Latent constructs model explaining the attachment-linked variation in autobiographical remembering

Abstract

In the current study, we proposed a latent constructs model to characterise the qualitative aspects of autobiographical remembering and investigated the structural relations in the model that may vary across individuals. Primarily, we focused on the memories of romantic relationships and argued that attachment anxiety and avoidance would be reflected in the ways that individuals encode, rehearse, or remember autobiographical memories in close relationships. Participants reported two positive and two negative relationship-specific memories and rated the characteristics for each memory. As predicted, the basic memory model yielded appropriate fit, indicating that event characteristics (EC) predicted the frequency of rehearsal (RC) and phenomenology at retrieval (PC). When attachment variables were integrated, the model showed that rehearsal mediated the link between anxiety and PC, especially for negative memories. On the other hand, for avoidance EC was the key factor mediating the link between avoidance and RC, as well as PC. Findings were discussed with respect to autobiographical memory functions emphasising a systematically, integrated framework.

Keywords:

• Autobiographical memory
• Adult attachment
• Phenomenology
• Latent constructs

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

¹ In order to test the consistency of the measurement model across different memory types, additional analyses were conducted with the data collected in our laboratory. We used data coming from self-defining, emotional, goal-related memories, all of which were collected with an almost similar procedure. The factor structure as well as factor loadings were found to be invariant across memory types, ∆χ² < 20.090 (∆df = 8). Links between the indicators and the latent constructs were found to be significant with the exception of self-definition and importance for goal-related memories. In further analyses, chi-square difference for the model of invariant intercepts was found to be significant; however, the intercepts for the self-definition and importance were not constrained, partial invariance of the intercepts were established. Partial invariance of the intercepts ensures the equality of the comparable elements in the model (Byrne, Shavelson, & Muthén, 1989; Dimitrov, 2006) and for current purposes, model of equal factor loadings was considered sufficient to demonstrate the consistency of the latent construct across different memories.

² We further tested the model in which error variances were also constrained to be equal for both types of memories; however, at this level, the model diverged significantly from the scalar-invariant model, even when we considered the artificial inflations in chi-square due to the number of constraints. This indicated that observed variables, in other words, unique items representing memory qualities, were not measured identically across groups. Measurement invariance at this final step was expected and actually we were primarily concerned for the variation at this level so that we could examine the item-specific differences arising from the individual differences in levels of anxiety and avoidance.

³ Although the effect was moderately significant (p = .05), the effect was considered significant since the values corresponding to 95% confidence intervals did not involve a zero point (Cheung & Lau, 2008; Preacher & Hayes, 2008).