Mechanisms of Contextual Risk for Adolescent Self-Injury: Invalidation and Conflict Escalation in Mother–Child Interactions

Abstract

According to developmental theories of self-injury, both child characteristics and environmental contexts shape and maintain problematic behaviors. Although progress has been made toward identifying biological vulnerabilities to self-injury, mechanisms underlying psychosocial risk have received less attention. In the present study, we compared self-injuring adolescents (n = 17) with typical controls (n = 20) during a mother–child conflict discussion. Dyadic interactions were coded using both global and microanalytic systems, allowing for a highly detailed characterization of mother–child interactions. We also assessed resting state psychophysiological regulation, as indexed by respiratory sinus arrhythmia (RSA). Global coding revealed that maternal invalidation was associated with adolescent anger. Furthermore, maternal invalidation and coerciveness were both related to adolescent opposition/defiance. Results from the microanalytic system indicated that self-injuring dyads were more likely to escalate conflict, suggesting a potential mechanism through which emotion dysregulation is shaped and maintained over time. Finally, mother and teen aversiveness interacted to predict adolescent resting RSA. Low-aversive teens with highly aversive mothers had the highest RSA, whereas teens in high–high dyads showed the lowest RSA. These findings are consistent with theories that emotion invalidation and conflict escalation are possible contextual risk factors for self-injury.

Acknowledgments

This work was supported by grants from the Pediatric Clinical Research Center at Seattle Children's Hospital, M01-RR 00037, and the American Foundation for Suicide Prevention to Cindy J. Smith, and by grants from the National Institute of Mental Health to Theodore P. Beauchaine, R01 MH63699 and to Sheila E. Crowell, F31 MH074196. We express thanks to Jennifer Gross, Barbara Kleine, Dana Kovalchick, Eileen Magill, Elise Mallman, and Andrea Moore.

Notes

¹This sample includes participants described previously by Crowell et al. (2005) and Crowell et al. (2008), where more detailed descriptive statistics are provided. Although certain results from global coding were reported in the latter publication, none of the prior analyses included the tests presented here. Results from the microanalytic coding system have not been reported elsewhere.

Note: T scores for the Achenbach scales are reported as mean (standard deviation).

^a Self, parent-, and teacher-reports on the Achenbach scales were assessed using the Youth Self-Report (Achenbach, 1991a), the Child Behavior Checklist (Achenbach, 1991b), and the Teacher Report Form (Achenbach, 1991c), respectively.

^b n = 16.

^c n = 20.

^d n = 16.

^e n = 20.

^f n = 9.

^g n = 15.

^a Number (and percentage) of self-injuring participants reporting at least one incident of this type.

Note: Robust standard errors are reported.

***p < .001.

Note: Robust standard errors are reported.

***p < .001.

²This hypothesis was also tested by including all three forms of unconditional aversive utterances in the same model. In this model, low-aversive utterances were used as the reference group, and the differences between low-, intermediate-, and high-aversive utterances was tested by including dummy-coded variables to indicate intermediate- and high-aversive utterances. The magnitude, significance, and direction of all effects were substantively identical. We therefore report results from the separate models to ease presentation and interpretation.

Note: Robust standard errors are reported.

*p < .05. **p < .01. ***p < .001.

³Similar to the model testing unconditional behaviors, this hypothesis was tested by including all conditional sequences in the same model. In this model, the low–low sequence was used as the reference group and the differences between low–low and all other sequences were tested by including dummy coded variables. The magnitude, significance, and direction of all effects were substantively identical. We therefore report results from the separate models to ease presentation and interpretation.

*p < .05. **p < .01. ***p < .001.

⁴This combination of codes was selected based on the significant associations between these variables, just reported, and to maximize the similarity of the global and microanalytic codes examined for these models.