Journal of Child and Family Studies

, Volume 18, Issue 4, pp 447–453

Sibling Conflict Resolution Skills: Assessment and Training


  • Brett W. Thomas
    • Mountain States Counseling & Psychological Services
    • Psychology DepartmentIdaho State University
Original Paper

DOI: 10.1007/s10826-008-9248-4

Cite this article as:
Thomas, B.W. & Roberts, M.W. J Child Fam Stud (2009) 18: 447. doi:10.1007/s10826-008-9248-4


Sibling conflict can rise to the level of a clinical problem. In Phase 1 a lengthy behavioral role-play analog sampling child reactions to normal sibling conflicts was successfully shortened. In Phase 2 normal children who lacked sibling conflict resolution skills were randomly assigned to a Training or Measurement Only condition. Training consisted of five clinic sessions focusing on verbal reasoning, assertiveness, and acceptance skills. Trained children outperformed Measurement Only children on the shortened role-play test. Parents of trained children, but not untrained children, perceived improved social competence in the home. The project provides an empirical foundation for future research with aggressive siblings.


ChildrenSiblingsAggressionSocial skills


A significant proportion of normal sibling interaction in the home setting has been observed to be aversive (Dunn 1983, 1988). Moreover, correlational analyses by the Patterson group have implicated the contribution of siblings to the development of coercive child behavior (Patterson 1984, 1986). Carter and Volkman (1992) and Schroeder and Gordon (1991) even suggest that sibling conflict should be recognized as a bonafide category of psychopathology. Therefore, the need to develop measurements and treatments of sibling conflict resolution skill deficits is apparent.

If parents report sibling aggression as a problem, simple parent collected home frequency counts of salient episodes of physical fighting can be readily obtained. The temporal consistency of parent coded sibling aggression frequency has been empirically demonstrated (Olson and Roberts 1987; Thomas 2002). Using home aggression frequency as an outcome measure, behavior therapists have designed treatment procedures to suppress sibling aggression. Combinations of immediate timeout, reinforcement for aggression inhibition, reinforcement for sibling cooperation, and social skill building have been consistently associated with reduced sibling aggression frequencies in the home (Allison and Allison 1971; Hawkins et al. 1966; Jones et al. 1992; O’Leary et al. 1967; Olson and Roberts 1987). Social skill training approaches have also been successfully implemented to improve the general relationships of normal siblings (Kramer and Radey 1997).

Aggression frequency is surely not a sufficient assessment strategy for sibling conflict. Developmentalists have documented the complexity of normal sibling interaction, most of which is prosocial. See Boer and Dunn’s (1992) text for reviews and a recent special edition of the Journal of Family Psychology (Kramer and Bank 2005) which frame the importance of the topic for child clinicians. Imitation, cooperative play, verbal justifications, affectionate actions, helping, teaching, and “mutual hilarity” are among the many prosocial sibling interactions that have been accurately measured (e.g., Brody et al. 1986; Dunn et al. 1990). Of central importance is the consistent finding that verbal justifications improve with development (e.g., Dunn and Munn 1987) and provide the basis for conflict repair in normal children (Putallaz and Gottman 1983). The Sibling Conflict Resolution Scale (SCRS), introduced by Roberts et al. (1992), was designed to evaluate prosocial methods of sibling conflict resolution at the micro-behavioral level. Clinicians need to know if verbal reasoning skills (“justifications” in the developmental literature), appropriate assertiveness, and acceptance of intractable sibling disagreement are within the repertoire of a clinic-referred, physically aggressive sibling dyad. If present, motivational interventions (e.g., immediate timeout, reinforcement for aggression inhibition) could be a sufficient intervention. If absent, skill building interventions would be indicated and used in combination with motivational strategies, since such children would lack the replacement skills for aggression.

The SCRS measures the overt reactions of 4- to 12-year old children to 33 scripted sibling conflicts. Conflicts and associated skills based on the content validity work of Arnold (1990) are presented in Table 1. Each conflict is described to the child by a narrator and presented by an actor holding a doll that represents the sibling. The narrator repeatedly prompts the child to resolve the conflict in a socially desirable manner “…with your best behavior; the way your mom, dad, and teacher want you to act toward your (brother/sister).” Responses are scored on a 5-point scale, where 1 = physical aggression, 2 = verbal coercion, 3 = neutral, 4 = verbal coping, and 5 = sophisticated verbal coping. Three samples of children have been administered the SCRS (Roberts et al. 1992, Samples 1 & 2; Thomas 2002, Sample 3). The SCRS total score has yielded acceptable test-retest reliability (r = .81, Sample 1; r = .64, Sample 2); child behavior can be coded accurately (r = .96, Sample 1; r = .92, Sample 2; r = .92, Sample 3); SCRS items are relatively homogenous (α = .84, Sample 1; α = .81, Sample 2). Total scores on the SCRS have covaried with child age (r = .69, Sample 1; r = .46, Sample 3), indicating that older children have acquired a better repertoire of prosocial reactions to sibling conflict than younger children. Thomas (2002) found that receptive language level and SCRS scores converged significantly, r = .33, supporting the association of language development and conflict resolution skill. It is essential to the criterion validity of the SCRS that its scores positively covary with indicators of normal developmental processes, such as age and language. The SCRS assumes that the complicated and subtle social discriminations required for successful resolution of sibling conflicts are acquired gradually during normal development across the pre-adolescent period (i.e., from 2- to 12-years of age). Thomas (2002) also found that sex of the child significantly incremented the prediction of SCRS skill beyond age, indicating that older female children were the most able to solve SCRS problems.
Table 1

Sibling conflict conditions and associated skillful resolutionsa


Skillful resolution(s)

Disagreement over shareable household item

Request sharing + give reason

Disagreement over non-shareable household item

Request turn-taking + give reason

Sibling noncompliance given:

Request/instruct + give reason, OR

    1. Request to use sib’s possession, OR

Make a “Deal”, OR

    2. Request to enter sib’s room, OR

Accept noncompliance

    3. Request to play, OR


    4. Instruction to perform adult-required work


Sib violates game rule

Verbal assertion + reason, OR

Warn pending discontinuation, OR


Sib teases verbally

Explore interest in playing, OR

Ignore, OR

Leave social context

Sib teases physically (“rude touch”)

Step back + verbal assertion + reason

Sib physically aggressive

Step back + verbal assertion + reason, OR

Leave social context

aBased on the work of Arnold (1990)

It was originally hypothesized that SCRS scores would be inversely correlated with sibling aggression (i.e., low SCRS scores would correlate with high sibling aggression frequencies). This hypothesis, however, was not supported for normal sibling dyads. The SCRS failed to correlate with parent-collected home frequencies of sibling coercion (r = −.138 and .028, in Samples 1 & 2 respectively). Repertoire levels of conflict resolution skills in normal sibling dyads appear to be independent of concurrent levels of social aggression in the home, possibly because of range restrictions (i.e., consistently low aggression rates in normal children) or motivational issues (i.e., normal children may inhibit sibling aggression despite a lack of conflict resolution skills, given effective parent management of child misbehavior in the homes of non-referred children). In summary, the SCRS appears to accurately measure a content valid set of sibling conflict resolution skills that are associated with age, linguistic level, and sex of the child. It is anticipated that the SCRS will eventually prove clinically useful for treatment selection tasks. Specifically, should an aggressive sibling dyad be treated with timeout for aggression, reinforcement for aggression inhibition, sibling conflict resolution skill building, or some combination of these protocols? To justify SCRS use with aggressive, clinic-referred sibling dyads, specific empirical foundations needed to be established. The current two-phased project was designed to provide that foundation.

In Phase 1 existing SCRS data were analyzed to determine if the measurement could be reduced in length without sacrificing needed psychometric properties. The prior versions of the SCRS (SCRS I and SCRS II, reported by Roberts et al. 1992) were too long (33–35 items) and too time-consuming (45–60 min) for clinical use. In Phase 2 normal children who did not display verbal reasoning to solve sibling conflicts on the shortened version of the SCRS were randomly assigned to a Training or Measurement Only condition to determine the effectiveness of a standardized skill building protocol to enhance sibling conflict skill repertoires.

Phase 1

Method and Results

An item analysis was performed on all available archival SCRS data. Three different samples of normal, recruited subjects totaling 95 children had completed the 33-item version of the SCRS (Roberts et al. 1992; Thomas 2002). Child age ranged from 4 to 12 years; children were recruited from parents attending Idaho State University or the community. The sample was predominantly European American and rural. The sample was randomly divided into two groups: an item-selection sample (n = 47) and a cross-validation sample (n = 48). Three item statistics were computed for the 47 subjects in the item-selection sample: item consistency, p-value, and item validity. First, only items that correlated .20 or higher with the total score were retained, yielding items that covaried with the SCRS total score (i.e., item consistency); second, items that were too easy (p >.90) or too hard (p <.10) were discarded, yielding items that produced variance among the participants; and third, only items that correlated .20 or higher with age were retained, since SCRS scores should improve with development (i.e., item validity). The item selection process reduced the test length from 33 items to 13 items. Most of the discarded items were a result of poor sensitivity to child age (18 of 20). The item consistency statistic was then re-computed with the new 13-item test which resulted in the elimination of one additional item. Four items that met criteria on two of the three indices and provided needed content validity by covering Arnold’s original 13 conflict scenarios were added back to yield the revised, 16 item test, hereafter, SCRS III. To ensure that SCRS III did not degrade the psychometric qualities of the longer 33-item SCRS, data from the independent cross-validation sample were obtained for the 16 selected items. A total score based on the 16-item test was constructed and that total score was analyzed for observer accuracy, internal consistency, and test validity (i.e., the correlation with child age). The new 16-item SCRS III yielded satisfactory inter-rater agreement ratio (89%), internal consistency (α = .83), and convergent validity (r = .58) on the independent cross-validation sample. The SCRS III was then used pre- and post-training during Phase II of the current project.

Phase II



Twenty normal children (12 males and 8 females) between the ages of 4 and 8 years (M = 5.5 years; range = 4.0–8.8 years) participated in the project. Children were recruited from parents attending Idaho State University (n = 6) or parent employees of Girls and Boys Town in Omaha, Nebraska (n = 14). Each participating child had at least one sibling within 3-years of his/her age (M age gap = 1.5 years). Children were predominantly of European-American ethnicity (78%) from families in the middle of Hollingshead’s Two Factor Index of Social Position (Myers and Bean 1968). Children were randomly assigned to a Training condition (n = 10) or Measurement Only condition (n = 10) group. The two groups were balanced for age and sex of the child. Seven children whose average baseline item score on the SCRS III was greater than 3.8 were excluded from further participation, since they consistently displayed verbal reasoning solutions to sibling conflict scenarios. The institutional review boards of Idaho State University and Father Flanagan’s Boys Home approved this study. Informed consent was obtained prior to the first assessment.


During the first (“pre”) and seventh clinic sessions (“post”) all children were administered the SCRS III, while all parents completed the Child Behavior Checklist (CBCL) (Achenbach and Rescorla 2000) and the Home and Community Social Behavior Scales (HCSBS) (Lund and Merrell 2001). All SCRS III administrations were videotaped. Parents received class credit or $10 per hour for participating in these two measurement sessions, which averaged 35 min in duration. Children were awarded a prize upon completion of the SCRS III. During clinic sessions 2 through 6 children in the Training condition participated in sibling conflict resolution skill training with a parent, which averaged 7.3 weeks in duration (range = 5–11 weeks). Parents were not compensated for their child’s enrollment in the skills program, but children always received a prize upon completion of that session’s lesson. Children in the Measurement Only condition were not provided any specific experiences between the pre-post measurement sessions, a period that averaged 6.3 weeks (range = 4–11 weeks). The skills program was made available to Measurement Only subjects upon completion of the waiting period and re-administration of the SCRS III and two questionnaires.

All training sessions followed a basic three-step sequence. Step 1: the therapist and an actor (who manipulated dolls to represent the siblings) modeled a conflict (e.g., over toy possession) and a coercive reaction (e.g., a physical fight). The child earned tokens by answering two questions about the scene: 1. “What was the problem?”; Example answer = “argue over toy”; 2. “What did the doll do wrong?”; Example answer = “fight”. If the child did not answer correctly, the answer was modeled and the question repeated once. Step 2: the same conflict was repeated, but this time the doll modeled a skillful response. The child was asked one question: “What smart thing did the doll do?” (Example answer = “share”). The same token reinforcement system and correction procedure were used as in Step 1. Step 3: the child was instructed to role play the skillful reaction to the conflict with the doll. Token reinforcement was provided for skillful responses; an error or no-response resulted in a second model of the skillful response (i.e., the therapist said, “Not quite, watch again.”). This process was repeated until the child role-played correctly or three trials were completed, whichever occurred first. All training sessions were videotaped.

At the outset of training sessions 2, 3, and 4, each scenario from the previous session(s) was reviewed using the role-play format (Step 3) described above. Scenarios reviewed were presented in a random order with different toys to promote generalization. During the review session (i.e., Session 5), all 14 scenarios introduced during training sessions 1–4 were reviewed in random order and with different toys. A criterion of 85% was established for Session 5. If a child did not correctly role-play at least 12 of the 14 conflict scenarios, missed items were to be reviewed until at least one correct role play of 12 scenarios was observed.


Observer Reliability

Twenty-six percent (10 of 39 tests) of the available SCRS III videotapes were selected at random and coded by an independent observer. Note that one SCRS III videotape of a Training condition subject malfunctioned, reducing the number of available tapes from 40 to 39. The mean inter-rater agreement ratio (item vs. item comparisons) was 75.6% (121/160). The SCRS III total score inter-rater reliability coefficient was r = .88. SCRS III data used in the analyses reported below were coded by observers blind to pre/post status and condition status of 93% of the tapes. Three tapes were scored by group consensus during the process of training the two independent coders.

Twenty percent (10 of 50 sessions) of the videotaped training sessions were selected at random and coded by an independent observer. These data were compared to the therapist’s recordings of child behavior obtained during the sessions. The mean inter-rater agreement ratio was 92.7% (89 of 96) for child responses to therapist questions and 94.2% (113/120) for child responses to role-play tasks.

Procedural Fidelity

A total of 608 conflict scenarios were presented during the administration of the SCRS III (16 items × 2 administrations × 19 subjects). Ten items (1.6%) were found to lack the required number of prompts, precluding a score for that item; a narrator or actor error was present in an additional 12 items (2.0%); token reinforcement for participation was not provided during 3.3% (14 of 418) of the scheduled items; and 35 of the 114 (31%) reminders to “…show me your best behavior…” were omitted.

Throughout the first 4 training sessions, participating children correctly answered an average of 47.2 of the 56 verbal questions (84.2%, range = 33–56) within the 1 to 2 trials programmed for each query. Children correctly role-played an average of 26.5 of the 28 conflict scenarios (94.5%, range = 21–28) within the 1 to 3 trials programmed for each scenario. All children in the training condition met the review session criteria by correctly role-playing an average of 13.5 of the 14 review scenarios (96.4%; range = 12–14). The 10 videotaped training sessions reviewed by a second observer discovered that the therapist omitted a programmed query to the child on 5 of 96 opportunities (5.2%), while no programmed prompts were omitted during 120 role-play scenarios. Six errors of commission (additional role-play trials or un-programmed reminders) were observed during the same 120 role-play scenarios.

Statistical Analyses

Analyses of Variance were performed on the data from the SCRS III, the CBCL Externalizing Scale, the HCSBS Social Competence Scale, and the HCSBS Antisocial Behavior Scale. The 2 × 2 ANOVAs consisted of one between subject factor, Conditions (Training vs. Measurement Only), and one within subject factor, Measurement (Pre vs. Post). The SCRS III data yielded a significant main effect for Measurement, F (1,17) = 10.78, p < .01. This was qualified, however, by a significant Conditions by Measurement interaction, F(1,17) = 8.9, p < .01, which accounted for 34% of the variance (η2 = .34). Simple effects were not readily testable, given unequal cell n. Consequently, four t-tests were performed to compare the cell means. Training subjects significantly improved their SCRS performance from pre- to post-training, t(8) = 4.35, p < .01. Moreover, Training subjects scored higher than Measurement subjects at post-training, t(17) = 4.37, p < .001. The two groups were not significantly different at pre-training, nor did Measurement subjects display any significant changes from pre to post. The SCRS III means and standard deviations are reproduced in Table 2.
Table 2

Means and standard deviations of SCRS III and HCSBS social competence data across conditions and measurement














Measurement only







HCSBSb social competence








Measurement only







aAverage item responses; n = 9 for Training and n = 10 for Measurement conditions

bT-scores; n = 10 for both conditions

There were no effects of Condition, Measurement, or interaction effects on the CBCL Externalizing and HCSBS Antisocial data. Group means were clearly in the normal range at baseline. Specifically, average pre-training CBCL Externalizing T-scores were 49.9 (SD 6.9) for the Training subjects and 47.9 (SD 12.1) for the Measurement Only subjects. Similarly, the pre-training HCSBS Antisocial T-scores were 53.1 (SD = 6.5) for the Training subjects and 51.3 (SD = 9.6) for the Measurement Only subjects. In contrast, the HCSBS Social Competence Scale data yielded a significant main effect of Measurement, F (1,18) = 5.0, p < .05, which was qualified by a significant interaction, F (1,18) = 5.0, < .05. The interaction effect accounted for 21% of the variance (η2 = .21). Tests of simple effects revealed a significant increase in the HCSBS Social Competency means from pre- to post-training for the Training condition participants, F(1,18) = 10.1, < .01. The remaining three simple effects were insignificant. HCSBS Social Competence Scale means and standard deviations can be found in Table 2.


The SCRS was successfully shortened by selecting items via an analysis of archival data. The new standardized, behavioral role-play test (the SCRS III) was accurately coded in the current project and found to be sensitive to the five session training protocol. Specifically, normal children with deficient SCRS scores who were randomly assigned to the protocol improved their repertoire of verbal reasoning, assertiveness, and acceptance to solve sibling conflicts. Children in the Measurement Only condition did not change during an equivalent temporal period and were less skillful at the post training assessment, suggesting that the training protocol was functionally related to the improved skills. Trained children manifested more consistent verbal reasoning post training on the SCRS III, improving from an average item response of 3.3 (“neutral”) at pre-training to 4.2 (“verbal coping”) at post-training. Moreover, parents perceived improved social competence in the home for children in the Training condition, but not for those in the Measurement Only condition, suggesting that the training effects observed in the clinic on the SCRS III generalized to the home setting.

Confidence in the outcomes is bolstered by observer accuracy data. Independent coders blind to experimental condition agreed with each other at the event level and at the total score level for SCRS III child responses. Moreover, child responses to therapist questions and instructions to role-play during the five training sessions were also accurately coded by the therapist. Procedural fidelity of the SCRS III administration and treatment protocol adherence during the training sessions were quantified and appear to be satisfactory with one exception. The SCRS III examiner failed to provide verbal reminders to display the socially desirable reaction to the conflict on 31% of the scheduled opportunities. The impact of this error is unknown, but would operate to suppress training effects, if anything. Clearly, the omission did not artificially inflate the detected differences between groups.

Several project limitations are also apparent. It is unknown if the children actually used their improved reasoning, assertiveness, and acceptance skills in the home setting. Parents did perceive changes in the expected directions for trained children on the HCSBS Social Competence Scale, but the effect was small (see Table 2) and failed to discriminate between the two groups at post-training. Moreover, the correspondence between parental summary judgments on a questionnaire and overt social behavior evaluated by objective observers is not likely to be strong. In addition, parents whose children participated in five training sessions on skills to improve sibling interaction over the course of almost 2 months may have been biased to perceive improvements in social competence. Had a placebo control rather than wait-list control been used, the training effect on the HCSBS Social Competency data might have been negligible. It is also unknown if the sibling aggression probability or frequency actually declined. No such measurements were obtained in the current project, nor was change in sibling aggression frequency expected in normal recruited siblings, given low aggression base rates and the lack of correspondence between SCRS scores and aggression in normal children. These participants were clearly normal, as evidenced by the recruiting process, and the T-scores on the CBCL Externalizing Scale and the HCSBS Antisocial Scale. It is unknown if a clinical sample would have improved to the same degree. Finally, the sample size was small (n = 10 per condition). On the one hand, the presence of a significant treatment effect with a small sample is impressive, given limitations on statistical power; on the other hand, generalization of the results to a larger population of normal sibling dyads in the 4–12 year age range is a noteworthy concern.

Can we teach oppositional, defiant, and disruptive 4–12 year olds who present with sibling aggression the same skill set used in this project and will those acquisitions effectively contribute to the reduction of coercive behavior and the enhancement of social competence among the siblings in the home setting? Both questions remain unanswered, but are important goals for future research to address. The current project provides the empirical foundation for that research by supporting the psychometric properties of a short, content valid laboratory measurement of sibling conflict resolution skills and by demonstrating the efficacy of a training protocol designed specifically to teach those skills.


Part of these data were originally presented as posters at the 2002 annual meeting of the Association for the Advancement of Behavior Therapy, Reno NV (Thomas et al. 2002) and the 2006 annual meeting of the Association of Behavior and Cognitive Therapy, Chicago (Thomas and Roberts 2006). A copy of the SCRS III manual that includes all procedural details is available for the cost of reproduction from the senior author upon request (Mark Roberts, Ph.D., Psychology Department, Stop 8112, 921 South 8th Ave., Idaho State University, Pocatello, ID 83209-8112).

Copyright information

© Springer Science+Business Media, LLC 2008