Findings Register

Navigation

Hub: 00 - Fact-Check MOC · Previous: 04 - Results Stage 1 · Next: 06 - Open Questions

Ten discrepancies with the direction each one cuts, the quoted passage behind each, and the checks that came back clean.

Every discrepancy found, with the direction it cuts. Direction matters more than count: errors that consistently understate a case suggest carelessness, errors that consistently overstate it suggest motivated reasoning. As of this writing the pattern runs against the book’s own argument, which is worth stating plainly and worth continuing to watch.

Quoted support

Added 2026-08-04 on receipt of the PDF. Findings 1 through 3 were previously recorded as confirmed without the quoted passage that standing rule 1 requires. Locators are book pages; add 15 for the PDF page.

Finding 1 — Ch. 2, p. 27. The sentence in full:

the term “massacre” was used to describe Palestinians killing Israelis 344 times, versus never for Israelis killing Palestinians. […] The use of “barbaric” was used 14 times to describe the killing of Israelis versus zero times for Palestinians. The term “savage” was used 16 times to describe the killing of Israelis versus zero times for Palestinians, and the term “slaughter” was used 120 times for the killing of Israelis and only once for Palestinians, for a total respectively of 494 to 0, or a ratio of 531-to-3.

The components give 344 + 14 + 16 + 120 = 494 on the Israeli side and 0 + 0 + 0 + 1 = 1 on the Palestinian side. The stated total of “494 to 0” contradicts the same sentence’s “only once for Palestinians,” and “531-to-3” matches neither the components nor the stated total. Three mutually inconsistent figures, confirmed as recorded.

Finding 2 — Fig. 2.1b, p. 29. Caption: “Emotive terms used about Palestinian and Israeli victims during the first 100 days of the conflict on TV.” The body text on the same page reads: “Narrowing our scope to the first 30 days of each conflict for cable news, we see a similar glaring double standard.” The values plotted are the 30-day values given in that paragraph. The caption is wrong, not the chart.

Finding 3 — text p. 27 vs Fig. 2.1a, p. 28. The text states “slaughter” was used “only once for Palestinians.” Figure 2.1a plots zero Palestinian “slaughter” mentions at all seven outlets.

Finding 10 — Ch. 2, p. 35 vs Ch. 7, p. 127. The same measurement, stated twice.

Chapter 2:

In our 100-day survey period, CNN used the “Hamas-run” label and related terms 277 times and MSNBC used it 146 times, despite neither using it once between October 7, 2023 and October 17, 2023.

Chapter 7:

As mentioned in Chapter 2, in our 100-day survey period, CNN used the “Hamas-run” label 247 times and MSNBC used it 137 times.

Same survey window, same networks, same label, and the second passage explicitly presents itself as a restatement of the first. The figures differ by 30 and 9 respectively, so this is not a single transposition.

The likely innocent explanation is visible in the wording: Chapter 2 counts the label “and related terms,” Chapter 7 counts “the ‘Hamas-run’ label.” Those could legitimately be two different measurements — the broader family versus the bare string. If so the defect is that two different quantities are presented as one figure under an explicit cross-reference, which a reader has no way to detect. Either way the restatement understates the book’s own case, and the discrepancy is real as published.

Resolving it requires knowing which terms were in each query, which is the unpublished coding layer described in 02 - Data Sources.

Withdrawn

The 1,053:43 entry, logged and withdrawn 2026-08-04. Earlier the same day a discrepancy was logged on the grounds that the 30-day TV emotive-terms comparison recorded in 08 - Search Terms and Their Allegations as 1,053:43 did not reconcile with the book’s own figures of MSNBC 348:12 and CNN 373:18, which total 721:30.

A full-text search of the book resolves it the other way: the string 1,053 appears nowhere in the book, and neither does 721. The error was in our note, not in the book. Note 08 has been corrected to 721:30 with the per-network breakdown.

Recorded rather than deleted because the direction of the mistake is worth noticing. It would have been logged as a book error that inflated a ratio, which is the direction this audit should be most suspicious of finding in its own work.

Verified as internally consistent

Worth recording, since a register of only errors misleads.

• Fig. 2.1a per-outlet figures sum correctly: massacre 80+43+12+33+124+50+2 = 344; slaughter 18+14+3+5+53+25+2 = 120.
• Fig. 2.1b TV counts match the source Nation study exactly across all four terms and both networks.
• Intercept summary statistics reproduce the article’s antisemitism/Islamophobia figures (549 vs 79, 87 percent) exactly.

Checked against the book text on 2026-08-04:

• Ch. 2, pp. 29–30. Both per-network 30-day summaries reconcile with Fig. 2.1b. MSNBC: 177+46+23+102 = 348 and 8+0+0+4 = 12, matching “a difference of 348 to 12, or a ratio of 29 to 1.” CNN: 225+56+13+79 = 373 and 16+0+0+2 = 18, matching “373 to 18, or a ratio of 21 to 1.” Both reconcile, and their total of 721:30 is what corrected the withdrawn entry below.
• Ch. 2, p. 31. Print “human shields” reconciles three ways: the stated 231 total, the per-outlet breakdown given in the text (NYT 150, WaPo 56, Politico 22, Axios 3), and Fig. 2.2a. The five Israeli-side mentions are all Washington Post, as claimed.
• Ch. 2, p. 30. TV “human shields”: 888 = CNN 438 + MSNBC 450, matching Fig. 2.2b.
• Ch. 2, p. 31. Print “right to defend itself”: Fig. 2.3 sums to 459 for Israelis (74+63+118+54+99+27+24) and 4 for Palestinians (CNN 1, NYT 1, USA Today 2), matching the text exactly.

The pattern so far is that the strongest findings are also the ones whose arithmetic holds. The defects cluster in the emotive-terms material in the opening pages of Chapter 2.

Standing assessment

Ten discrepancies, nine confirmed and one unresolved, none individually fatal. Four are pure typography or citation metadata. The substantive ones (1, 8, 9, 10) all reduce rather than inflate the strength of the book’s case. That is evidence of a citation and figure apparatus that was not tightly checked, not of a thumb on the scale.

The serious problems with the book are not in this register. They are structural: the mismatch between a twelve-month thesis and predominantly 30-to-100-day data, the absence of any engagement with the obvious objection that emotive terms track violence-form rather than victim identity, and the gap between measuring asymmetry and asserting causation. See 06 - Open Questions.