Table 1
Mean and standard deviations of the main variables by experimental condition in Studies 1 and Study 2.
| STUDY | MUNICH – US CONCILIATORY | MUNICH – US UNCOMPROMISING | VERSAILLES – US CONCILIATORY | VERSAILLES – US UNCOMPROMISING | |
|---|---|---|---|---|---|
| N1 = 31/N2 = 112 | N1 = 34/N2 = 79 | N1 = 33/N2 = 110 | N1 = 34/N2 = 91 | ||
| Expectations about Iran’s future behavior | 1 | –11.77 (43.76) | –8.09 (45.56) | 7.70 (47.68) | –12.06 (37.64) |
| 2 | 13.93 (39.23) | 7.13 (40.6) | 21.22 (39.35) | 14.81 (35.94) | |
| Approval for US policy | 1 | 3.74 (1.22) | 3.91 (.98) | 4.11 (1.19) | 3.97 (1.02) |
| 2 | – | – | – | – | |
| Inevitability of future negative outcome | 1 | – | – | – | – |
| 2 | 3.47 (1.41) | 3.44 (1.13) | 3.20 (1.34) | 3.46 (1.28) | |
| Foreseeability of future negative outcome | 1 | – | – | – | – |
| 2 | 2.99 (1.45) | 2.82 (1.38) | 2.87 (1.48) | 2.97 (1.37) | |
| Current Knowledge | 1 | 4.05 (1.16) | 3.83 (1.27) | 3.56 (1.23) | 3.53 (1.18) |
| 2 | 3.86 (1.07) | 3.99 (1.09) | 4.13 (1.08) | 3.88 (.94) |
[i] Note: Standard deviations appear in parentheses. For the 1st variable (Expectations about future outcomes), the scale ranges from –100 to 100 (a more positive score = higher expectations for the positive future outcome relative to the negative one). All other scales are 7-points Likert-type ones. Cells containing dashes indicate that the corresponding variable was not measured in that study. n1 = sample size for study 1; n2 = sample size for study 2. Depending on the condition, past knowledge related to the Munich Accord or to the Versailles Treaty.
Figure 1
In Study 1 (top row) and Study 2 (bottom row), expectations that Iran will use its nuclear energy for civilian rather than military purposes in the future (on a scale from minus 100% = military purpose, to plus 100% = civilian purpose), depending on the “historical lesson” participants were reminded of (Versailles vs. Munich) and the current policy they believed the US had adopted toward Iran (Conciliatory vs. Uncompromising). Results of the moderation by current knowledge appear respectively on the left (–1 standard deviation) and on the right (1 standard deviation).
Table 2
Effect of the experimental manipulation (WWII Lesson ⋅ US Policy) and the moderation by Current Knowledge on the dependent variables in Studies 1 and 2.
| EXPECTATIONS ABOUT FUTURE OUTCOMES | APPROVAL FOR US POLICY | PERCEIVED FORESEEABILITY OF FUTURE OUTCOME | PERCEIVED INEVITABILITY OF FUTURE OUTCOME | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| STUDY 1 | STUDY 2 | STUDY 1 ONLY | STUDY 2 ONLY | STUDY 2 ONLY | |||||||||||
| β | T | P | β | T | P | β | T | P | β | T | P | β | T | P | |
| Intercept | –.01 | –.11 | .915 | .01 | .12 | .901 | .00 | .02 | .986 | –.01 | –.12 | .904 | –.00 | –.09 | .928 |
| Current Knowledge | –.02 | –.26 | .790 | .13 | 2.75 | .006 | –.07 | –.68 | .498 | –.03 | –.53 | .593 | –.09 | –1.74 | .081 |
| WWII Lesson | –.09 | –1.03 | .303 | –.09 | –1.77 | .078 | –.09 | .92 | .358 | –.03 | –.55 | .580 | .04 | .85 | .393 |
| US Policy | .07 | .74 | .460 | .08 | 1.61 | .109 | –.01 | –.11 | .922 | –.00 | –.09 | .926 | –.04 | –.85 | .393 |
| Lesson × Policy | –.15 | –1.68 | .096 | .01 | .32 | .744 | –.07 | –.78 | .435 | .04 | .86 | .389 | .04 | .88 | .380 |
| Lesson × Current Knowledge | –.04 | –.41 | .679 | –.05 | –1.14 | .271 | –.05 | –.49 | .626 | –.01 | –.30 | .762 | .02 | .34 | .732 |
| Policy × Current Knowledge | .13 | 1.50 | .135 | .02 | .46 | .647 | .06 | .65 | .519 | –.02 | –.36 | .719 | –.06 | –1.25 | .213 |
| Lesson × Policy × Current Knowledge | .17 | 1.91 | .058 | .12 | 2.47 | .014 | .05 | .47 | .638 | –.04 | –.81 | .417 | –.06 | –1.13 | .258 |
[i] Note: Results obtained using Model 3 of Process (Hayes, 2012). Alpha level in Study 1 = .05, Study 2 = .0336; p values below alpha level are indicated in bold font. Sample sizes: NStudy1 = 132; final NStudy2 = 392. The dependent variables used in each model appear in columns (only the 1st DV appears in both studies 1 & 2; the rest was only measured in one of the two). All variables are standardized.
Table 3
Means (and standard deviations) of the main variables in Study 3 (total N = 361).
| OPERATION CAUSED 9/11 CONDITION | OPERATION PREVENTED WORSE ATTACK CONDITION | |
|---|---|---|
| N = 195 | N = 166 | |
| Baseline expectation that Italian intervention will be successful (covariate 1) | 3.20 (1.59) | 3.3 (1.61) |
| Baseline expectation that a terrorist attack will occur in Italy (covariate 2) | 4.49 (1.07) | 4.79 (.93) |
| Expectations about future success of Italian intervention (DV1) | –4.89 (31.91) | –13.07 (32.26) |
| Expectations about future terrorist attack in Italy (DV2) | 6.84 (1.86) | 7.09 (1.74) |
| Current Knowledge (moderator 1) | 3.51 (.96) | 3.55 (1.08) |
| Confidence in predictions | 3.85 (1.34) | 3.88 (1.22) |
| Endorsement of the analogy between 1998 US intervention & Italian intervention against ISIS (moderator 2) | 3.83 (1.39) | 3.94 (1.29) |
[i] Note: Standard deviations appear in parentheses. Expectations about future outcomes (DV1) range from –100 to 100 (a more positive score = higher expectations for the positive future outcome relative to the negative one) and the likelihood of a future terrorist attack in Italy (DV2) ranges from 1 (0% certainly not) to 11 (100% certainly). All other variables were measured on 7-points, Likert-type scales, where higher number = higher expectations/current & past knowledge/analogy endorsement/confidence.
Figure 2
In Study 3, expectations that the Italian intervention against ISIS will be successful in the future (on a scale from minus 100% = not successful and will increase ISIS’ expansionist ambitions, to plus 100% = highly successful in diminishing ISIS’ power), depending on the “historical lesson” made accessible to participants (i.e., that there was a strong causal link in the past between the 1998 US intervention and the 9-11 attack vs. no such link). Results of the moderation by current knowledge appear respectively on the left (-1 standard deviation) and on the right (1 standard deviation).
Table 4
Main and moderation effects on the dependent variables of Study 3.
| EXPECTATIONS ABOUT FUTURE SUCCESS OF ITALIAN INTERVENTION | EXPECTATIONS ABOUT A FUTURE TERRORIST ATTACK IN ITALY | |||||
|---|---|---|---|---|---|---|
| β | T | P | β | T | P | |
| Intercept | –.00 | –.03 | .977 | –.00 | –.00 | .999 |
| Current Knowledge | .01 | .25 | .799 | .06 | 1.36 | .173 |
| Covariate (corresponding baseline measure) | .34 | 5.94 | <.001 | .62 | 15.83 | <.001 |
| Experimental condition | –.14 | –2.81 | .005 | –.02 | –.54 | .587 |
| Experimental condition × Current Knowledge | .08 | 1.64 | .101 | .00 | .01 | .987 |
| Intercept | –.00 | –.03 | .973 | –.00 | –.01 | .991 |
| Analogy endorsement | –.22 | –4.58 | <.001 | .13 | 3.25 | .001 |
| Covariate (corresponding baseline measure) | .35 | 6.31 | <.001 | .62 | 15.72 | <.001 |
| Experimental condition | –.13 | –2.68 | .008 | –.03 | –.65 | .515 |
| Experimental condition × Analogy endorsement | .04 | .81 | .421 | .01 | .28 | .781 |
[i] Note: These results were obtained using Model 1 of Process (Hayes, 2012). Alpha level = .0336; p values below this alpha level are indicated in bold font. Total N = 361. Experimental condition = either being told that the 1998 US intervention caused the September 11th 2001 attack vs. that 1998 US intervention prevented a worse attack; Analogy endorsement = endorsement of the analogy between the 1998 US intervention against Al-Qaeda and the 2017 Italian intervention against ISIS.
Table 5
Main and moderation effects on the dependent variables of Study 4.
| EXPECTATIONS ABOUT POSITIVE FUTURE IN CATALONIA | APPROVAL FOR FUTURE INVESTMENT IN CATALONIA | |||||
|---|---|---|---|---|---|---|
| β | T | P | β | T | P | |
| Intercept | –.00 | –.00 | .998 | .00 | .01 | .993 |
| Self-reported Current Knowledge | .10 | 2.30 | .022 | –.04 | –.86 | .390 |
| Covariate (corresponding baseline measure) | .46 | 10.49 | <.001 | .19 | 3.81 | <.001 |
| Causal role manipulation | .02 | .55 | .583 | .06 | 1.27 | .205 |
| Causal role × Self-reported Current Knowledge | –.02 | –.48 | .631 | .07 | 1.48 | .138 |
| Intercept | .00 | .00 | .998 | .00 | .00 | .992 |
| Objective Current Knowledge | –.06 | -1.37 | .392 | –.07 | -1.52 | .129 |
| Covariate (corresponding baseline measure) | .48 | 10.81 | <.001 | .18 | 3.69 | <.001 |
| Causal role manipulation | .02 | .54 | .586 | .06 | 1.29 | .197 |
| Causal role × Objective Current Knowledge | –.01 | –.25 | .805 | –.04 | –.89 | .372 |
[i] Note: These results were obtained using Model 1 of Process (Hayes, 2012). Alpha level = .0336; p values below this alpha level are indicated in bold font. Total N = 397. The p values are reported here in bilateral, but according to the preregistration of this study, one-tailed tests were conducted (i.e., the reported p values in the text were split in two). Causal role (experimental) manipulation = perceived minor vs. major role of 1991 Kosovo referendum in subsequent Kosovo War; Self-reported Current Knowledge = current knowledge as measured in Studies 1-3 through self-report scales; Objective Current Knowledge = current knowledge as measured through true/false questions.
Table 6
Two internal meta-analyses of cumulative main and moderation effects of exposure to historical analogy on predictions across Studies 1–4.
| STUDIES | INTERNAL META-ANALYSES | |||||
|---|---|---|---|---|---|---|
| 1 (N = 132) | 2 (N = 392) | 3 (N = 361) | 4 (N = 397) | FIXED-EFFECTS (N = 1282) | RANDOM-EFFECTS (N = 1282) | |
| Main effect of analogy on predictions | –.13 | .00 | –.14** | .02 | –.04 [–.09, .01] | –.05[–.14, .03] |
| Interaction effect Analogy × Current Knowledge on predictions | .17† | .12* | .09† | –.02 | .07** [.02, .13] | .08† [–.00, .16] |
| Simple effect at -1SD of current knowledge | –.32** | –.11 | –.22*** | .04 | –.12***[–.17, –.07] | –.15† [–.29, .01] |
| Simple effect at +1SD of current knowledge | .02 | .14† | –.05 | .00 | .03 [–.02, .09] | .03† [–.06, .12] |
[i] Note: All effect sizes are Pearson’s r. The random effect meta-analysis was obtained using the Sidik-Jonkman method. *** p < .001, ** p < .01, * p < .05, † p < .10. 95% confidence intervals are provided in brackets.