*3.3. The Effects of Mindfulness Depend on Acceptance*

Table 3 shows the estimated coefficients of the model including the low-level variables (Figure 2). We found a significant effect of lockdown on non-judging but not on observing nor on non-reacting. Anxiety was significantly reduced by both non-judging and nonreacting but was not influenced by observing. Depression was not predicted by observing nor by non-judging, but it was reduced by non-reacting. Lastly, anxiety increased both components of the sleep-wake cycle (sleep and wake), whereas depression did not influence any of them. Regarding indirect effects, we found a significant path from lockdown to anxiety through non-judging (b = 1.01, CI = [0.36, 1.86], SE = 0.37, β = 0.17), two significant paths from non-judging to sleep (b = −0.12, CI = [−0.29, −0.02], SE = 0.07, β = −0.12) and wake (b = −0.17, CI = [−0.34, −0.06], SE = 0.07, β = −0.20) through anxiety, and two significant paths from lockdown to both sleep (b = 0.55, CI = [0.04, 1.47], SE = 0.36, β = 0.05) and wake (b = 0.82, CI = [0.21, 1.81], SE = 0.40, β = 0.08) through non-judging and anxiety. Lockdown had no significant direct effect on anxiety, depression, sleep, or wake. However, both observing and non-judging had a significant direct effect on sleep (the first positive, the second negative). To sum up, this model confirmed that the main direct and indirect effects of time were mostly dependent on acceptance (in particular, on the non-judging facet). Observing had a direct deleterious effect only on sleep, while non-reacting reduced both anxiety and depression; none of these effects, however, were influenced by time. Even in this case, testing the model while removing the covariates did not significantly alter the results. The simplified model including only significant paths and no covariates revealed acceptable fit statistics: χ2(10) = 9.89, *p* = 0.45, CFI = 1.00, TLI = 1.00, RMSEA = 0.01, SRMR = 0.09.


**Table 3.** SEM estimated coefficients for model 2.

Note. b = unstandardized coefficient, CIlower and CIupper = lower and upper 95% bootstrapped confidence intervals of b, SE = standard error, β = standardized coefficient. Significant paths are reported in boldface. Significant level is indicated as follows: \* *p* < 0.05, \*\* *p* < 0.01.

**Figure 2.** Second model including all the subscale scores. Continuous arrows represent significant paths, while dotted arrows represent non-significant paths. Standardized coefficients are reported only for significant paths. For the sake of clarity, only the direct paths from lower levels to higher levels that were significant are shown: e.g., the paths from lockdown to distress and sleep variables are not shown because they were not significant.
