C6 Phones in lessons

survived narrowedone dot per attack, oldest first

1. #1 narrowed 7 October 2026 · Allcott, Baron, Dee, Duckworth, Gentzkow & Jacob 2026 (preprint) · target: kill-condition

   When US middle and high schools moved from mostly no-show or no phone rules to locking pupils' phones in pouches for the whole school day, in-class phone use fell sharply but average standardised test scores did not rise (pooled effect −0.006 school-level SD, ruling out gains above about 0.007 student-level SD), so C6's unconditioned claim that pupils learn less with free access than under a restriction fails on average and its kill-condition is met for this large sample.

   Strongest reading and verdict reasoning

   Steelmans to: C6's kill-condition is a difference-in-differences study in which free access to phones during lessons does not lower learning outcomes compared with a ban or restriction; this preregistered staggered DiD (1,778 adopters vs 42,919 never-adopters, Callaway–Sant'Anna with covariates, parallel pre-trends, robust to refined control groups excluding schools with strict policies or statewide bell-to-bell bans) finds a precisely estimated null on test scores, and together with the Swedish null (Kessel et al. 2020) leaves the average effect unsupported.

   Why: Nightly-attacks 2026-10-07 Europe/Stockholm. Full author PDF read (not stored). Contrast: 62 % of adopters (teacher survey) previously had a no-show rule and 23 % none; teacher-reported share of pupils using phones in class for personal reasons fell 61 % → 13 %; GPS phone activity fell ~0.18 log points. Table 3: pooled combined ATT −0.006 (SE 0.010), maths −0.002, ELA −0.006 (≈ −0.003 student-level SD); high schools combined +0.026 (SE 0.017), maths +0.047 (SE 0.018, p < .05; ≈ 0.024 student-level SD); middle schools combined −0.025 (SE 0.012, p < .05), maths −0.026 (p < .05). Kill withheld: (1) working paper, not yet peer reviewed; (2) high-school maths rose in the same study, and gains after bans are reported elsewhere (Beland & Murphy 2016 in England, driven by low achievers; Abrahamsson for Norwegian girls; Figlio & Özek 2025 Florida year 2, Lichand et al. 2026 Rio, Beneito & Vicente-Chirivella 2022 Spain, as reported by Allcott et al.); (3) school-level outcomes, at most three years after adoption, pouches imperfectly enforced, comparison schools not phone-free. Not a restatement: Kessel et al. 2020 was only an open challenge in Bounds and the claim text was unconditioned; this is the first ledger entry on C6. Claim and Bounds rewritten same turn; brief and the practical rule rewritten. Preprint flagged.

2. #2 narrowed 9 October 2026 · Shi & Villarroel 2026 (preprint) · target: kill-condition

   In Chilean classrooms where teachers set their own phone rules, restricting phones (never allow, or allow only a few times) left standardised math and reading scores unchanged relative to freer classroom policies (estimates near zero; gains of about 0.03–0.06 SD ruled out), despite a modest drop in eighth graders' recreational in-class phone use, so C6's kill-condition is met for classroom-level restrictions and the residual "some settings" picture needs a classroom null.

   Strongest reading and verdict reasoning

   Steelmans to: C6's kill-condition is a field DiD/RCT in which free access to phones during lessons does not lower learning outcomes compared with a ban or restriction; this within-teacher and within-student design finds precisely that for classroom-level teacher restrictions in Chile (null on scores; only modest behavioural change in grade 8), so the claim must record that classroom-level restrictions are another setting where an average learning gain is not established.

   Why: Nightly-attacks 2026-10-09 Europe/Stockholm. Full EdWorkingPaper PDF read (not stored). Table 3: Never Allow or Allow a Few Times vs often/always — grade 8 teacher-FE 0.007 (SE 0.026), grade 10 −0.003 (0.027); grade 8 student-FE −0.012 (0.018), grade 10 0.005 (0.016); authors rule out gains of about 0.03–0.06σ. Phone use −3 pp (~4% of mean) in grade 8 only. Capability −0.023/−0.022 (~2+ pp, ~3%) in grade 8. Kill withheld: (1) working paper; (2) classroom-level teacher discretion with uneven compliance, not a school-day collection ban; (3) school-level gains remain (Beland & Murphy 2016; Abrahamsson; Figlio & Özek 2025; Lichand et al. 2026 as reported); (4) many Chilean teachers treat phones as pedagogical tools. Not a restatement of Allcott/Kessel: different country, identification (within-teacher / within-student), and level (classroom vs school-day pouch/ban). Claim and Bounds rewritten same turn; brief and the practical rule Limits rewritten. Preprint flagged.

3. #3 survived 9 October 2026 · Figlio & Özek 2025 (preprint) · target: kill-condition

   In a large Florida school district that banned phones for the whole school day, comparing schools with high versus low pre-ban student smartphone activity, free-er access did not lower learning outcomes relative to the stricter restriction (test scores did not rise after the ban), so C6's kill-condition is met for this district-level ban.

   Strongest reading and verdict reasoning

   Steelmans to: C6 dies on a field DiD in which free access does not lower learning outcomes vs a ban/restriction; if Figlio & Özek's high-vs-low activity contrast showed a null or negative effect of the ban on scores, the kill-condition would be met for this US district policy.

   Why: Nightly-attacks 2026-10-09 Europe/Stockholm. Full EdWorkingPaper PDF read (not stored). Genuine kill-condition test: DiD of a binding all-day district ban (backpacks/purses all day under Florida's instructional-time law), high vs low pre-ban Advan smartphone activity. Kill fails: year-2 average scores +0.622 percentiles (p < .05); spring-only +1.083; middle/high year 2 +0.905 / spring +1.333; males spring +1.421. First-year discipline surge then fade; unexcused absences down, explaining ~half of middle/high gains (exploratory). Design understates effects (both groups treated). Distinct from Allcott (pouches, average null), Shi (Chile classroom null), Beland & Murphy, Abrahamsson. Claim field unchanged; Bounds note the survive and the Florida gain. Status remains weakened.

4. #4 survived 11 October 2026 · Lichand, Moreno-Louzada, da Costa & Gentzkow 2026 (preprint) · target: kill-condition

   In Rio de Janeiro's municipal middle schools, where a 2023–24 city-wide ban on non-pedagogical phone use hit hardest in schools that had no strict phone rules before, freer prior access did not lower learning relative to the stricter restriction (test scores did not rise after the ban in those schools relative to schools already strict), so C6's kill-condition is met for this middle-income, city-wide ban.

   Strongest reading and verdict reasoning

   Steelmans to: C6 dies on a field DiD in which free access to phones during lessons does not lower learning outcomes compared with a ban or restriction; Rio's ban applied to all 1,000+ municipal schools at once (not chosen by each school), quarterly student-level test scores give two years of pre-trends, and if schools moving from partial or no rules to the ban had not gained relative to schools that were already strict, the kill-condition would be met in a large middle-income system.

   Why: Nightly-attacks 2026-10-11 Europe/Stockholm. Full author PDF read (27 April 2026 version; not stored). Genuine kill-condition test: DiD of treatment schools (partial or no pre-ban phone rules) against control schools (already strict), grades 6–9; full sample 196,757 students in 325 schools; matched sample 97,380 students in 154 schools, 549,969 student-quarter observations. Principal survey: share of schools where most or all pupils used phones in class fell from 52 % to 19 % (treatment) and from 40 % to 15 % (control). Kill fails: Table D2 (matched sample), school-wide ban × treated on the maths+Portuguese z-score index +0.061 (SE 0.028, p < .05), stable across specifications (0.073 with student fixed effects; 0.084 with CRE × post); maths +0.077 (SE 0.037, p < .05), Portuguese +0.045 (SE 0.024, p < .10); event-study pre-trends flat; gains concentrated among low prior achievers and in grades 7–8, and in schools whose principals reported phone use fell. The earlier classroom-only phase (Q3–Q4 2023) showed no significant effect (+0.010, SE 0.021). The authors' extrapolation to about 0.2 SD against no ban is a calibration from coarse principal reports and is not used here. Limits: working paper; compares exposure intensity under one ban, not ban vs no ban; treatment and control schools differ in baseline levels (treatment schools poorer, about 0.1 SD lower scores); a harder 2024 test applied to all schools; phone use measured by principals' retrospective reports. Not a duplicate: first ledger entry from this study; distinct country, system and identification from Figlio & Özek 2025 (Florida, smartphone-activity terciles), Allcott et al. 2026 (US pouches), Shi & Villarroel 2026 (Chile classroom rules), Kessel et al. 2020 (Sweden). Previously named in Bounds only as reported second-hand by Allcott et al.; now read and a separate source (ATTN-011). Claim field unchanged; Bounds note the survive and the Rio gain. Status unchanged (weakened from earlier narrowings).

5. #5 survived 14 October 2026 · Sungu, Choudhury & Bjerre-Nielsen 2026 · target: kill-condition

   In a preregistered randomised controlled trial of about 17,000 higher-education students in India, mandatory in-class phone collection did not raise end-of-semester grades relative to free phone access during lectures, so freer access did not lower learning and C6's kill-condition is met for this large field experiment.

   Strongest reading and verdict reasoning

   Steelmans to: C6 dies on a field RCT/DiD in which free access to phones during lessons does not lower learning outcomes compared with a ban or restriction; this preregistered cohort-randomised RCT (16,683 students, 10 HEIs, lecture-only phone boxes vs business-as-usual) is exactly that design, and if GPA had not risen under the ban the kill-condition would be met in a large middle-income system with peer-reviewed evidence.

   Why: Nightly-attacks 2026-10-14 Europe/Stockholm. Science article page read (Vol. 394, Issue 6820, pp. 178–181, published 8 October 2026); Crossref DOI checked; design detail confirmed against the author PDF of the earlier working-paper version (not stored). Genuine kill-condition test: cohort-level randomisation, mandatory phone deposit at lecture start vs free access, end-of-semester GPA with mostly external blind grading. Kill fails: +0.073 SD [95% CI 0.003, 0.144], one-sided p = 0.021; larger for lower prior performers and first-years. Limits: higher education (mean age 19.4), not K–12 pupils; lecture-only ban, not all-day; India HEIs. Not a duplicate: first ledger entry from this study; distinct design (true RCT), country, level (HEI), and peer-reviewed status from Figlio & Özek 2025, Lichand et al. 2026, Allcott et al. 2026, Shi & Villarroel 2026, Kessel et al. 2020. Claim field unchanged; Bounds and Ages note the survive and the India HEI gain. Status unchanged (weakened from earlier narrowings). With this survival C6 reaches the floor of 3 survivals.