re-analysis of the Commission on Community Interrelations experiments

Answering an Antisemitic Remark, 1945 to 1949

A project of the Lab for the Global Study of Antisemitism, Anne Tanenbaum Centre for Jewish Studies, University of Toronto. Ron Levi and Sol Goldberg.

X in

6 publications · 1,130 checks on the transcription · 47 data tables · 3 attitude experiments, 1946 · 7 preference rounds, 1945–1946 · 1 replication, 1949

Citron (1950, 1952) experiments: transcribed dataset (attitude change)

Companion codebooks: CODEBOOK_preference.md (Chapter Two, pref_*), CODEBOOK_training.md (Chapters Four and Five, train_*), CODEBOOK_hck1953.md (the 1949 anti-Negro-remark study, hck1953_*).

Levi & Goldberg, transcribed 2026-10-13 from the page images of

Every row carries source and page (journal page numbers for the article, typed page numbers for the dissertation). Citron published no respondent-level data; everything here is a published aggregate.

Check before use: python3 scripts/verify_dataset.py recomputes 342 of Citron's printed totals, ANOVA entries and cross-table identities from these files. All pass. The one inconsistency inside Citron's own tables (dissertation Table 17, an MS printed as .00 that should round to .01) is listed in the script.

The experiments in one paragraph

All three ran in 1946. Passers-by near Times Square (and, on two days, in downtown Newark) were paid half a dollar to watch a short play in an employment-office waiting room. "Jones" loses his turn to "Goldstein" and says "These Jews are all alike." Another man answers him with one of two arguments, American tradition ("We don't want that kind of talk in America") or individual differences ("You find good and bad in every group"), in one of two manners, low emotion (calm) or high emotion (indignant). Each audience saw two versions of the scene with two different answers (design.csv). Audiences filled out attitude questionnaires before and after. Jewish audience members were dropped from the analysis. In Experiment III the answerer was "Stevenson" (non-Jewish) on four days and "Levy" (says "I'm Jewish myself") on four days, played by a different actor.

Sign convention

mean_change keeps Citron's sign: negative = change toward less prejudice (toward the pro-Jewish end of the scale). Reverse it for "positive = good" displays.

Tables

File One row per Contents
cells.csv experiment × outcome × answer × setting mean change, n; within-cell variance where printed. 58 rows.
sessions.csv Exp III audience session (date × first answer) citizenship change across the first answer, mean and n. 32 rows.
margins.csv printed total every row, column and grand total Citron printed, for verification.
anova.csv ANOVA line every analysis-of-variance table (article 3, 6, 11; dissertation 17–31).
tests.csv reported test t-tests against zero, Bartlett tests, other reported statistics.
citron_standardized.csv answer × outcome article Table 12: Citron's mean / SD, with his stars and weighted averages.
distributions.csv change score dissertation Table 15: individual change scores, citizenship, 1st answer.
items.csv item × before/after Exp I and II questionnaire items, wording and response percentages.
outcomes.csv outcome what each measure is and what happened between its two administrations.
answers.csv answer argument, manner, Citron's label, opening line of the script.
settings.csv setting place, answerer, actor, dates.
design.csv audience group which two answers each group heard, in which order.
sample.csv characteristic sample descriptions from the text.
sample_exp3_detail.csv characteristic × category Exp III sample, Times Square and Newark counts and percentages (dissertation appendix pp. 192-193).
instruments.csv instrument × item face sheet (p. 176) and the wording and instructions of Scales I to III.
scripts_text.csv script line the four answer scripts and the introduction, line by line (article appendix pp. 125-126).
summary_claims.csv claim Citron's own one-line summary of each study (dissertation Chapter Six, pp. 164-167).

Keys

experiment: exp1 (28 Mar 1946), exp2 (1 and 3 Apr 1946), exp3 (eight days, 15 Oct–20 Dec 1946).

answer: low_american, high_american, low_individual, high_individual (Citron's own shorthand: "low American" = calm American tradition); none for the effect of Jones's remark alone.

setting: ts_exp1, ts_exp2 (Stevenson, Times Square); ts_stevenson, ts_levy, newark_levy (Exp III); ts_all (both Times Square answerers, used only for Jones's remark).

outcome (details in outcomes.csv):

outcome measure what lies between the two administrations
exp1_six_item, exp2_six_item six-item score first answer; Jones repeats his remark
like_jones Scale I (5 points) first answer; a game; Jones repeats his remark
jones_right Scale II (5 points) first answer; a game; Jones repeats his remark
citizenship_remark Scale III (9 points), 1st → 2nd Jones's remark only
citizenship_answer1 Scale III, 2nd → 3rd the first answer only. Citron (1952, p. 73): "the best single measure of the effect of the answer"
citizenship_answer2 Scale III, 3rd → 4th a game; Jones repeats his remark; the second answer. Collected from 7 Nov on (6 of 8 days); cells are classified by the second answer, so they come from the groups that heard the paired answer first

Things a user should know

  1. Unit of assignment. Answers were assigned to whole audiences of 12–30 people. Exp I had one audience per answer; Exp II two; Exp III 32 sessions (sessions.csv). Citron's between-session mean squares are in anova.csv ("day to day").
  2. Same people, several outcomes. In Exp III the four outcomes come from the same respondents, so estimates pooled across them are not independent.
  3. Within-cell variances are printed only for citizenship_answer1 (dissertation Table 13). For Exp I and II they can be recovered from the per-answer t-tests in tests.csv (SD = |mean| × √n / t); verify_dataset.py shows these reproduce the within-groups MS and Citron's standardized means in Table 12.
  4. Citron's standardization (Table 12) divides each mean by the SD of changes for that answer, not by the pooled within-group SD.
  5. Jones's remark raised prejudice in Times Square (+.13) and lowered it in Newark (−.18); Citron calls the Newark result a "boomerang effect". It concerns the remark, not the answers.
  6. Other chapters. The audience-preference experiments (dissertation Chapter Two) are in the pref_* files with their own codebook (CODEBOOK_preference.md); the training experiments (Chapters Four and Five) are in the train_* files (CODEBOOK_training.md).
  7. Citron's two summaries differ. The 1950 article reports the calm-American / indignant-individual pattern as a cumulative interaction across experiments. The 1952 dissertation summary (p. 166) says only that "the manner of an answer is not related to change in attitude-set" and that the American tradition reply produced significant change; it does not mention the interaction.

Corrections relative to the May 2026 files

The May file citron_1950_cell_data_long_v3.csv (now in ../legacy_2026-05/) recorded the Stevenson / low individual within-cell variance as .80. The page reads .64, and only .64 reproduces Citron's printed pooled variances (.81, .95, .87) and the Snedecor within MS of .811.