The Influence of Maternal Reflective Functioning and Expressed Emotion on Children’s Attachment among Children with or at Risk of Behavioural Problems

Abstract

Abstract

Background: This study examined whether levels of parental reflective function (RF), parental expressed emotion (EE) and children’s attachment styles are significantly related in a sample of children with high levels of conduct disorder (CD) symptoms.

Method: The sample (n = 143) consisted of children aged 5-7 years at risk of behaviour problems. Participants were recruited from a borough of London and a unitary authority in the south west of England. Data for the three main variables and confounders were collected using semi-structured interview, direct observation and questionnaires from both parents and children. The Parent Development Interview (PDI) was used to assess RF; the Five-Minute Speech Sample to assess EE and the Manchester Child Attachment Story Task (MCAST) to assess child attachment.

Results: Global levels of maternal RF did not significantly differ between the securely and insecurely attached groups of children. Mothers of securely attached children, however, had higher RF ratings on the negative interactions and anger subscales of the PDI compared to mothers of insecurely attached children. No significant difference was found in parental EE between secure and insecurely attached children. High EE-warmth was associated with high global RF, but there was no significant relationship between EE-criticism and RF. Multiple logistic regression found no significant relationships between parental RF, parental EE and child attachment.

Conclusions: These findings may suggest that attachment classification influences the levels of maternal RF in specific negative situations. Conversely it is possible that high maternal RF in such situations enables mothers to respond more sensitively to their child, leading to more secure attachment. The finding that maternal RF and EE were not associated with child attachment may suggest these variables are not strongly related, the sample is too small to detect any effect or that the specific sample lacks variability in scores. To the author’s knowledge this is the first study to test for these relationships with confounders included in the model, which may explain the null findings.