Inconclusive Prospective Effect of Dispositional Mindfulness on Looming: A Simulated Reanalysis and Comment on Prieto-Fidalgo and Calvete (2024)

Prieto-Fidalgo and Calvete (2024) analyzed data from two waves of measurement (4 months apart) on, for example, dispositional mindfulness (i.e., trait mindfulness) and looming (i.e., focusing on possible threats and perceiving them as imminent and intense) in a sample of vocational training students in Spain (between 15 and 26 years old, 37% females, n = 575) with cross-lagged panel models (i.e., statistical models where subsequent scores on two or more constructs are regressed on prior scores on the same constructs). Prieto-Fidalgo and Calvete reported statistically significant negative cross-lagged effects of prior non-judging and non-reactivity, two facets of dispositional mindfulness, on subsequent looming when adjusting for prior looming. Prieto-Fidalgo and Calvete interpreted their findings in causal terms, concluding, for example, that non-judging and non-reactivity protect against looming.

However, it is known that adjusted cross-lagged effects may be spurious (i.e., not indicating genuine/true effects) due to correlations with residuals (i.e., deviations between individual’s observed and true scores) and regression to the mean (Campbell & Kenny, 1999; Castro-Schilo & Grimm, 2018; Sorjonen et al., 2019). Given these known limitations of cross-lagged effects, the objective of the present commentary was to scrutinize the conclusions by Prieto-Fidalgo and Calvete (2024) through triangulation. If prior mindfulness had, as proposed by Prieto-Fidalgo and Calvete, a truly decreasing effect on looming, prior mindfulness could be expected to have had (1) a negative association with subsequent looming while adjusting for prior looming and (2) a positive association with prior looming while adjusting for subsequent looming. This positive association would suggest that high prior mindfulness had neutralized high prior looming and permitted individuals to attain the same subsequent level of looming as those with lower prior looming and lower prior mindfulness; (3) a negative association with the subsequent-prior looming difference.

We computed the effects presented above in data simulated to have the same sample size and correlations between variables as in the empirical data analyzed by Prieto-Fidalgo and Calvete. The simulation and analysis procedure is illustrated in the flowchart in Fig. 1. We employed simulated (i.e., computer-generated) data as the empirical data were not accessible to us. The corresponding author of the study by Prieto-Fidalgo and Calvete did not respond to our request for the data. We emphasize that both the standardized effect (i.e., effect between variables that have been rescaled to the same metric with a mean of zero and a standard deviation of one) of X on Y₂ while adjusting for Y₁ (Eq. 1, Cohen et al., 2003) and on the Y₂ − Y₁ difference (Eq. 2, Guilford, 1965) are functions of correlations between the variables. Hence, these effects will be the same in data, empirical or simulated, where the correlations between variables are identical and simulated data can be employed to evaluate what the effects are in corresponding empirical data.

Our simulations were based on the assumption that regression effects are functions of correlations (Eq. 1 and Eq. 2) and that regression effects in simulated data with the same correlations between variables, therefore, will be the same as in corresponding empirical data. The models in our triangulation were chosen because they all estimate effects of a predictor on the change in an outcome variable between measurements (and this is what we wanted to evaluate) in three complementary ways: (1) while adjusting for a prior score on the outcome; (2) while adjusting for a subsequent score on the outcome; (3) a crude effect not adjusting for a prior or a subsequent score on the outcome. The first of these models correspond to those analyzed by Prieto-Fidalgo and Calvete while models 2 and 3 are complementary, i.e., they evaluate the same thing but in different ways. By analyzing complementary models, we can evaluate the robustness of Prieto-Fidalgo and Calvetes’ findings and the validity of their conclusions. The triangulation method employed here has been used previously (e.g., Sorjonen & Melin, 2024a, 2024b, 2025). Analyses and the simulation for the present commentary were conducted with R 4.4.0 statistical software (R Core Team, 2024) using the MASS package (Venables & Ripley, 2002). The analytic script, which also generates the simulated data, is available at the Open Science Framework at https://​osf.​io/​y2bvf/​.

In accordance with conclusions by Prieto-Fidalgo and Calvete and a hypothesis (i.e., a scientific assumption) of a truly decreasing effect, prior non-judging (β = − 0.162 [− 0.238; − 0.086], p < 0.001) and non-reactivity (β = − 0.134 [− 0.206; − 0.063], p < 0.001) had statistically significant (i.e., with a low estimated probability to be a mere coincidence) negative associations with subsequent looming while adjusting for prior looming. However, in contrast to a hypothesis of a truly decreasing effect, prior non-judging (β = − 0.223 [− 0.296; − 0.149], p < 0.001) and non-reactivity (β = − 0.030 [− 0.103; 0.043], p = 0.424) did not have positive associations with prior looming while adjusting for subsequent looming. This indicates, for example, that among persons with the same subsequent level of looming (e.g., a standard score of zero), those with a high prior level of non-judging (e.g., 1) had had a lower prior level of looming (1 × − 0.223 = − 0.223) compared with those with a low prior level of non-judging (e.g., − 1 and − 1 × − 0.223 = 0.223). Hence, individuals with a high prior level of non-judging were predicted to have experienced a larger increase in looming between the timepoints (0 − (− 0.223) = 0.223) compared with individuals with the same subsequent level of looming but with a lower prior level of non-judging (0 − 0.223 = − 0.223). This suggests that a low, and not a high, prior level of non-judging had neutralized a high prior level of looming and permitted persons to attain the same subsequent level of looming as those with a lower prior level of looming but with a higher prior level of non-judging. Also failing to support a hypothesis of a truly decreasing effect, prior level of non-judging (β = 0.039 [− 0.043; 0.121], p = 0.348) and non-reactivity (β = − 0.069 [− 0.150; 0.013], p = 0.100) did not have statistically significant negative associations with the subsequent-prior level of looming difference. The six analyzed models and their results are summarized in Table 1. We re-run the analyses above ten times using different seeds for the random data generation. In each case, the results were identical to those reported above. This is due to the fact that regression effects are functions of correlations between variables (Eq. 1 and Eq. 2), which means that if correlations are identical, regression effects will also be identical.

These diverging results with decreasing, increasing, and null associations between prior dispositional mindfulness on subsequent change in looming indicate that the conclusions by Prieto-Fidalgo and Calvete (2024) of a truly decreasing, i.e., protective, effect were premature and questionable. It is important for researchers to be aware that adjusted cross-lagged effects, as well as other correlations, do not prove causality. Otherwise, they may overinterpret results, something that seems to have happened to Prieto-Fidalgo and Calvete. More broadly, the present findings warn against overly optimistic assumptions that observational (i.e., non-experimental) data, e.g., on dispositional mindfulness and mental health outcomes, allow strong causal inference if analyzed with “the right” statistical method. The random-intercept cross-lagged panel model (RI-CLPM, Hamaker et al., 2015; Mulder & Hamaker, 2021) is an extension of the traditional cross-lagged panel model and it has been proposed to be better than the traditional model at evaluating if data contains true causal effects or not. However, there are indications that RI-CLPM is susceptible to bias and spurious findings in a similar way as the traditional model (e.g., Sorjonen et al., 2023, 2024). As it seems, no magical statistical wand that can reveal causality in observational data with absolute certainty has been discovered yet. For now and for extra scrutiny, we advocate using, as we did here, triangulation and fitting complementary models to data. If findings from these complementary models point in the same direction, causal inference is substantiated (but never finally proven). However, if, as in the present commentary, results do not point in the same direction, caution is advised and it is probably best to refrain from causal conclusions. It should be noted that the present triangulation method is mainly intended to be used to scrutinize causal claims based on cross-lagged effects in observational data. It has little or no bearing on conclusions based on analyses of experimental data, e.g., randomized controlled trials (RCTs) with mindfulness interventions.

The present paper is a methodological commentary on the study by Prieto-Fidalgo and Calvete. Readers interested in more extensive reviews of research on dispositional mindfulness are referred to, for example, Rau and Williams (2016), Tomlinson et al. (2018), and Karl and Fischer (2022). For information and discussions on causal effects of mindfulness interventions, see the review by Creswell (2017), as well as RCTs (e.g., Albertson et al., 2015; Economides et al., 2018; Flett et al., 2019; Hammerdahl et al., 2025; Helminen et al., 2025; Jackman et al., 2025; Kastner, 2025; Vollbehr et al., 2025). For studies employing the triangulation method used here, see, for example, Sorjonen and Melin (2024a, 2024b, 2025).