As shown in , participants with ADHD score lower on the composite measure of mindfulness and three of the four subscales, Describing, Acting in Awareness, and Accepting without Judgment compared to non-ADHD participants (see ). On the Observing subscale, no group differences are evident and, in fact, the mean score for the ADHD participants is slightly higher (36.22) than non-ADHD controls (34.65). On the TCI scales, participants with ADHD report themselves as more Novelty-Seeking and more Self-Transcendent than non-ADHD participants, replicating the findings found in a previous study on an independent set of participants (Lynn et al., 2005
). In addition, ADHD participants score lower on Self-Directedness than controls, a finding not present in our earlier study (Lynn et al., 2005
) but that replicates others (Anckarsater et al., 2006
; Cho et al., 2008
TCI and KIMS scores by lifetime ADHD diagnosis
Our hypotheses required inclusion of NS, HA, and the three character scales (ST, SD, and C) from the TCI. In addition, demographic variables (age, sex, ethnicity, and education) were included as covariates. We tested the relationship of ADHD with mindfulness by using hierarchical regression analysis entering demographic variables in step 1, followed by ADHD in step 2, and the five TCI scales in step 3. This methodology allows us to assess the percent variance in mindfulness scores accounted by each set of variables and to compare the relative change between steps. Various diagnostics were conducted to verify that the assumptions of hierarchical regression were not violated. Specifically, none of the variables in the full model exceeded a variance inflation factor of 2, the Shapiro-Wilk W test for normality and quantile-quantile plot confirmed normality of the residuals, absence of heteroscedacity was confirmed by examining a residual plot, and a linear relationship between the main effects and the outcome were verified through scatter plots.
As shown in , demographic variables account for little of the variance in mindfulness (7%) in contrast, ADHD is strongly associated with mindfulness (presence of ADHD=lower mindfulness) increasing the variance accounted for from 7% to 23%. As can be seen in Step 3, individual differences in trait mindfulness are also influenced by personality as reflected by an increase in percent variance from 23% (Step 2) to 51% (Step 3). Overall, 51% of the variability in mindfulness could be accounted for by differences in demographics (education), ADHD, and specific facets of personality; the three most influential variables (as indexed by the magnitude of standardized betas) are ADHD, SD, and ST.
Summary of hierarchical regression analysis of mindfulness score (KIMS sum) on ADHD and selected personality variables (n=105)
In an ad hoc analysis, we explored the same set of variables shown in Step 3 with each of the four KIMS subscales to determine if any one scale was differentially affected by ADHD. In those analyses, it was clear that ADHD was significantly associated with the subscale ‘acting in awareness’ (standardized beta = −.48, p<.01) and not with the other three scales (Accepting, Describing, and Observing had standardized betas of −.07, −.21, and −.14, respectively; all were non-significant in the model). To further explore the relationship of ADHD and the ‘acting in awareness’ scale, we applied component analysis to the 18 ADHD symptom items and the 10 ‘acting in awareness’ items after scaling them to similar 0, 1, and 2 response categories (by collapsing the two categories ‘never’ and ‘rarely true’ on the KIMS into one group). The first five eigenvalues (9.3, 2.6, 1.7, 1.5, and 1.2) suggested the presence of a relatively large general factor but a single component extraction revealed that the ADHD items loaded more highly on the factor while the KIMS subscale items had low loadings. Using a two component solution with a Promax rotation, the 18 ADHD items loaded on a first factor while six KIMS items loaded on a second factor. Four KIMS items (items #3 – I am easily distracted, #11 – I drive on autopilot, 15 - When reading, I focus all my attention on what I’m reading (reverse scored), and #23 – I don’t pay attention to what I’m doing) cross-loaded on both factors. The correlation among the Promax rotated components is 0.33. These results suggest that the two questionnaires measure two correlated constructs with their overlap largely reflected by a subset of items that measure attention.