Although the ASD literature is replete with research about emotion recognition and findings are mixed (e.g., Jones et al., 2011
), relatively less is known about emotion dysregulation with the notable exception of anxiety (reviewed by White, Oswald, Ollendick, & Scahill, 2009
). This is curious as many youth with ASD who are referred for clinic or school evaluations experience intense emotional reactions (e.g., anger, irritability), and these reactions play an important role in referral for pharmacotherapy, isolation from peers, parental stress, and family distress. Nevertheless, their co-occurrence raises questions as to whether they are epiphenomena of the ASD clinical phenotype, clinical features of a distinct subtype of ASD, or pathogenically similar to phenomena in non-ASD samples. Evidence of an association between emotion dysregulation and ASD symptomatology in non-ASD referrals suggests shared risk factors, supports dimensional models of symptom behaviors (Barneveld et al., 2011
; Drabick, 2009
), and underscores the potential value of pursuing similar pathogenic models in youth with ASD. The results of the present study indicate many clinically referred, non-ASD youth with ODD (AIS and NS) exhibit elevated symptoms of ASD and SSD compared with Controls (Prediction 1), and this appears to be the case regardless of age, informant, or source-specific versus source-exclusive comparisons, with one notable exception (i.e., mother-rated communication deficits among younger youth, ). Consistent with the procedure for constructing ODD symptom groups, source-exclusive AIS differed from Controls (who according to both informants had few ODD symptoms) only when the same informant's ratings served as the basis for defining AIS group and ASD and SSD symptom severity ( and ). The extant literature and our reported findings indicate that the emotion dysregulation and interpersonal conflicts that define ODD, to include peer conflicts (Drabick & Gadow, 2012
; Gadow & Drabick, submitted for publication
), may share similarities with communication and social skills deficits of ASD and SSD; indeed, perhaps the most salient similarity is that social interactions are challenging and thus generate a range of intense emotional reactions. It is also possible that ASD and SSD represent divergent extremes of similar processes (Crespi & Badcock, 2008
; Russell-Smith, Maybery, & Bayliss, 2010
). Our expectation that source-specific symptom groups based on teachers' () versus mothers' () ratings would reveal more pronounced group differentiation (AIS, NS, Controls) was not supported (Prediction 4). Indeed, differences between AIS and NS groups were observed more frequently among mother-defined groups (AIS > NS for 7 of 10 comparisons; ) than teacher-defined groups (AIS > NS for 1 of 10 comparisons; ). Consistent with this pattern, there was a slightly greater number of group differences for mothers' () versus teachers' () ratings of source-exclusive groups compared to cross-informant groups.
4.1. Differential validity of AIS and NS (Prediction 2)
Consistent with our expectation that emotion dysregulation (i.e., anger, irritability) would share more commonality with ASD than NS, within-informant group comparisons indicated youth with source-specific AIS had more severe ASD and SSD symptoms than the source-specific NS groups, but this was the case only for mothers' ratings of younger children, mother's ratings of social deficits in older youth (), and teachers' ratings of SSD in younger children (). Nevertheless, these results provide additional support for the notion that AIS and NS represent divergent phenomena in non-ASD clinically referred youth, and extend this observation to ASD and SSD symptoms. Although the AIS and NS dichotomy is sometimes referred to as the affective and behavioral aspects of ODD, respectively (Burke & Loeber, 2010
), it is possible that these symptom groups pertain to different types of affect with unique neurobiologic substrates and phylogenetic histories (e.g., NS may be more associated with novelty-seeking/exploratory behaviors; Alcaro, Huber, & Panksepp, 2007
; Drabick & Gadow, 2012
4.2. Similarities between ASD and SSD (Prediction 3)
In general, both ASD and SSD symptoms evidenced a similar pattern of source-specific ( and ) and source-exclusive group differences ( and ), which was not unexpected as both symptom domains appear to be interrelated (Gadow & DeVincent, 2012
). We are mindful of the extraordinary conceptual issues surrounding the differential validity of ASD and SSD symptoms (see Starling & Dossetor, 2009
) and would simply add that some questions may not be resolvable with our current nosology. For example, a youth who is cognitively rigid and consequently encounters difficulties complying with authority figures and getting along with peers would exhibit at least one symptom of ASD, but would likely be classified as ODD in everyday clinical settings. Our current categorical nosology may not be well-equipped to address these issues (e.g., Crespi, 2010
; Meyer-Lindberg, 2010
; Panksepp, 2006
), including difficulty with differential diagnosis, informant perceptions, and use of multiple informants (Drabick, 2009
; Gadow & DeVincent, 2012
4.3. Behavioral variation and informant discrepancies (Prediction 5)
For decades, investigators have reported modest levels of agreement between different informants' ratings of child psychopathology, which is illustrated by the findings of an influential meta-analysis conducted by Achenbach, McConaughy, and Howell (1987)
who examined inter-rater correlations from 119 studies. They found the average
correlation between parents' and teachers' ratings of child behavior was low (r
= .27). Although there has been a long-standing tendency to dismiss the significance of informant differences as being “measurement error” or “methodological nuisances” (see De Los Reyes, 2011
), findings of studies conducted in multiple countries have found differences in the environmental, biological, and behavioral concomitants of cross-situational, source-specific, and source-exclusive ODD (e.g., Dirks et al., 2011
; Drabick & Gadow, 2012
; Drabick et al., 2007
; Gadow & Nolan, 2002
; Gadow, DeVincent, & Drabick, 2008
; Gadow, Chernoff, et al., 2010
; Munkvold, Lundervold, Lie, & Manger, 2009
; Offord et al., 1996
; Severa, Lorenzo-Seva, Cardo, Rodríguez-Fornells, & Burns, 2010
; Wood, Rijsdijk, Asherson, & Kuntsi, 2009
) and between mothers' and teachers' perceptions of therapeutic improvement in ODD symptoms consequent to intervention (e.g., Gadow, Nolan, Sverd, Sprafkin, & Schneider, 2008
). As we have discussed elsewhere (Gadow & Drabick, submitted for publication
), these source-exclusive groups differ in a wide range of background characteristics and school-functioning variables, some of which (e.g., parental discipline, failure to do school work) are likely behavioral antecedents of or triggers for intense emotional reactions among youth with AIS.
Variation in behavioral, physiological, and morphological characteristics (traits) in response to different environmental variables (phenotypic plasticity
) is a fundamental concept in evolutionary biology and a pervasive feature of life on this planet (Piersma & van Gils, 2010
) and plays an important role in human health and disease (e.g., Hochberg et al., 2011
). It can be either reversible or permanent and provides a conceptual model for understanding informant discrepancy. A child can behave very differently in different settings (intra-individual variation), and children vary in their ability to modulate their own behavior according to the demands of the situation (inter-individual variation). Reversible, intra-individual differences in behaviors modulated by environmental variables (phenotypic flexibility
or behavioral plasticity
) can be highly stable in specific environments and show little correlation with behaviors in different settings (context specific
) (e.g., Komers, 1997
; Piersma & Drent, 2003
; Wilson, 1998
). A parallel concept (developmental plasticity
) applies to irreversible, inter-individual variation in traits resulting from gene × environment interactions during developmental periods. In other words, developmental plasticity refers to an organism's ability to adjust its developmental trajectory in response to environmental cues. Both phenomena apply to child neurobehavioral syndromes and are probably best illustrated in the case of ADHD (which is highly co-morbid with both ODD and ASD (Gadow et al., 2006
; Gadow, DeVincent, & Drabick, 2008
)) simply because it is the most common child psychiatric disorder and consequently the most studied.
Many children with ADHD evidence dramatic within-individual changes in behavior depending on contextual features (e.g., task demands, novelty, level of structure, adult presence), illustrated by compliant behavior when in the physician's office (Sleator, 1982
). Behavioral plasticity among children with ADHD is also evident in response to different activities within the same school setting (e.g., Whalen et al., 1978
; Zentall & Zentall, 1975
). Moreover, their reactions to environmental variation evidence between-individual variation compared with typically developing peers without ADHD (e.g., Porrino et al., 1983
). Consistent with these observations, there is growing evidence suggesting that common genetic polymorphisms interact with environmental factors to influence within- and between-individual differences in behavioral plasticity (e.g., Bakermans-Kranenburg & vanIJzendoorn, 2007
; Belsky et al., 2009
; Dmitrieva, Chen, Greenberger, Ogunseitan, & Ding, 2010
; Martel et al., 2011
; Reiner & Spangler, 2010
Our research extends this line of inquiry to inter-individual variation within the ODD clinical phenotype by comparing children who were seemingly less behaviorally plastic in terms of emotion regulation (i.e., youth who were rated as having AIS according to both mothers and teachers) with peers who appeared to be more plastic (i.e., exhibited AIS according to only one informant) ( and ). Phenotypic flexibility in youth with ADHD (e.g., Marwit & Stenner, 1972
), ODD (e.g., Gadow & Drabick, submitted for publication
), or AIS is likely mediated at least in part by cognitive and other child-specific variables (e.g., common gene variants; Belsky et al., 2009
) that modulate reactions to task demands and contextual features. In summary, from our perspective, informant discrepancy is not only expected but also clinically relevant, and may be explained in part in terms of mechanisms that underlie behavioral plasticity.
4.4. Cross-informant syndromes (Prediction 6)
Our expectation that cross-informant AIS would be associated with more severe ASD and SSD symptoms than source-exclusive AIS was generally not supported with two notable exceptions: (a) teachers' ratings of communication deficits (but not SSD) symptoms among 6–11 year olds (AIS:T > AIS:M + T), and (b) mothers' ratings of the AIS:T group () and teachers' ratings of the ASI:M group () (i.e., cross-informant ratings of the source-exclusive groups.) The fact that the younger, teacher-exclusive group had more severe ASD symptoms than the cross-informant group suggests a more environmentally sensitive condition. As previously noted, one plausible explanation for the former seemingly counter-intuitive finding is that different pathogenic processes may be linked to the informant or the environment that serves as the informant's frame of reference (e.g., De Los Reyes & Kazdin, 2005
; Drabick et al., 2008
). For example, communication and social skills play an important role in successful school functioning, and it is therefore not unexpected that children in the teacher-exclusive AIS group would obtain higher scores for these symptoms than the AIS:M + T group. This is also consistent with our previous findings of a differentially higher rate of early language problems in these same children with AIS:T (see Gadow & Drabick, submitted for publication
4.5. Age (Prediction 7)
Consistent with our prediction that younger and older youth would exhibit a different pattern of group differences in SSD severity, informants indicated differences between source-specific AIS and NS groups in younger (AIS > NS > C; ) but not older (AIS,NS > C; ) youth. Interestingly, this same pattern of group differences was obtained for both mothers' and teachers' ratings. There were also age-related differences in findings for source-exclusive groups. Unlike source-specific groups, comparisons for each informant and each age cohort resulted in a unique outcome ( and ), though the cross-sectional design precludes our ability to draw conclusions about developmental differences related to age or informant.
4.6. Strengths and limitations
The present study has several strengths, including an operationalized, dimensional approach to parsing ODD symptoms, based on the recommendations for DSM-5
of an expert consensus panel that reviewed relevant data (Pardini et al., 2010
), and assessing ASD and SSD symptoms (Barneveld et al., 2011
); a large study sample; consideration of different age groups; and comparison of different informants. Nevertheless, many different biologic, cognitive, and environmental processes can lead to seemingly similar behavioral outcomes (Drabick, 2009
), well illustrated by deficits associated with ASD versus SSD (e.g., Crespi & Badcock, 2008
; Russell-Smith et al., 2010
), and these variables were not considered in the present study. Moreover, although relatively unstudied, it is likely that the social and emotional symptoms that define ODD are pathogenically heterogeneous, and much additional research will be required to understand how they relate to AIS, NS, ASD, and SSD. Given the preponderance of males in the AIS and NS subgroups, we were unable to consider whether patterns of findings differed based on gender. The fact secondary school teachers spend considerably less time with individual students than their colleagues in elementary schools likely impacts opportunities to observe intense emotional reactions by the former.
Information about SSD symptoms (and to some extent ASD symptoms in older youth) are often obtained from self-report (e.g., rating scales, clinical interviews) measures in clinical research, which in the case of SSD may be particularly important for the assessment of positive symptoms such as hallucinations and delusions. Moreover, there is generally poor agreement between parent and youth self-report of psychiatric symptoms (Achenbach et al., 1987
; Gadow, Chernoff, et al., 2010
). Nevertheless, this does not invalidate mothers' and teachers' ratings of SSD as potentially useful markers of phenotypic heterogeneity or their use in improving understanding of pathogenic processes. For example, poor insight correlates with SSD symptom severity among individuals with adolescent-onset psychosis (Parellada et al., 2009) and ASD is characterized by poor self awareness; therefore, comparisons between younger ASD and non-ASD samples may greatly benefit from the use of caregiver report. Nevertheless, generalization of this study's findings is bounded by assessment and informant considerations.
Age was an important variable in the pattern of diagnostic group differences and though clinically informative, a cross-sectional design cannot address developmental processes. It is reasonable to hypothesize that older youth may represent a somewhat different segment of the clinical population consisting of both early-onset cases with protracted difficulties refractory to caregiver efforts or environmental modifications, as well as youth with recent-onset disturbances modulated in part by different biological process. It would be important to learn whether co-occurring ASD or SSD symptoms in youth with AIS or NS are risk factors for later mental health concerns, and in the case of SSD, whether these relations are similar in ASD and nonASD samples.
Finally, we did not collect information about setting-specific (e.g., individual freedom to chose activity, peer or sibling behaviors, empathy, demands on working memory) or important child-specific (e.g., sensitivity to reward, planning abilities, emotion recognition) variables that may induce or exacerbate intense emotional responses or youth's ability to regulate their reactions. Additional research will be required to determine their role as potential mediators of ODD symptom mechanisms and ASD/SSD processes.
4.7. Clinical and research implications
Owing to the inevitable overlap among youth classified as source-specific versus source-exclusive, it was not possible to test which strategy was superior; however, a more reasonable objective is to determine under what circumstances a particular strategy may be more useful. We illustrate this situation with the following example: in the case of short-acting stimulant medication, it could be argued that if problem behaviors occur primarily in the home (or school), then the evaluation of treatment effects may be better served with parents' (or teachers') assessments, and treatment could be administered in such a way as to address the most problematic setting. Thus, the classification strategy resulting in the greatest degree of phenotypic homogeneity for the purposes of a specific application (e.g., intervention, determination of course) would likely be the most ideal. Unfortunately, in research settings, this may encourage consideration of data from one informant only, which consequently undermines the study of informant discrepancies.
A related issue pertains to qualitative differences in information obtained from different informants. For example, mothers observe child behavior in a much wider range of settings (within-child) than teachers, which is often interpreted to mean their observations are more clinically valid because they capture the youth's “true” behavior. Conversely, a case can be made for the advantages of teachers' ratings as the school provides a standard setting with a restricted range of environmental experiences with many peers and is therefore better able to illuminate individual differences (between child). Moreover, teachers share fewer similarities in genetic background and environmental experiences with their pupils than parents, which is important given the reciprocal nature of social interactions. In view of this complexity, it is truly impressive that any scientific progress is made in understanding emotional responses in children with the aid of conventional assessment instruments.
Proscriptions for action in clinical and research settings are conflicted. In clinical applications, obtaining information from multiple informants, particularly diagnostic evaluations and response to pharmacotherapy, has been advocated for several decades. In clinical research settings, things are a bit muddled. As previously noted, informant discrepancies are often considered a nuisance for many reasons (including an obstacle to publication success; De Los Reyes, 2011
). Nevertheless, we would encourage other investigators to examine informant discrepancies with the goal of generating a better understanding of etiology, response to intervention, and predictors of long-term outcome.
As previously discussed (Gadow et al., 2004
), informant discrepancy has considerable significance for the interpretation and conduct of clinical research because investigators have generally used the findings of structured interviews with the primary caregiver to construct clinical phenotypes, occasionally conducting similar interviews with the youth. In parallel fashion, many psychologists and school-based investigators have relied heavily on information obtained from the youth's teacher to address inter-individual differences. More recently researchers have incorporated information obtained from multiple informants (i.e., the “or rule”) to define clinical constructs. As the results of the present study indicate, however, these different strategies for defining clinical phenotypes may lead to very different conclusions about similarities and differences between diagnoses and inferences about the magnitude of therapeutic improvement.