Chávez and Cuma Chávez, is a 384-page advanced course organized by cultural domain, milpa and market, kites and clay, medicine and politics, and it exists as a born-digital draft that one of the authors handed over directly, with permission to use it. Tucked behind its eight chapters sits the real treasure: a brand-new reference grammar by Judith Maxwell, eleven domains from morphophonology to pragmatics, with roughly 1,300 numbered interlinear examples of professionally parsed modern Kaqchikel. Twenty years separate the two books. The same linguists wrote both.

Which means the textbook, whose whole personality is the receipt, suddenly had something no receipt can buy: a second witness.

• •

It is worth pausing on why this matters more than another gigabyte of corpus would. A bigger corpus multiplies attestation, but it cannot multiply authority: when your one reference grammar errs, or simplifies, or quietly changes its mind, no amount of raw text will tell you which of those has happened. What you want is a second description of the same language, at the same register of care, by people whose judgment you already trust, and ideally by the very same people, so that every disagreement is a datum about the language or the standard rather than about differing competence. Two Maxwells constitute a control group of one another. Where they agree, a claim is as settled as fieldwork gets. Where they disagree, something real happened in the intervening twenty years, and the disagreement itself is publishable information.

So the audit machinery from the first review was run again, nine passes over the whole textbook this time, every paradigm cell, spelling and usage claim checked against the sequel's grammar and its corpus. The mechanics were less trivial than they sound. The sequel's PDF text layer mixes curly and straight apostrophes for the same words, so every search had to normalize first or silently miss half its attestations. Worse, Maxwell's new gloss abbreviations are A for absolutive and E for ergative, the exact mirror of the textbook's Set A (ergative) and Set B (absolutive); any comparison that string-matched labels instead of translating them would invert every person-marker check in the book. The audit translated. What came out the other side was a stack of confirmations, a short list of corrections, and twenty-two documented disagreements between the two Maxwell-team books themselves, which are logged in full at the bottom of this page.

• • •

Start with the vindications, because the first review owes somebody an apology. The textbook writes roma "because of" throughout its chapters. Patal Majzul's dialect survey assigns ruma to the majority of towns, Tecpán included, and the first review, siding with the survey and the modern corpora, listed the book's cheat-sheet geography among the errata. Enter the second witness: in the sequel's running Kaqchikel, roma outnumbers ruma 490 to 25. The book's own primary authorities, writing at length in the language they describe, prefer the form the survey marginalizes, by a factor of twenty. The chapters stand; the survey note now cites both authorities; and the reviewer notes for the record that the corpus he waved at the defendant has been outvoted.

The sequel also closed questions the first review could only flag. The potential-aspect repair vowel, xtiqa- not xtqa-, which the review argued from a handful of attestations, is now confirmed 46 to zero in the sequel's text. And three of the book's honesty flags, the bracketed [?] marks it prints on forms no source would vouch for, have been retired by direct attestation. The directional ok "inward", which the first book showed exactly once, turns out to be a full member of the sequel grammar's set: Xok ok pa rujul ri umül, "The rabbit went into its hole" (p. 230). The progressive najin, which the book dared teach only in the set phrase najin jöb' "it is raining", is the sequel's everyday workhorse: najin yetzijon "they are talking" (p. 53). And the constructed frame Tew ninna' "I feel cold", admitted in the first edition to be an extrapolation, is attested straight: nuna' tew "he feels cold" (p. 53). This is the verification habit paying its dividend. A book that marks its guesses can cash them in the moment evidence arrives; a book that prints guesses in confident type has nothing to redeem.

• • • •

The second witness did not only hand out prizes. Three of the textbook's confident claims went down on its testimony, and each is instructive about the risk of arguing from one book's silence.

First, the prohibition on -ajo' "want" with a chi clause. The first edition reasoned: the source never shows "want that you do it", therefore do not build it. The sequel builds it constantly, and in the friendliest possible place, a shopkeeper's question: ¿Achike nawajo' chi ninb'än awoma? "What do you want me to do for you?" (p. 160). Absence of evidence in a 300-page primer was never evidence of absence, and the chapter now teaches the pattern as good Kaqchikel. Second, the etymology of the directional aqanäj "upward", which the book had folded together with aqanaj "leg", two spellings of one word, the leg you rise on. A pretty story, and wrong: Maxwell's new grammar explains that the other directionals are worn-down verbs of motion, but "go up" in Kaqchikel is derived from a positional root, so no verb existed to wear down; the slot is filled instead by an adverb, cognate with a K'iche' verb of ascending, that merely resembles the leg word (p. 233). Third, the plural of the affectionate diminutive: the book confected taq ti ak'wala' "the little children" by stacking its plural particle on its diminutive, and the sequel supplies the form the language actually uses, xtäq, a reduced ch'utäq, plural of ch'uti'n (p. 265). All three corrections are in the current text, each citing the draft by page.

━━━━━

Now the part the whole post is named for. Set the textbook aside entirely and read the two source books against each other, and the same authors, twenty years apart, can be caught teaching opposite rules with equal confidence.

The headline case is the second-person object before a plural subject. ¿La ütz awäch? teaches a reduction: "we help you" is röj yaqato', "they help you" rije' yakito', the object marker absorbed. The sequel states, in its grammar and in a footnote that knows exactly what it is doing, that at- does not reduce before qa- or ki-, only before ru-, and its texts duly write xatqacha', yatkik'uxlaj, xatkikamisaj (p. 214). Not a nuance, not a register: the same cell of the same paradigm, taught oppositely by the same hands. The textbook follows the earlier book, and now says out loud that the later one differs.

Then there is manta, the counterfactual particle, where the sequel manages to disagree with the first book and with itself in the space of two hundred pages. The first book's grammar table glosses Manta k'o nupwaq, xkib'e pa jotöl as "If I had money, I would go to the States", a positive contrary-to-fact. The sequel legislates the opposite: manta is reserved for negative counterfactuals, "if X had not", with xa ta handling the positive ones, as in Xa ta ïn b'eyom "If I were rich" (pp. 277-278, 287). It then breaks its own statute in its own exercises, where a manta sentence demands the positive reading (p. 119), and elsewhere uses manta as a purpose clause, "lest" (p. 164). The first review, meanwhile, had scolded the textbook's Chapter 20 for hedging, for warning learners that the polarity of manta flips and that "context decides". Between one book's rule, the other book's opposite rule, and the other book's violations of its own rule, that hedge now stands as the single most accurate description of manta in print. Sometimes the honest shrug is the science.

The pattern repeats in miniature with wi, the location particle. The sequel's prose asserts that only ta may come between a verb and its directional (p. 230); its own corpus, sixty pages later, writes xinya' wi kan "where I left it" (p. 286), agreeing with the first book, and with the textbook's rule, against its own grammarian. And my favorite finding of the entire audit is orthographic. The sequel's grammar chapters cite the relational nouns as -uma and -uk'in, the u-forms, the ones the dialect survey endorses. The sequel's running Kaqchikel, the dialogues and readings its authors composed as language rather than about it, writes roma 490 times to 25 and rik'in 195 to 47. Maxwell writes ruma when doing grammar and roma when writing Kaqchikel. The descriptivist and the speaker share one hand and disagree about one vowel, and the speaker, as usual, wins.

•
━━━━━

It would be easy, and wrong, to read the log below as an errata sheet for Maxwell. Almost none of it is error. Two decades passed between the books; in those decades a generation of speakers grew up, the ALMG standard matured, spellings were relitigated, and the same linguists went on listening. Some entries are the language moving: the focus antipassive loosening its grip on relative clauses, family terms shedding their plurals, lax vowels spreading through the pronouns. Some are the standard moving: rin becoming rïn, weekday spellings resettling. Some are simply the difference between a beginners' simplification and an advanced book's fuller truth, which is not a contradiction but a curriculum. And a few, the best ones, are one author's rule colliding with the same author's usage, which is what happens to every linguist honest enough to leave a corpus lying next to a grammar. Drift between two editions of the truth is not noise. It is the signal of a living language being described by living people.

That is what the textbook now does, and it is the real moral of this sequel of a review. The verification habit that the first review tested, every form attested or derived or flagged, turned out to be exactly the apparatus a second witness needs: the flags told the audit where to look, the citations told it what would break, and the disagreements slotted into the book as documented variation instead of silent contradiction. A book checked against one witness is consistent. A book checked against two is starting to be true.