Cross-check · Internal review

The Two Writing-Research Documents, Compared

Conflicts, coverage gaps, and a fact-check of both against the primary sources

Compiled August 7, 2026 · Covers teaching-writing-research.html (v1, Aug 1) and teaching-writing-research-2.html (v2, Aug 3) only

Throughout, V1 means teaching-writing-research.html (Aug 1) and V2 means teaching-writing-research-2.html (Aug 3). Findings are marked FIX for a factual error, TIGHTEN for something defensible but overstated or mis-sourced, HOUSEKEEPING for a mechanical defect, and VERIFIED for a claim checked against the original and confirmed.

The verdict in six lines

Neither document supersedes the other. V2 is far more reliable on evidence; V1 holds three things V2 dropped entirely — the curriculum-market verdict, the grade-level scope table, and the read on the student's own writing.
The one conflict that matters is daily dosage. V1 prescribes roughly 15–20 minutes a day; V2 prescribes 45–60. That is a 3× difference in the core instruction and the two cannot both be implemented.
V2 has one genuine fabrication and one garbled statistic — a quotation that isn't in its source, and a study count off by an order of magnitude. Both are fixable in a line each.
V1's errors are errors of attribution and overstatement, not invention: it credits a recommendation to the wrong federal guide, oversells one method's evidence base, and links to the wrong standards strand.
Everything else held. Roughly 35 individual figures were checked against the original reports; all but the ones listed in §4 matched exactly — including several that read like AI overstatement and turned out to be verbatim.
No dead or invented links in either document. All 37 external URLs in V2 and all 8 source entries in V1 resolve.

Outright errors (both in V2)

Overstated or mis-sourced

Conflicts between the two

~35

Figures verified correct

1. What each document is

	V1 — `teaching-writing-research.html`	V2 — `teaching-writing-research-2.html`
Date / size	Aug 1, 2026 · ~16 KB	Aug 3, 2026 · ~73 KB
Purpose	A practical brief: what the research says, what to buy, what the student's first pieces show, and a year map.	A rigorous evidence review: effect sizes, evidence tiering, and a ~45-source shelf.
Evidence handling	Narrative. Names findings, no numbers, no tiering.	Every claim carries a citation and an evidence-type badge (META / STUDY / GUIDE / SURVEY / THEORY / WISDOM).
Specific to the student	Yes — a full read of the first two assignments.	No. Entirely generic; never refers to her work.
Covers what to teach	Yes — the modes an eighth grader must cover, plus a quarterly map.	No. Method only; no scope or sequence.
Self-description	—	Calls itself a "Companion to teaching-writing-research.html." That is accurate, and it is the right way to read the pair.

2. Where they conflict

Six real disagreements. Only the first two change what happens day to day.

Question	V1 says	V2 says	Which is better supported
Daily dosage	10 min freewrite + 5 min sentence work, plus 2–3 lessons a week. Roughly 15–20 min/day.	3–5 min quickwrite + 25–40 min sustained composing on the current piece. Roughly 45–60 min/day.	V2, but with a caveat: its hour-a-day floor comes from a federal recommendation the panel itself rates minimal evidence (see §4). The deliberate-practice argument behind it is stronger than the citation is.
Ship cadence	A piece goes through the full process and gets shared every 1–2 weeks.	Ship every 2–4 weeks — and specifically to a reader beyond the mentor.	V2 on both counts. The audience research measured real readers against writing-for-a-grade, so the "beyond the mentor" constraint is doing real work. V1's examples (read aloud at dinner, mailed to a grandparent) already satisfy it.
The Writing Revolution's standing	Its core claim is "backed by strong results"; the book is "the one purchase worth considering."	Whole-method trials are thin; tagged as practitioner wisdom resting on well-evidenced components — "a reasonable bet, honestly labeled."	V2. The components (sentence combining, expansion) are well evidenced; the packaged method is not independently trialed. V1 overstates.
What the federal guides recommend	"The What Works Clearinghouse echoes this: sentence construction … is one of its formal recommendations."	The secondary guide has three recommendations: explicit strategies (strong), integrate reading and writing (moderate), use assessment to inform instruction (minimal).	V2. V1's claim is true of the elementary guide (grades 1–6) and false of the secondary one — see §4.
Shape of sentence work	5 minutes, daily.	10 minutes, 3×/week.	Either. Same weekly volume, different rhythm — but a handout can only implement one, so pick deliberately rather than by accident.
How hard to push on grades	"Feedback beats grades" — stated as a preference.	Don't grade drafts; ideally don't grade much at all. Grade only finished work, against criteria she has already self-assessed against.	V2. Same direction, much firmer, and grounded in a specific finding: comments alone outperformed grades and grades-with-comments.

3. Coverage gaps: what only one has

This is the main reason to keep both files. The gaps run in both directions and barely overlap.

Only in V1 — and V2 offers no substitute	Only in V2 — and V1 offers no substitute
The curriculum-market survey. IEW, Brave Writer, The Writing Revolution, and the free options, each with a verdict — including the conclusion that nothing free rises to "excellent." This is the answer to "should I buy anything?", and V2 dropped it entirely. The grade-level scope. The narrative / informative / argument / foundations table — what the year has to cover. V2 is all method and contains no account of content whatsoever. For a curriculum document this is the largest single hole. The read on the student's actual writing. Five named strengths and five growth areas drawn from her first two pieces. V2 never touches her work. The quarterly map and the interest-driven project ideas attached to each quarter. Named free tools for mechanics reference and convention practice.	Every number, plus the evidence tiering that distinguishes a meta-analysis from a master teacher's opinion. V1 asserts; V2 shows its work. The cognitive model — knowledge telling vs. transforming, the working-memory bottleneck, deliberate practice. This is the part that lets a teacher improvise correctly mid-lesson instead of following a script. Motivation as a cause of skill, not a nicety — including the governing rule that any practice visibly costing enthusiasm yields to the writer. The one-on-one adaptation. How to translate classroom-validated practices to a mentorship: the mentor becomes the collaborator; peer feedback becomes outside readers. What actively backfires: packaged trait rubrics, the five-paragraph formula as a destination, workshop ritual without instruction, grading everything, marking every error. AI guidance. Second reader between conferences, fine; ghostwriting, fatal. Nothing in V1. The mechanics of feedback: the conference protocol, co-built rubrics, the portfolio, the "moves I own" list, quarterly review. The source shelf — ~45 entries with working links.

Only in V1 — and V2 offers no substitute

Only in V2 — and V1 offers no substitute

The curriculum-market survey. IEW, Brave Writer, The Writing Revolution, and the free options, each with a verdict — including the conclusion that nothing free rises to "excellent." This is the answer to "should I buy anything?", and V2 dropped it entirely.
The grade-level scope. The narrative / informative / argument / foundations table — what the year has to cover. V2 is all method and contains no account of content whatsoever. For a curriculum document this is the largest single hole.
The read on the student's actual writing. Five named strengths and five growth areas drawn from her first two pieces. V2 never touches her work.
The quarterly map and the interest-driven project ideas attached to each quarter.
Named free tools for mechanics reference and convention practice.

Every number, plus the evidence tiering that distinguishes a meta-analysis from a master teacher's opinion. V1 asserts; V2 shows its work.
The cognitive model — knowledge telling vs. transforming, the working-memory bottleneck, deliberate practice. This is the part that lets a teacher improvise correctly mid-lesson instead of following a script.
Motivation as a cause of skill, not a nicety — including the governing rule that any practice visibly costing enthusiasm yields to the writer.
The one-on-one adaptation. How to translate classroom-validated practices to a mentorship: the mentor becomes the collaborator; peer feedback becomes outside readers.
What actively backfires: packaged trait rubrics, the five-paragraph formula as a destination, workshop ritual without instruction, grading everything, marking every error.
AI guidance. Second reader between conferences, fine; ghostwriting, fatal. Nothing in V1.
The mechanics of feedback: the conference protocol, co-built rubrics, the portfolio, the "moves I own" list, quarterly review.
The source shelf — ~45 entries with working links.

4. What doesn't check out

Status: corrected in the source documents, retained here as the record

Every finding in this section has since been fixed in V1 and V2 themselves. The findings are deliberately left in place, stated as they were found, so there is a permanent record of what was wrong and what the correction was — do not expect to see these errors in the current files. The one exception is the dosage conflict in §2, which is a disagreement between the two documents rather than an error in either, and remains open.

V2 — two errors, three imprecisions

Fix these two

FIX The Hillocks study count is wrong by an order of magnitude. §1 describes it as "the first great meta-analysis, covering ~500 experimental studies from 1963–1982," and §6 repeats "the least effective thing measured across 500 studies." Per Hillocks' own published synthesis — the PDF V2 already links — several hundred studies were screened and the meta-analysis rested on 60 well-designed studies with 72 experimental treatments, sitting inside a broader review of about 2,000 studies. The number 500 appears nowhere. Correct to: 60 studies / 72 treatments, drawn from a review of ~2,000.
FIX The National Commission on Writing quotation is not in the report. §5.8 puts "if students are to learn, they must write" inside quotation marks. The actual passage runs: students must struggle with the details, wrestle with the facts, and rework raw information into language they can communicate — "In short, they must write." Correct to: quote the real sentence, or drop the quotation marks and paraphrase openly.

Tighten these three

TIGHTEN Two figures are attributed to the wrong version of the same study. "123 studies, 154 effect sizes" and the −0.32 for grammar are the numbers from the peer-reviewed journal article. The Carnegie report reports 142 studies / 176 effect sizes and gives grammar no number at all — there it is a sidebar note, not one of the eleven elements. The document's own method note claims these figures "were read directly from the Carnegie reports," which is not where they came from. Fix: cite the journal version for those two figures, or switch to 142/176.
TIGHTEN The league table's first band is labelled "eleven elements" but has twelve rows. Same root cause: the journal's eleven include grammar and exclude writing-for-content-learning; the Carnegie eleven are the reverse. The table merges both lists. Fix: relabel the band, or split grammar into its own row-group with its own source line.
TIGHTEN The hour-a-day dosage floor is presented more firmly than its source supports. It rests on the elementary guide's "provide daily time for students to write," which the panel rates minimal evidence. V2 is scrupulous about this elsewhere — it explicitly flags the secondary guide's third recommendation as minimal — so the omission reads as inconsistent rather than dishonest. Since this is the source of the biggest conflict in §2, the caveat matters.

V1 — one attribution error, three overstatements

Corrections

FIX The federal recommendation is credited to the wrong guide. "The What Works Clearinghouse echoes this: sentence construction (combining and expansion) is one of its formal recommendations." That is Recommendation 3 of the elementary guide (grades 1–6, moderate evidence). The secondary guide (grades 6–12) has three recommendations and sentence construction is not among them. In a document about an eighth grader, this is misleading as written.
TIGHTEN "A meta-analysis of hundreds of studies." It is 123 documents yielding 154 effect sizes (journal version), or 142 studies yielding 176 effect sizes (Carnegie report). "Hundreds" overstates it.
TIGHTEN The Writing Revolution's evidence base. "Backed by strong results" is true of the method's components and not of the packaged method, which lacks independent trials. See §2.
TIGHTEN The standards link points to the wrong strand. It targets the History/Social Studies/Science & Technical writing standards, which contain no narrative standard — while the table directly above it describes narrative, informative, and argument. The general grade-8 writing standards are the right target.

5. What was verified and holds

Roughly 35 individual figures were checked against the original documents. Everything not listed in §4 matched — including several claims that read like AI overstatement and turned out to be verbatim accurate.

Confirmed against the primary sources

VERIFIED All eleven headline effect sizes in V2's league table — 0.82, 0.82, 0.75, 0.70, 0.55, 0.50, 0.32, 0.32, 0.32, 0.25, 0.23 — are exact matches to the Carnegie report, in the right order.
VERIFIED The strategy-instruction breakdown: 1.14 for the self-regulated variant vs. 0.62 for other strategy teaching; 1.02 for struggling writers vs. 0.70 across the full range; 0.70 for word processing with low-achieving writers; 0.46 for weaker writers in the paired sentence-combining study. All exact.
VERIFIED "The process approach is near zero without trained teachers" — this reads like an exaggeration and is the report's own finding: 0.46 with training, "negligible" without, except 0.27 in grades 4–6.
VERIFIED The grammar numbers: −0.32 in the journal version, −0.29 in Hillocks. Also Hillocks' inquiry 0.56, evaluation scales 0.36, sentence combining 0.35, free writing 0.16 — all consistent with the source.
VERIFIED The classroom-exposure statistics, every one: 260 classrooms, 20 schools, five states, 1,520 teachers surveyed, 7.7% of observed class time spent writing a paragraph or more, 1.6 pages a week in English and 2.1 across the other three subjects combined. One thing V2 understates — those 20 schools were selected for reputations of excellence in teaching writing, which makes 7.7% worse than it sounds, not better.
VERIFIED The secondary practice guide's three recommendations and their evidence levels, exactly as V2 states them.
VERIFIED The national assessment figures: about 27% of eighth graders at or above Proficient, 20% below Basic.
VERIFIED V1's ranking is internally sound. Its top-seven list, though it carries no numbers, orders the practices consistently with the actual effect sizes — including placing word processing above sentence combining, which is correct (0.55 vs. 0.50).
VERIFIED Link integrity. All 37 external URLs in V2 and every source entry in V1 resolve, including the homeschool-blog links that looked invented. The handful of 403 responses are publisher bot-blocking on DOI and journal hosts, not dead links.

6. Housekeeping

As with §4, all three items below have since been fixed; they are recorded here as found.

HOUSEKEEPING V2 One broken internal link. The "how to read this document" paragraph links to the source shelf as #sources, but the section's actual id is s10. There is no sources anchor in the file, so the link does nothing. The table of contents links correctly; only this one is wrong.
HOUSEKEEPING V1 No print stylesheet. V1 has no @media print block at all, so it prints with screen margins, live link colors, and tables sliced across page breaks. V2 has one.
HOUSEKEEPING V1 Naming. Two references to the student by relationship have been changed to "the student." Neither document ever contained her given name.

7. How this was checked

Both documents were read in full. Every effect size, survey statistic, study count, and recommendation was then checked against the primary source rather than from memory — the original reports were downloaded and searched directly:

The Carnegie meta-analysis report (full PDF) — for the eleven effect sizes, the strategy-instruction breakdown, the training/no-training split, the study and effect-size counts, and the grammar sidebar.
The journal abstract of the same meta-analysis — which resolved where the 123/154 counts and the −0.32 grammar figure actually come from.
Hillocks' own published synthesis (PDF) — for the study count, which is where V2's largest error surfaced.
The classroom-exposure study (full PDF) — for all seven of its statistics.
Both federal practice guide pages — for the recommendation wording and evidence levels.
A source search for the disputed quotation, which returned the report's actual sentence.
An automated status check of all 45 external links across the two documents.

Bottom line

V2 is the more trustworthy document on evidence and should be treated as the authority wherever the two disagree. But it is genuinely a companion, not a replacement: the curriculum-market verdict, the grade-level scope table, and the read on the student's own writing exist only in V1, and nothing in V2 covers them. The corrections in §4 and §6 have been applied to both files; what remains open is the dosage conflict in §2, which is a choice to be made rather than an error to be fixed.