The study showed that parents of children with ADHD are more hostile and critical towards their children than parents of healthy controls. These results were significantly influenced by childrens' OPB, since oppositionality ratings entered as a covariate removed this significant effect. This confirms Taylor's  hypothesis that comorbid OPB contributes to the emergence of high EE in parents of ADHD. Indeed, in an earlier study, mothers of ADHD children who had responded positively to methylphenidate treatment, their expression of warmth increased and expressed criticism towards their child decreased . Further, in a recent study by Sonuga-Barke et al. , ADHD children whose parents were highly critical and lacked warmth had significantly higher levels of ODD and CD. This study also found gene-environment interactions involving EE and polymorphisms of the serotonin and dopamine transporters suggesting that maternal expressions of warmth and hostility may act together with genetic factors in altering the severity of ADHD .
Contrary to our expectations, high criticism did not predict high oppositionality ratings but high warmth predicted low scores instead. This result was specific for the ADHD group and highlights the potentially protective role of parental warmth. For instance, Tully et al.  found that enhanced maternal warmth was predictive for low parent and teacher ADHD ratings, and Caspi et al.  showed that the monozygotic twin who received more maternal negativity and less warmth had more antisocial behavior problems compared to the twin receiving more warmth.
There were no significant differences in baseline cortisol between cases and controls. We were not able to confirm our hypothesis that ADHD children would show lower cortisol levels in response to a psychosocial stressor. This decrease of cortisol levels over the assessment period was absent in the ADHD children (negative condition). Control children and ADHD children in the positive control condition on the other hand did show a decrease in cortisol over time as did control children in the negative condition. This is in line with the findings by Hooley et al.  who only found effects of maternal criticism on remitted depressed patients. The majority of studies on ADHD children with comorbid OPB found decreases in cortisol levels [20, 55], though these results are not entirely consistent. There were only six children in our study with Conner oppositional ratings below T-scores of 60 in both parent and teacher questionnaires, thus the majority of children in our study presented comorbid OPB within the clinical range. Severe oppositional problems predicted weaker adrenocortical reactivity for children with the predominantly inattentive and hyperactive, but not for the combined ADHD subtype . Combined subtype ratings on the Conners' parent and teacher scales were overall high in our sample which might explain the discrepancy in findings. Also stress tasks in the studies mentioned were not based on parental EE. It might be that parental criticism is a highly salient psychosocial stressor for subjects with psychiatric disorders. The participants in the study by Hooley et al.  were fully remitted for a minimum of five months (range five to 109 months) and still reacted strongly to maternal criticism. There is also evidence that early disruptions in the parent-child-relationship, as is often found in families with ADHD children, produces increased cortisol levels in preschoolers. Further these heightened levels of cortisol predict increased behavioral and emotional problems in the school-aged child .
Family adversity, which can accompany ADHD, is related to higher and less-regulated cortisol activity in school-aged children and adolescents [57, 58]. Children with behavioral problems have shown an increased cortisol response during a parent-child conflict-discussion task and were associated with dysfunctional parenting . Typically, children with the largest increases in cortisol were rated as less capable of regulating negative emotions and aggression [60, 61]. An impairment in the regulation of emotion, as part of the core symptom of impulsivity, is often observed in children with ADHD . Effects of emotional states on saliva cortisol responses could also be demonstrated for situations encountered in daily life [63, 64]. In both studies, emotions of loneliness, sadness, and anger, as assessed with diary reports, were associated with specific cortisol patterns. Depressive symptoms led to a decrease of basal cortisol levels and higher cortisol levels were observed for moments adolescents were alone rather than with others, and for trait anger. Responses to the negative condition by ADHD children suggest an emotional stress response maybe similar to feelings of anger/annoyance. ADHD children did perceive the criticism expressed by their parents, but this was independent of the experimental condition, since no significant correlations between the final perceived criticism rating and condition could be obtained (see table 3) for ADHD children. Thus, apparently, children with ADHD do not explicitly perceive an acute emotional stressor, but rather respond to it on a physiological level, whereas control children do show an explicit reaction to the priming condition (see table 3), but this is not reflected in a physiological response.
In agreement with our mediation hypothesis (see above), we demonstrated that positive EE were significantly linked to the psychosocial stress response which in turn attenuated the direct path between EE and oppositionality. Our results support the important role of positive/warm interactions between parent and child and its effect on oppositional behavior. Since we were able to show this for cases and controls, positive parenting might be a generally protective parenting factor. However, it would be very interesting to replicate findings in a larger sample with separate analyses for ADHD cases and healthy controls. High psychosocial stress on the other hand, as shown experimentally, results in increases of the cortisol response and the tension felt. This in turn might lead to feelings of stress/anger that are expressed in oppositional behavior as described in behavioral models on ADHD and comorbid OPB . Since the emotion provocation task seems to be very useful in manipulating EE response, future studies should further explore the association of ADHD and OPB with respect to high EE.
Limitations to the study
The power analysis prior to testing showed that our sample lacked five participants for optimal sample size. A post-hoc power analysis with the given sample size and 5 repeated measures between 2 groups established a satisfying power of .94 with Lambda = 12.81, F(1,121) = 3.81, α = .05, f = 0.25. Thus interpretation of results should not be overly biased.
For ethical reasons, we were not able to obtain blood samples from the participating families. Due to this, we were unable to assess the critical ACTH component of the HPA-axis. We also did not measure other stress variables such as heart rate or skin conductance response that could have confirmed the cortisol stress responses.
Since children with ADHD significantly differed from healthy control children in the Conners' subscale anxious-shy, possible influences on cortisol response had to be taken into account, even though children with positive screenings of internalizing disorders were not included in the study. We collapsed the parent and teacher ratings of the Conners' subscale anxious-shy into one measure and entered this as a covariate in an ANOVA with baseline cortisol as the dependent variable. Anxiousness did not influence levels of baseline cortisol significantly (F(1,112) = 0.100, p .752), though assessment of comorbid disorders on HPA-axis should be addressed in future studies on this topic to further confirm this result.
The sample in this study consisted of males and females, though it was predominantly male, representing the four to one prevalence of males to females in populations at large [66–68]. Gender did not affect levels in cortisol in our study, but a replication of these results in a sample of ADHD girls is recommended.
The age range in this sample was fairly large. It is well known that ADHD symptoms change over time, and for children reaching adolescence a reduction in impulsive/hyperactive behavior is often found . But when entered as a covariate, age did not significantly influence our results.
Oppositional behavior was solely assessed with the Conners Rating Scales. A more thorough assessment of this comorbid condition, especially with respect to the differentiation between ODD and CD might contribute to more specific results.
EE were assessed with the short FMSS and not the original Camberwell Family Interview (CFI). Assessment of EE with the CFI takes approximately one to two hours followed by three to four hours rating. Assessment and rating of EE with the FMSS is much faster and correspondence between CFI-EE ratings and those from the FMSS have proven to be highly satisfactory , that the use of this measure seemed justified.
Positive and negative comments were not subjected to qualitative analysis. But since cortisol response of the ADHD children in the positive condition was so akin to that of the controls, we were able to show that ADHD children and controls respond similar to positive feedback and differently to negative feedback. The qualitative difference in negative feedback, that is highly likely, is reflected in the significant differences in high EE ratings by ADHD parents.
The cross-sectional study design did not allow us to determine the direction of effects. Future studies should report path models that show the structural relations between the variables analysed.
As shown in this and several other studies, parental high EE is a highly salient psychosocial stressor and proves to be relevant for comorbid OPB in ADHD patients. Approaches focusing on the reduction of parental hostility and the enhancement of warmth should be a target for therapeutic interventions to reduce or even prevent comorbid disorders such as OPB. Even though most behavior therapy programs include sessions on positive parent-child interactions their premises need to be further explored and might lead to an improvement of behavioral therapy elements. Furthermore, cortisol levels might serve as a relevant indicator for therapeutic success. Children with disruptive behaviors and low basal cortisol levels had a better response to intervention by parent training as assessed by changes in cortisol levels and disruptive behavior scores .