Zooming in on the transmission of adults’ unresolved loss/abuse to infant disorganised attachment: A latent class analysis

Background
Infant disorganised attachment is predicted by parents’ unresolved states of mind about loss or abuse, which are identified based on lapses in the monitoring of reasoning, discourse, and behaviour surrounding loss/abuse in the Adult Attachment Interview (AAI; George et al., 1984, 1995, 1996; Main & Hesse, 1990). Recent meta-analytic results indicate an effect size of r = .21 for this association (Verhage et al., 2016). Research has assumed that unresolved states of mind are a unitary construct despite the fact that unresolved loss/abuse is identified based on a large variety of indicators. This study attempted to gain deeper insight into unresolved states of mind and the transmission to infant disorganisation. The following research questions were investigated: 1) Can groups of interviewees be identified who show similar profiles of indicators of unresolved loss/abuse (i.e., lapses in reasoning/discourse/behaviour)? 2) How are these profiles associated with unresolved states of mind and infant disorganised attachment?

Methods
Data included 999 caregiver-child dyads within 11 studies in the Collaboration on Attachment Transmission Synthesis. AAI data were available from 999 dyads and infant disorganisation (measured in the Strange Situation; Ainsworth et al., 1978) was available from 747 dyads. Latent class analysis (LCA) was performed on subsamples of adults with applicable loss (n = 872) and abuse (n = 238). Indicators of unresolved loss/abuse were dichotomised as coder ratings were inconsistently available (present/not present). Latent class probabilities were used to predict unresolved score (continuous: 1-9) and infant disorganisation (continuous: 1-9), using multilevel linear regression. This study was pre-registered on OSF (https://osf.io/w65fj/).

Results
Table 1 shows preliminary results from the LCA. For both unresolved loss/abuse, the two-class solution was the best fit. As seen in Figure 1, latent profiles of unresolved loss were distinguished based on the probability of showing six specific indicators. The latent profiles of unresolved abuse were distinguished based on the probability of showing disorganised speech or unsuccessful denial of abuse. In the total sample (N = 999), higher unresolved scores were significantly associated with infant disorganisation (B = 0.08, SE = 0.04, p = .035). The latent class probabilities for unresolved loss and abuse were not significantly associated with unresolved score and infant disorganisation (ps > 0.135).

Conclusions
The findings do not indicate distinct subtypes of unresolved loss. Rather, they suggest high- and low-lapsing subgroups. For unresolved abuse, findings indicate two subgroups, one characterised by a high probability of showing disoriented speech and one characterised by a high probability of showing unsuccessful denial of abuse. The indicators distinguishing the subgroups of unresolved loss/abuse were among the most frequently occurring lapses. It may be possible that these are particularly important for identifying unresolved loss/abuse. However, further scrutiny is needed, given that no information was available regarding the strength and frequency of occurrence of indicators within interview transcripts. The latent profiles of unresolved loss/abuse were not associated with infant disorganised attachment. As a next step, statistical learning techniques will be used to predict infant disorganisation based on specific indicators of unresolved loss/abuse.