Response Patterns on the Parent–Child Relationship Inventory in a Simulated Child Custody Evaluation

Abstract

Custody evaluation likely induces a motivation for parents to distort self-reports of parenting behavior. This study examined thesusceptibility of the Parent–Child Relationship Inventory (PCRI) to exaggeration of positive parenting in an instructional manipulation simulating a custody evaluation versus research study in 64 university students who were parents of children between the ages of 3 and 15. Results suggested that the simulated custody context influenced PCRI scores in the expected direction. However, the social desirability indicator built into the measure was effective at detecting the attempt to present an overly favorable image of the parent–child relationship.

KEYWORDS:

• forensic evaluation
• child custody
• parent child relations
• psychological assessment
• social desirability
• response bias
• test validity
• test reliability

Notes

Note. First three subscales are reported as t-scores; Social Desirability is reported as a raw score. Participants completed the measure first under one manipulation and then under the other during the second administration.

Note. Values above the diagonal are research; values below the diagonal are custody.

Correlations ≥ |.25| are significant at p = .05; correlations ≥ |.32| are significant at p = .05.

Note. Control model df = (5, 55). Full model df = (6, 54).

*significant at α = .05/3.

% Var = proportion of total variance in INV_C uniquely explained by each predictor.

Note. Control model df = (5, 55). Full model df = (6, 54).

*significant at α = .05/3.

% Var = proportion of total variance in COM_C uniquely explained by each predictor.

Note. Control model df = (5, 55). Full model df = (6, 54).

*significant at α = .05/3.

% Var = proportion of total variance in LIM_C uniquely explained by each predictor.

SAS GLM was used for the main analysis at alpha = .05/2. The test for flatness was significant (F (2, 61) = 14.26, p < .0001, semi-partial η² = .175); a follow-up t-test analysis revealed that at α = .05, COM (M = 49.77, CI = 47.56 – 51.97) differed from LIM (M = 56.22, CI = 53.81 – 58.63), but neither differed from INV (M = 53.95, CI = 51.60 – 56.31). Using Wilks’ criterion, the profiles (see Figure 1) did not deviate from parallelism, (F [2, 61] = 0.31, p = .74, partial η² = .005).