ABSTRACT
People generally perceive a stronger link between smoking and cancer than between cancer and smoking. Generally, prior research on asymmetrical causal reasoning has not distinguished predictive (searching for effects) and diagnostic reasoning (searching for causes) from the order in which causes and effects are presented. Across 6 studies (overall N = 627), we show that order and reasoning have an additive influence on the causality perception: causes, spatially or temporally presented before the effect, strengthen the causality attribution associated to predictive (vs. diagnostic) frames. Moreover, we show that order and reasoning frame are bi-directionally related, as the cause-first order triggers predictive reasoning and vice versa, and people mentally maintain the cause-first order when envisaging a causal relation. Besides its methodological contribution to the causal reasoning literature, this research demonstrates the powerful role of word order in causal reasoning. Implications for the role of word order in communication and risk prevention are discussed.
Acknowledgements
We would like to thank Alice Spollon for collecting the data of Study 1.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Maria Laura Bettinsoli http://orcid.org/0000-0002-5860-2675
Notes
1 When analyzing the effect of positive and negative correlations within a full factorial linear model with order seen and perceived correlation direction as within participants factors and versions and sentences as random factors, the effect of order seen was not modified by the direction of correlations. Correlation direction only produced a main effect, with negative correlations (M = 70.09, SD = 29.70) being associated with greater perceived causality compared to positive correlations (M = 62.81, SD = 28.63), F(1,152) = 18.23, p < .001.
2 We also ran a General Linear Mixed Model in which we included the order, the type of sentence and the interaction between order and type of sentence as fixed effect, and participants, the order and the sentences as random. The model revealed two main effects of Order F(1,156) = 7.48, p = .007 and Type of Sentence F(1,11) = 78.58, p < .001, but no interaction. As in the first analysis, when Cause preceded the Effect, participants perceived the relation between the elements as stronger than when presented in Effect–Cause order. In Addition, as highly expectable some pairs of elements were rated as stronger related (e.g. Smoking–Cancer) than others (e.g. Potassium–Hypertension), which can be due to a higher familiarity and salience of some of the elements presented in the sentence.
3 As for Study 1, we ran a General Linear Mixed Model in which we included the order, the type of sentence and the interaction between order and type of sentence as fixed effect, and participants, the order and the sentences as random. As in Study 1, the model revealed two main effects of Order F(1,99) = 4.08, p = .045 and Type of Sentence F(1,11) = 82.05, p < .001, but no interaction. Again, when Cause preceded the Effect, participants perceived the relation between the elements as stronger than when presented in Effect–Cause order. As in Study 1, some pairs of elements were rated as more strongly related (e.g. Smoking–Cancer) than others (e.g. Potassium–Hypertension).
4 We ran two additional analyses in which we either (1) excluded participants who failed to identify the correct causality direction (i.e. Erroneously identifying a pair as Cause–Effect when it was Effect–Cause, and vice versa) or (2) taking into account in the statistical model the order identified by participants, which means taking into account the correct causal direction, but also those 10% of incorrect causal directions. Both models revealed the same pattern of results in line with our hypotheses, respectively (1) F(1,95) = 4.18, p = .043 and (2) F(1,99) = 3.97, p = .048.