Mother-Child Reminiscing At-Risk: Maternal Attachment, Elaboration, and Child Autobiographical Memory Specificity

Abstract

Mother-child reminiscing, the process wherein mothers and their children discuss past events and emotional experiences, has been robustly linked with child outcomes, including autobiographical memory. To advance previous work linking elaborative maternal reminiscing with child autobiographical memory specificity, the ability to generate and retrieve specific memories from one’s past, it is essential to make distinctions among aspects of elaboration and to consider how maternal risk factors may influence the reminiscing context. The present study evaluated (1) an interaction among emotional and structural elaboration predicting child autobiographical memory specificity, and (2) the potential moderating role of maternal adult attachment. Participants consisted of 95 preschool-aged children and their mothers. The sample was predominantly low-income and racially diverse. Dyads completed a reminiscing task which was coded for emotional and structural elaboration. Mothers completed the ECR-R to assess attachment-related avoidance and anxiety, and children completed the AMT-PV to assess memory specificity. Results indicated that the association between structural reminiscing and child memory specificity was moderated by emotional elements of reminiscing. At high levels of emotional elaboration, mothers with high levels of structural elaboration had children with more specific memory than mothers with low levels of structural elaboration. Moreover, emotional elaboration (1) predicted less specific child memory without high structural support, and (2) negatively predicted child specificity at high levels of maternal attachment avoidance and anxiety, a profile associated with fearful avoidance. Future directions and implications are discussed.

Autobiographical memory (AM) refers to the recollection of personally experienced events and is determined by cognitive, affective, and social influences (Fivush, 2011; Reese, 2009). Autobiographical memory specificity is one aspect of AM which refers to the ability to recall specific memories of one-time personally experienced events (see Valentino, 2011; Williams et al., 2007 for reviews). Difficulty in retrieving specific memories in response to cue words is referred to as overgeneral memory (OGM), or reduced AM specificity. Rather than referencing one exact and precise event, overgeneral memories include categorical responses that refer to repeated events or memories that refer to extended periods of time (Williams et al., 2007). Reduced memory specificity (i.e., OGM) has been observed among individuals with depression and trauma-related psychopathology and has been associated with the emergence and clinical course of these disorders, although further work is needed to clarify the nature and direction of these associations (see Moore & Zoellner, 2007; Sumner, Griffith, & Mineka, 2010, for reviews). As such, identifying predictors of AM specificity in children is a critical research priority that may inform our understanding of processes that relate to the onset and maintenance of reduced memory specificity as well as normative AM retrieval. Moreover, it is particularly crucial to identify predictors of memory specificity during the preschool period, which is regarded as the time during which skills central to the ability to recall and express AM emerge (Nelson & Fivush, 2004).

Early theoretical models of OGM posit that reduced AM specificity results from the functional avoidance of negative memories, which then generalizes to all retrieval (Williams, 1996). Further updates to this model primarily focus on cognitive mechanisms such as executive functioning in accounting for the etiology of OGM (e.g., Williams et al., 2007). In contrast, the developmental psychopathology (DP) model of OGM (Valentino, 2011) expands on these accounts by applying sociocultural approaches to memory development to the understanding of OGM and AM specificity. This model emphasizes the integration of normative developmental research on memory with OGM, highlights the relevance of predictors of typically developing child AM (e.g., maternal elaborative reminiscing) for understanding reduced specificity, and acknowledges that variables may positively or negatively predict specificity, depending on moderating factors in the broader developmental context. To date, little work has considered factors at multiple ecological levels nor evaluated moderators of child memory specificity; doing so would provide critical empirical evidence to evaluate the DP model of OGM.

Mother-child reminiscing, the process wherein mothers and their children construct interactive dialogues to discuss past events (Fivush, Haden, & Reese, 2006), is identified in the DP model as a robust contextual predictor of preschool children’s AM, particularly in the context of reminiscing about children’s emotional experiences. The ways in which mothers discuss and elaborate on past events with their children socializes child AM by influencing how children recollect their past and cope with and understand their emotions (see Fivush et al., 2006; Nelson & Fivush, 2004; Valentino, 2011 for reviews). As such, the DP model hypothesizes that elaborative reminiscing will be a strong predictor of child memory specificity as well. Importantly, recent work has distinguished between emotional elaboration (e.g., how often mothers reference emotions and ask emotional questions) and structural elaboration during reminiscing (e.g., how often mothers provide new, non-emotional information; ask for contextual details such as where events took place) (Fivush & Vasudeva, 2002; Larkina & Bauer, 2010; Sales & Fivush, 2005). Structural elaboration was found to be unrelated to child AM specificity in our previous report with this sample (Valentino et al., 2014), which did not consider the quantity of maternal emotional elaboration. To extend and clarify this previous report, the current investigation more comprehensively evaluates the association between maternal reminiscing and child memory specificity by considering both structural and emotional elaboration.

Moreover, the nature of associations between elaboration and child memory specificity is likely influenced by the presence of maternal risk factors. Maternal adult attachment is one individual difference factor that has been consistently associated with child memory (see Alexander, Quas, & Goodman, 2002 for a review), and may be particularly relevant for AM specificity, which involves a child’s ability to respond to emotion cue words with specific memory exemplars. Maternal attachment, which involves a mother’s behavioral tendencies and interpersonal style within relationships, may likely affect a mother’s ability to socialize children’s recall of emotional memories during reminiscing. Thus, the current study investigates reminiscing in relation to child AM specificity, and further examines how these associations are affected by the context of maternal adult attachment.

Mother-Child Reminiscing

Mother-child reminiscing plays a critical role in the emergence of preschool-aged children’s AM, as the ways in which mothers engage in memory-sharing with their children predicts child remembering abilities and how children talk about their past to others (see Fivush et al., 2006, for review). Children of more highly elaborative mothers tend to construct and share more detailed, elaborated, and coherent accounts of their personal past during parent-child reminiscing conversations (Reese, Haden, Fivush, 1993), and experimentally training mothers in elaboration is associated with more sophisticated child memory (e.g., Van Bergen, Salmon, Dadds, & Allen, 2009). Mothers high on elaboration frequently incorporate questions and statements that add new information to the dialogue, provide evaluative feedback to their children, and encourage children’s interactive participation. In contrast, mothers low on elaboration tend to ask few or redundant questions that do not add new information to the dialogue nor properly scaffold their child’s participation.

Maternal elaboration may be either emotional (referencing and asking about specific emotions) or structural (providing and asking about non-emotional event details). Although some studies have demonstrated inter-correlations among structural and emotional elaboration (e.g., Larkina & Bauer, 2010), others have found structural and emotional elaboration to be unrelated (e.g., Bost et al., 2006; Fivush & Vasudeva, 2002). Thus, the emotional content of reminiscing may affect the way structural elaboration relates to child memory specificity. Highly structured discussions of emotional experiences may promote child AM specificity by facilitating children’s ability to understand and to jointly construct a coherent narrative of past emotional events, including discussion of causes, consequences, and resolution of emotions (Koren-Karie, Oppenheim, & Getzler-Yosef, 2008). In contrast, emotional elaboration without sufficient structural elaboration, particularly for the discussion of negative events, may negatively predict memory specificity, as raising emotional content without explanation of its causes or fluent discussion of resolution may activate children’s emotional arousal without the maternal support needed to understand the emotional experience. In fact, Valentino and colleagues (2014) found that maternal reminiscing quality, as defined by observational ratings of coherence and sensitive guidance, was positively related to child memory specificity. Importantly, reminiscing quality differs from and has been found to be unrelated to the frequency of structural and emotional elaboration (McDonnell et al., 2013). Therefore, the current investigation considers the quantity of both structural (non-emotional) and emotional elaboration in relation to child memory specificity.

Maternal Adult Attachment

Central to attachment theory is the principle that individuals form internal working models that include beliefs about the self in relation to others that influence how a person interprets and relates to their social world in adulthood (Cassidy, 2000). Adult romantic attachment is conceptualized as varying along two continuous dimensions of attachment insecurity: avoidance and anxiety (Fraley & Roisman, 2014; Hazan & Shaver, 1987; Shaver & Mikuliner, 2002). Importantly, maternal adult attachment has been consistently related to children’s memory and suggestibility for stressful information, although maternal attachment has not been examined in association with child AM specificity (see Alexander et al., 2002b; Bruck & Melnyk, 2004 for reviews).

Attachment avoidance is a deactivating strategy which refers to discomfort with and withdrawal from close relationships and is manifested as defensive distancing from dependency, emotion, or intimacy (Bartholomew, 1990). Attachment anxiety reflects hyper-activation of the attachment system, and refers to efforts to maximize or elicit closeness in personal relationships due to excessive fear of abandonment or rejection and hypervigilance to negative emotions (Fraley & Shaver, 2000). Importantly, individuals high on both avoidance and anxiety are said to possess a fearful-avoidant attachment style. Fearful avoidance has been associated with disorganized attachment and refers to an intense desire for closeness coupled with a negative, incoherent sense of self that manifests as avoidance of relationships due to fear of rejection (Bartholomew & Horowitz, 1991; Simpson & Rholes, 2002). Fearful avoidance has been most associated with disruptions in sense of self and emotions, as individuals with fearful avoidance have been shown to have higher levels of emotional maladjustment than individuals with other forms of insecure attachment (Anderson & Alexander, 1996; Carnelley, Pietromonaco, & Jaffe, 1994; Feeney, 1999; Murphy & Bates, 1997). Although previous work has examined maternal attachment within the contexts of mother-child interactions such as reminiscing (e.g., Fivush & Sales, 2006) and child memory independently (e.g., Alexander et al., 2010; Chae et al., 2014), no examinations have considered attachment, reminiscing, and child memory together.

Maternal Attachment, Mother-Child Reminiscing, and Child Memory

Parents with secure attachment styles tend to be more elaborative during reminiscing (Bost et al., 2006; Coppola, Ponzetti, & Vaughn, 2013; Reese, 2008). However, little existing research has considered dimensions of maternal attachment insecurity (avoidance and anxiety) in relation to reminiscing, as most examinations used dichotomous measures of attachment (i.e., secure versus insecure). Arguably, maternal attachment may influence how a mother socializes child AM during reminiscing, as avoidance or anxiety may affect a mother’s ability to construct memory narratives that facilitate remembering by organizing a child’s past emotional experiences into an autobiography.

Maternal attachment avoidance and anxiety may each negatively affect child memory specificity. First, original theories of OGM suggested that reduced specificity resulted from the functional avoidance of negative memories (Williams, 1996). Defensive avoidance is central to descriptions of adult attachment avoidance, which may be involved in the development of child OGM if mothers avoid discussions of emotion with their children (Valentino, 2011). Regarding attachment anxiety, Fivush and Sales (2006) found that maternal attachment anxiety was correlated with higher levels of elaboration during reminiscing of stressful events and with higher levels of child internalizing and externalizing symptoms (Fivush & Sales, 2006). These findings suggest that attachment anxiety may moderate associations between reminiscing and child outcomes.

Although no empirical work has examined how high anxiety and avoidance together influence the reminiscing discourse, mothers high on both avoidance and anxiety may have more difficulty engaging in elaborated and coherent emotional discussions than mothers high on avoidance or anxiety alone (Simpson & Rholes, 2002). Thus, maternal attachment avoidance and anxiety may together affect how a mother is able to engage in reminiscing that socializes children’s ability to recall specific emotional memories. Emotional elaboration may impede, rather than facilitate, children’s AM if mothers are unable to structure emotional discussions to generate a coherent narrative that involves the child in fluently discussing causes, consequences, and resolution of emotions. The current study addresses this gap by investigating both attachment avoidance and anxiety, and their interaction, as moderators of associations between reminiscing and child OGM.

Furthermore, the maternal attachment system may be activated or expressed differently across types of events being recalled and discussed. For example, Fivush and Sales (2006) demonstrated different patterns of relations among maternal attachment and reminiscing variables as a function of event type. In particular, attachment anxiety correlated with higher levels of maternal narrative engagement and exploration during reminiscing discussions of stressful (e.g., emergency room trips) but not conflict events. In addition, maternal attachment avoidance has been shown to negatively predict child memory in stressful contexts (e.g., Alexander et al., 2002a; Chae et al., 2014). Given that the attachment system is activated under conditions of stress (Bowlby, 1988), it is likely that associations between attachment and reminiscing in relation to child memory would be strongest in the context of stressful event discussions. Moreover, the attachment system is also activated by separation cues, although little work has considered reminiscing discussions of separation events, in particular (Bowlby, 1969). Therefore, the current investigation considers patterns of relations among maternal attachment, reminiscing, and child memory within multiple event categories.

The Current Study

In the present investigation, the following aims and hypotheses were examined:

1. Structural and emotional elaboration, and their interaction, were evaluated in relation to child specificity. We anticipated an interaction wherein emotional reminiscing would negatively predict child memory specificity at low levels of structural elaboration, and positively predict specificity at high structural elaboration.

2. Maternal adult attachment was evaluated as a moderator of associations between reminiscing and child AM specificity. We anticipated that reminiscing would negatively predict child memory among mothers high in attachment anxiety and among mothers high in attachment avoidance. Because profiles of fearful avoidance (high avoidance and high anxiety) have been conceptualized as most associated with maternal disorganized emotions and behavior, we anticipated that this profile would create a reminiscing context that confers the highest level of risk for child outcomes.

3. Associations between maternal attachment and child specificity were examined across event categories. We anticipated that associations between profiles of fearful avoidance (high avoidance and high anxiety) and child memory specificity would be strongest within discussions of stressful or separation experiences, as compared to happy or ongoing conflict events.

Method

Participants

Participants included 95 mother-child dyads who were recruited from the community in a medium-sized Midwestern city as part of a larger study examining mother-child reminiscing (see Valentino et al., 2014). The sample was racially diverse and predominantly low-income, with mothers’ ages ranging between 20 and 53 years (M=30.65, SD=7.30) and children’s (54.2% male, 45.8 % female) ages ranging between 3.5 and 6 years (M=5.55, SD=0.87). A higher risk sample reflecting a broad range of socio-demographic variability was recruited in order to assess a greater range of elaboration and attachment insecurity. Specific demographic information is provided in Table 1. All participants were primarily English-speaking.

Table 1.

Demographic Characteristics as a Percentage of the Sample

Characteristic	Percentage (n=95)
Maternal race/ethnicity
African American	60.6
Caucasian	21.3
Latino/Hispanic	9.6
Asian	1.1
Multiracial	7.4
Annual family income
Less than $6,000	26.1
$6,000 – $16,999	41.3
$17,000 – $28,999	21.7
Over $29,000	10.9
Maternal education level
≤ 11th grade	21.1
High school / GED	28.4
Full trade school	1.1
Some college	35.8
Associate’s degree	13.8
Maternal marital status
Single	53.2
Married	26.6
Cohabitating	7.4
Divorced	7.4
Separated	4.3
Widowed	1.1
Child race/ethnicity
African American	60.2
Caucasian	14.0
Latino/Hispanic	6.5
Multiracial	19.4

Procedure

Mothers and their children were recruited and brought into the lab as part of a larger study examining mother-child reminiscing (see Valentino et al., 2014). The subset of measures relevant for the present investigation are subsequently reported. After mothers provided documented informed consent, trained research staff administered individual assessment measures to mother and child separately from one another. The memory tasks were completed before other verbal parts of the assessment battery so that those tasks did not influence autobiographical memory retrieval. Halfway through the session, mother-child dyads participated in the joint reminiscing task. These interactions were video-taped, audio-recorded, and then transcribed verbatim by undergraduate research assistants.

Measures

Mother-child reminiscing.

Mothers and children were instructed to discuss four past events with each other. Before beginning the task, mothers chose four events to discuss with their child: (1) a happy event, (2) an acutely stressful event (e.g., visit to emergency room), (3) an event involving an ongoing conflict between mother and child, and (4) an event involving separation between mother and child (e.g., child going to daycare or staying with relative overnight) (e.g., Valentino et al., 2014). Negative event discussions were predominantly elicited because reminiscing about negative emotions has been shown to be more predictive of child functioning (e.g., Sales & Fivush, 2005) and more elaborative (Sales, Fivush, & Peterson, 2003). Mother-child pairs sat in a small assessment room and were instructed to discuss these events as if they were at home. All dyads discussed the happy event first, and then the order was counterbalanced. The reminiscing conversations usually lasted between 5 and 15 minutes.

Autobiographical Memory Test – Preschool Version (AMT-PV; Nuttall, Valentino, Comas, McNeill, & Stey, 2014).

The AMT-PV is a downward extension of the adult AMT that has been empirically validated as developmentally appropriate for use with young children from a range of socio-economic backgrounds. The AMT-PV consists of 10 cue words (5 positive and 5 negative) that are presented orally (read aloud by experimenter) and visually (word shown on paper) in a fixed order, alternating between positive and negative cues (i.e., happy, mad, surprised, sad, lucky, scared, strong, tired, smart, hungry). Trained experimenters instructed the participants to recall a specific memory in response to each cue word (e.g., ‘think of one time that you felt _____ and tell me about it’). The participant had 1 minute to generate a specific memory for each cue word, and was prompted for a specific memory as necessary (i.e., “can you tell me just one time when you felt that way”) within one minute by the experimenter, who was trained not to explain the cue word or give additional instructions to the child. The AMT is the most widely used paradigm for assessing memory specificity and overgenerality in the literature (Moore & Zoellner, 2007). The AMT-PV has been shown to be valid through item response theory and confirmatory factor analyses, which confirmed that the preschool version maintains the same underlying 1-factor structure as the original AMT (Nuttall et al., 2014).

Experiences in Close Relationships (ECR-R; Fraley, Waller, & Brennan, 2000).

The ECR-R is a self-report questionnaire designed to assess individual differences with respect to attachment-related avoidance (e.g., discomfort with close relationships) and attachment-related anxiety (e.g., fear of abandonment or rejection in close relationships) that was administered to mothers in the current study. A total of 36 items are assessed on a Likert scale ranging from 1 (strongly disagree) to 7 (strongly agree), with higher scores reflecting higher levels of attachment avoidance or anxiety. Both scales have been shown to be internally consistent, relatively stable over time, and convergent with related attachment measures including the Adult Attachment Interview (Shaver, Belsky, & Brennan, 2000; Shaver & Mikulincer, 2002). The ECR was selected for use in this study for two primary reasons. First, this measure ideally addresses the hypotheses of the current investigation because of its ability to capture a continuum of attachment-related avoidance and anxiety, as opposed to categorical measures of adult attachment (Mikulincer & Shaver, 2008). Second, use of this measure builds on previous work which has examined the ECR in relation to reminiscing (e.g., Fivush & Sales, 2006) and prior studies which have demonstrated associations between the ECR and child memory (e.g., Alexander et al., 2002a; Chae et al., 2014).

Peabody Picture Vocabulary Test (PPVT-4; Dunn & Dunn, 2007).

The PPVT-4 was administered in standard fashion to mothers in the current investigation to assess receptive vocabulary skills.

Expressive Vocabulary Test (EVT-2; K. Williams, 2007).

The EVT-2 was administered to children in the current investigation to assess expressive language abilities.

Coding and Reliability

Mother-child reminiscing.

Reminiscing conversations were transcribed and coded for quantity of (1) structural elaboration and (2) emotional elaboration. Structural and emotional reminiscing have previously been previously distinguished in the literature (e.g., Bost et al., 2006) and similar measures have been referred to as elaboration versus emotional integration (Fivush & Vasudeva, 2002), and factual/cognitive versus emotional elaborations (Sales & Fivush, 2005). Both measures of maternal elaborative reminiscing were coded using a frequency-based scheme in which each maternal utterance (subject-verb preposition) is coded for elaborative content (see Fivush & Sales, 2006; Fivush & Vasudeva, 2002; Van Bergen et al., 2009 for related schemes). Structural elaboration was calculated as the mean frequency of Wh-questions (open-ended elaborative questions, i.e. “who”, “what”, “where”, “why”, or “how”), Y/N questions (closed-ended ‘yes or no’ elaborative questions), elaborative statements (provide the child with new information about the event), and confirmations (maternal affirmations of child contributions to the conversation) averaged across all four event discussions. Emotional elaboration was calculated as the mean frequency of emotion questions (ie., “how did you feel?”), positive emotion attributions (ie., “you felt happy”), and negative emotion attributions (ie., “you felt scared”) averaged across all four event conditions.

Coding of emotional and structural elaboration was mutually exclusive, except in the case of questions such as “how did you feel?”, which were coded to indicate the presence of both an open-ended (structural) question and an emotion question in order to be able to examine the extent to which a mother structured an emotion inquiry with wh-question language. To maintain the independence of structural and emotional elaboration, an adjusted Wh-question variable was computed by subtracting the total number of emotion questions from the total number of wh-questions. This variable was then used to compute an adjusted structural elaboration variable in the same manner described above. Emotion explanations (statements that explain how the mother knew the child felt certain ways) and causes (statements referencing a reason why the child felt an emotion) were also coded, but were observed with such low frequency that they could not be included in analyses. The structural (α = .75) and emotional (α = .83) elaboration composite variables demonstrated high internal consistency and were calculated in a manner consistent with prior work (e.g., Valentino et al., 2014). Additionally, structural and emotional reminiscing variables were computed in the same way within each of the four event discussions (happy, stressful, conflict, separation). Twenty percent of the reminiscing transcripts were double coded by two undergraduate researchers and intraclass correlation coefficients ranged from .94 to .99.

AMT-PV.

Children’s responses to the AMT-PV were transcribed and coded for the number of specific memories. The child’s first memory response to each cue word (prior to any experimenter prompting) was coded for memory specificity, with a score of 0 defined as overgeneral and a score of 1 as specific. These responses were summed to create the total AMT variable of memory specificity. Specific memories were defined as references to events that lasted for a day or less. Overgeneral memories were further classified as categorical (reference to repeated events), extended (reference to events that lasted for longer periods of time), or ‘no response’ (failure to recall a memory). Substantive analyses were conducted based on the number of specific memories as is standard and recommended in the extant literature (Moore & Zoellner, 2007; Nuttall et al., 2014; Valentino et al., 2014). Further, specificity was utilized as the dependent variable because the AMT-PV was validated using the number of specific memories, and due to the lack of clarity, especially with the preschool age range, regarding how to categorize non-responses (e.g., as overgeneral or not) (Griffith et al., 2012). In addition, child responses after experimenter prompting were also coded for specificity to assess whether children were ever able to generate a specific memory exemplar for each cue word. A total specificity score after prompting was constructed to ensure that results held even after children received scaffolding from the interviewer to prompt for additional details. Consideration of specificity after prompting thus minimizes scoring concerns by addressing the potential confound of young children having difficulty remembering task instructions or responding to our memory protocol due to language and/or narrative difficulties. Reliability analyses revealed high inter-rater reliability on 20% of the sample between two coders who were unaware of the study’s hypotheses, α = 0.96.

Results

Descriptive Data

Mothers and children were excluded from analysis if language scores assessed during the laboratory visit were more than two standard deviations below the mean of the standardization sample because the reminiscing task relied heavily on verbal responding. Overall, 102 participants completed the study, but 7 were excluded because of these factors. Participants who were excluded did not significantly differ from those retained on primary demographic variables, including maternal age, education, race or ethnicity, current employment status, total family income, and current marital status. However, excluded mothers had lower levels of structural reminiscing compared to the entire sample, suggesting that they had significant difficulty engaging in this verbal task, t(101) = −3.8, p < .01. No group differences were observed for the remaining study variables, including child memory specificity and maternal attachment avoidance and anxiety.

All variables were examined for normality; no significant outliers were detected and variables were square-root transformed in the case of significant skewness or kurtosis. Observed descriptive statistics regarding maternal elaborative reminiscing, maternal attachment, and child AM specificity are presented in Table 2. Descriptive statistics regarding the proportion of AMT responses across items are presented in Table 3; on average, 47% of child memory responses were specific. Given the large percentage of ‘no memory’ responses, a proportion score was created to assess the percentage of specific responses a child provided out of all memory responses they contributed (i.e., number of specific memories / (10 – number of memory omissions). The pattern of results did not change using this proportion score as the dependent variable, and all findings remained significant. Moreover, results of substantive analyses did not change using the child total specificity score after prompting, which minimizes the potential confound of young children having difficulty responding to the task by showing that all findings held even after children received additional prompting. As such, results are reported with the number of specific memories (prior to experimenter prompting) utilized as the dependent variable to be consistent with prior work, the majority of which focuses on number of specific responses because the meaning of the ‘no memory’ response remains unclear (Griffith et al., 2012; Moore & Zoellner, 2007).

Table 2.

Descriptive Statistics and Correlations Among Primary Study Variables

Measure	M	SD	N	1.	2.	3.	4.	5.	6.	7.	8.	9.
1. Attachment avoidance	3.09	1.25	95	---
2. Attachment anxiety	2.52	1.15	95	.44**	---
3. Structural elaboration	5.43	3.32	93	−.03	.04	---
4. Emotional elaboration	3.65	1.85	93	.06	.08	.45*	---
5. Child memory specificity	4.70	2.98	90	.05	−.11	.07	−.19+	---
6. Child EVT	94.15	11.16	93	−.17	−.14	.04	−.08	.39**	---
7. Child age	5.49	.87	95	−.08	−.05	.17	−.10	.35**	.22*	---
8. Maternal PPVT	88.80	10.10	95	−.16	−.03	.13	.11	.03	.32**	.04	---
9. Maternal age	30.65	7.30	94	−.05	.08	.22*	.03	.09	.09	.48**	.00	---

Note.

^*p < .05

^**p < .01

⁺p = .07.

EVT = Expressive Vocabulary Test; PPVT = Peabody Picture Vocabulary Test. Structural and emotional elaboration variables have been square-root transformed to account for significant skewness.

Table 3.

Proportions of Child AMT Responses

Item	Specific	Categorical	Extended	No Response
Happy	.49	.10	.04	.37
Mad	.55	.14	.03	.28
Surprised	.62	.07	.01	.30
Sad	.51	.19	---	.30
Lucky	.45	.11	.02	.42
Scared	.49	.24	.01	.26
Strong	.36	.21	---	.44
Tired	.40	.31	.01	.28
Smart	.43	.18	.05	.34
Hungry	.44	.21	.04	.32
Average	.47	.18	.02	.33

Overall, the mean frequency of maternal elaborations averaged across all four event discussions was consistent with prior reminiscing research (e.g., Reese & Newcombe, 2007). There were no significant differences in maternal reminiscing with boys versus girls. No significant gender differences were found for memory specificity. Moreover, the overall level of structural and emotional elaboration was not significantly related to the order of event reminiscing. All of the reminiscing variables were significantly positively skewed; square-root transformations significantly reduced skewness and thus were used in analyses. The pattern of results did not differ depending on whether the adjusted or unadjusted Wh-question variable was incorporated into the broader structural elaboration composite. Therefore, results using the unadjusted structural elaboration variable are reported for all analyses to maintain consistency with prior work (e.g. Valentino et al., 2014).

Correlations among the primary study variables are presented in Table 2. As expected, attachment avoidance and attachment anxiety were significantly positively correlated with one another, which is consistent with recent estimates of their inter-correlation (e.g., r(91) = .43, p < .001 in Chae et al., 2014). This reflects conceptualizations of adult attachment as varying along continuous dimensions of avoidance and anxiety, wherein profiles of fearful avoidance are distinguished by high levels of both dimensions (Bartholomew, 1990; Simpson & Rholes, 2002). Moreover, the primary reminiscing variables were significantly inter-correlated. Child memory was significantly positively correlated with child age and child expressive vocabulary; thus, these variables were considered as covariates in all substantive analyses. Maternal age was positively correlated with structural elaboration, but was unrelated to child memory specificity. Inclusion of maternal age in all models did not change the results, and therefore maternal age was not considered as a covariate in substantive analyses.

Main Analyses

Substantive analyses were conducted in Mplus (Mplus version 7.1; Muthen & Muthen, 1998–2009) using full information maximum likelihood estimation to handle missing data. To test our first hypothesis regarding the association between structural and emotional elaboration in relation to child AM specificity, a moderated regression framework was utilized (Aiken & West, 1991). In addition to structural and emotional elaboration and their interaction, child age and language were included as continuous covariates given their association with child specificity. All predictors and covariates were grand-mean centered, and the model was saturated. Results indicated that elaborative emotional reminiscing negatively predicted specificity, although this was qualified by a significant interaction between structural and emotional elaboration, β = 0.23, p < .05 (see Table 4). This model explained a significant proportion of variance in child memory specificity, R² =.28, p < .01.

Table 4.

Interaction between Emotional and Structural Elaboration Predicting Child Memory Specificity

		Child Memory Specificity
		Unstandardized Estimate (SE)	Standardized Estimate (SE)
Model R2	.28**
Child Age		.78 (.33)*	.22 (.09)*
Child EVT		.09 (.03)**	.34 (.09)**
Emotional Elaboration		−1.15 (.49)*	−.25 (.11)*
Structural Elaboration		.26 (.38)	.07 (.11)
Emotional * Structural		1.21 (.50)*	.23 (.10)*

Note.

^*p < .05

^**p < .01.

EVT = Expressive Vocabulary Test.

To evaluate the nature of the relationship among these variables, the interaction was probed by calculating and plotting simple slopes for high (one SD above the mean) and low (one SD below the mean) levels of structural elaboration (see Figure 1). Simple slope analysis revealed that at low levels of structural elaboration, emotional elaboration was negatively associated with child memory specificity (b = −2.14, p < .01). At high levels of structural elaboration, there was no significant association between emotional elaboration and memory (b = −.16, p > 05). Furthermore, at low levels of emotional elaboration, there was no significant difference in child specificity between low and high levels of structural elaboration. Whereas at high levels of emotional elaboration, mothers with high structural elaboration had children with more specific memory compared to mothers with low levels of structural elaboration.

Figure 1. Interaction Between Emotional and Structural Elaboration Predicting Child Memory Specificity.

Note. Error bars represent standard errors. The simple slope at low structural elaboration is significant, and the simple slope at high structural elaboration is non-significant.

To assess our second hypothesis concerning the relationship between reminiscing and child AM specificity within the broader context of maternal attachment, a moderated regression analysis was conducted. To consider whether attachment avoidance and or attachment anxiety would moderate a relationship between structural and or emotional reminiscing with child memory specificity, product terms among the variables were entered to evaluate a four-way interaction between avoidance, anxiety, structural, and emotional elaboration. All possible two-way (structural by emotional, avoidance by structural, anxiety by structural, avoidance by emotional, anxiety by emotional, and avoidance by anxiety) and three-way interactions (emotional by structural by avoidance, emotional by structural by anxiety, structural by avoidance by anxiety, emotional by avoidance by anxiety) were considered in addition to the four-way interaction (see Table 5). Child language and child age were also entered as continuous covariates. Results indicated that the four-way interaction was non-significant (β = −.11, p > .05). However, the three-way interaction between emotional elaboration, attachment avoidance, and attachment anxiety significantly predicted memory specificity (β = .33, p < .01). The model was saturated and explained a significant portion of the variance in child specificity, R² = .44, p < .001.

Table 5.

Interaction between Attachment Avoidance, Attachment Anxiety, Structural, and Emotional Elaboration Predicting Child Memory Specificity

	Child Memory Specificity
	Unstandardized Estimate (SE)	Standardized Estimate (SE)
Child Age	.58 (.31)+	.17 (.09)+
Child EVT	.10 (.02)**	.35 (.09)**
Structural elaboration	.14 (.36)	.04 (.10)
Emotional elaboration	−2.12 (.52)**	−.47 (.11)**
Attachment avoidance	.40 (.25)	.17 (.11)
Attachment anxiety	−.37 (.29)	−.14 (.11)
Emotional * Structural elaboration	1.90 (.52)**	.37 (.10)**
Avoidance * Structural elaboration	−.24 (.35)	−.08 (.11)
Anxiety * Structural elaboration	−.14 (.36)	−.04 (.10)
Avoidance * Emotional elaboration	.44 (.47)	.11 (.11)
Anxiety * Emotional elaboration	−.22 (.46)	−.05 (.12)
Avoidance * Anxiety	−.24 (.21)	−.12 (.10)
Emotional * Structural * Avoidance	.02 (.65)	.00 (.12)
Emotional * Structural * Anxiety	−.02 (.73)	.00 (.12)
Structural * Avoidance * Anxiety	.06 (.29)	.02 (.11)
Emotional * Avoidance * Anxiety	1.14 (.41)**	.33 (.12)**
Emotional * Structural * Avoidance * Anxiety	−.56 (.56)	−.11 (.11)

Note.

⁺p = .07

^*, p < .05

^**p < .01.

EVT = Expressive Vocabulary Test.

To evaluate the nature of the relationship among these variables, the three-way interaction between emotional elaboration, anxiety, and avoidance was evaluated separately and probed by calculating and plotting simple slopes for high (one SD above the mean) and low (one SD below the mean) levels of attachment avoidance (see Figure 2). At high levels of attachment avoidance, emotional elaboration and attachment anxiety significantly interacted to predict child memory specificity, β = −.32, p = .05 (see Figure 2). Simple slope analysis revealed that among individuals with low attachment anxiety, there was no significant association between emotional elaboration and child specificity, b = −.001, p > 05. However, among individuals with high attachment anxiety, emotional elaboration negatively predicted child specificity, b = −2.96, p < .05 (see Figure 2). At low levels of attachment avoidance, emotional elaboration and attachment anxiety did not interact, β = .24, p > .05.

Figure 2. Emotional Elaboration Interacts with Attachment Anxiety at High Attachment Avoidance.

Note. Error bars represent standard errors. Simple slope is significant at high anxiety, and non-significant at low anxiety.

Event Category Analysis

Correlations among the event category reminiscing variables are presented in Table 6. Attachment avoidance was negatively correlated with structural elaboration for separation events. No other significant correlations emerged between either of the attachment variables with any of the category reminiscing variables. Emotional elaboration for separation events negatively predicted child AM specificity. No other reminiscing variables were significantly correlated with child specificity. The majority of the category reminiscing variables were inter-correlated with one another.

Table 6.

Correlations Among Event Category Variables

Measure	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.
1. Attachment avoidance	---
2. Attachment anxiety	.44**	---
3. Structural elab: happy	−.08	−.01	---
4. Emotional elab: happy	.04	.04	.08	---
5. Structural elab: stress	.12	.16	.45**	.17	---
6. Emotional elab: stress	.01	.02	.27*	.35**	.56**	---
7. Structural elab: conflict	.06	.02	.45**	.26*	.45**	.29**	---
8. Emotional elab: conflict	.07	−.04	.19+	.29**	.10	.19	.29**	---
9. Structural elab: separation	−.29**	−.08	.46**	.25*	.40**	.29**	.51**	.29**	---
10. Emotional elab: separation	.04	.10	.05	.28**	.26*	.44**	−.05	.06	.22*	---
11. Child memory specificity	.05	−.12	.05	−.03	.13	−.16	.07	−.20+	−.06	−.22*	---

Note.

⁺p<.10

^*p < .05

^**p < .01.

Elab = elaboration.

Four parallel moderation analyses were conducted to evaluate the hypothesis that associations between maternal attachment and child AM specificity would be strongest within reminiscing discussions of stressful or separation experiences. Bonferroni-corrections were applied to all analyses to account for multiple tests. The emotional elaborative reminiscing variables for (1) stressful, (2) separation, (3) happy, and (4) conflict event discussions were utilized for analysis. Structural elaborative reminiscing variables were not considered further, as no significant results were found involving structural elaboration in relation to maternal attachment in substantive analyses. First, to consider whether attachment avoidance and attachment anxiety would moderate a relationship between elaborative emotional reminiscing of stressful events and child memory specificity, product terms among the variables were entered to evaluate a three-way interaction between avoidance, anxiety, and emotional elaboration of stressful events. Child language and child age were also entered as continuous covariates. Results indicated that for stressful event discussions, the three-way interaction significantly predicted memory specificity (β = .45, Bonferroni-corrected p =.016). The model was saturated and explained a significant portion of the variance in child specificity, R² = .35, p < .001. As above, the interaction was probed by calculating simple slopes for high (1 SD above mean) and low (1 SD below mean) levels of attachment avoidance. At high levels of attachment avoidance, emotional elaboration of stressful events and attachment anxiety significantly interacted to predict memory specificity (β = −.41, p < .01) in a manner consistent with hypothesis 2 (see Figure 2). Simple slope analysis revealed that among individuals with low attachment anxiety, there was no significant association between emotional elaboration of stressful events and child specificity, b = .24, p > 05. However, among individuals with high attachment anxiety, emotional elaboration of stressful events negatively predicted child specificity, b = −1.97, p < .01. At low levels of attachment avoidance, elaboration and attachment anxiety did not significantly interact, β = .21, p > .05.

Parallel analyses were conducted to consider whether this interaction was significant within other event categories. The three-way interaction among elaboration, attachment avoidance, and attachment anxiety was non-significant for emotional elaboration of separation events (β = .12, p > .05), happy events (β = .10, p > .05), and conflict events (β = .17, p > .05). Thus, in partial support of hypothesis 3, maternal attachment avoidance and anxiety interacted with emotional elaboration of stressful event discussions in relation to child memory specificity.

Post-Hoc Analyses

Post-hoc analyses were conducted to further clarify and characterize differences between mothers (1) high and low on structural elaboration, (2) high and low on emotional elaboration, and (3) high on attachment avoidance and anxiety relative to other attachment styles. Median splits were utilized to compute two dichotomous variables distinguishing high versus low structural and emotional elaboration. Independent samples t-tests were conducted to compare the emotional reminiscing variables for mothers high and low on structural elaboration. Compared to mothers low on structural elaboration, those high on structural elaboration had significantly higher levels of emotional elaboration across all event categories, with the exception of emotional elaboration for separation events (see Table 7). Independent samples t-tests were conducted as above to compare mothers high and low on emotional elaboration on structural reminiscing variables. Compared to mothers low on emotional elaboration, those high on emotional elaboration had significantly higher levels of structural elaboration within stressful event discussions, although no significant differences were found for structural elaboration of happy, conflict, or separation events (see Table 8).

Table 7.

Post-Hoc Analysis of Reminiscing Differences between Mothers High and Low on Structural Elaboration

	Group Means (SD)
Reminiscing variable	Low Structural	High Structural	t
Emotional elab: happy	1.60 (0.78)	2.02 (0.94)	t(90) = −2.33*
Emotional elab: stress	1.12 (0.96)	2.01 (0.95)	t(90) = −4.50***
Emotional elab: conflict	.66 (0.89)	1.11 (1.00)	t(90) = −2.28*
Emotional elab: separation	1.11 (0.83)	1.43 (1.02)	t(88) = −1.61, ns

Note. p<.05 = *, p<.01=**, p<.001***, ns = p > .05.

SD = standard deviation. Reminiscing variables have been transformed due to skewness.

Table 8.

Post-hoc Analysis of Reminiscing Differences between Mothers High and Low on Emotional Elaboration

	Group Means (SD)
Reminiscing variable	Low Emotional	High Emotional	t
Structural elab: happy	3.23 (1.01)	3.51 (1.10)	t(90) = −1.25, ns
Structural elab: stress	2.94 (1.02)	3.71 (1.15)	t(90) = −3.36**
Structural elab: conflict	3.07 (1.19)	3.23 (1.00)	t(90) = −0.70, ns
Structural elab: separation	3.05 (0.87)	3.38 (0.92)	t(88) = −1.78, ns

Note. p<.05 = *, p<.01=**, p<.001***, ns = p > .05.

SD = standard deviation. Reminiscing variables have been transformed due to skewness.

To further clarify the nature of these differences, proportion scores were computed to assess the relative utilization of each elaborative element. For example, proportion of open-ended questions was computed as the frequency of WH questions out of all structural elaborative elements (e.g., WH questions / (WH questions + YN questions + confirmations + elaborations)). Proportion scores were computed in the same way for each of the structural elaborative elements. Proportion of emotional elaborative elements was likewise computed. For example, proportion of negative attributions was computed as the frequency of negative attributions out of all emotional elaborative elements (e.g., neg attr / (neg attr + pos attr + emotion questions)), and variables were computed in the same way for positive attributions and emotion questions. No significant differences were detected between mothers high versus low on structural elaboration for any of the proportion scores. In contrast, mothers high on emotional elaboration had a significantly higher proportion of negative attributions than mothers low on emotion elaboration, t(91) = −2.15, p < .05. Moreover, mothers high on emotional elaboration had (1) a higher proportion of closed-ended (YN) questions, t(91) = −2.61, p < .05, and (2) a lower proportion of structural elaborative statements wherein mothers added new, non-emotional information to the dialogue, t(91) = 3.82, p < .001.

Third, to consider differences in reminiscing across attachment groups, mothers were assigned to one of four attachment groups using median splits: (1) high avoidance, high anxiety, (2) high avoidance, low anxiety, (3) low avoidance, high anxiety, and (4) low avoidance, low anxiety. However, no significant group differences were detected in any of the afore-mentioned reminiscing variables utilized in the current investigation using one-way ANOVA analyses (all p-values = .09 – .97).

Discussion

The findings of the current investigation advance our understanding of predictors of child autobiographical memory specificity in three primary ways. First, the association between structural reminiscing and child memory specificity was moderated by emotional elements of reminiscing. Second, maternal attachment moderated associations between reminiscing and child memory, such that emotional elaboration negatively predicted child memory specificity at high levels of attachment anxiety and attachment avoidance. Third, analyses by event category indicated that the three-way interaction between emotional elaboration, attachment avoidance, and attachment anxiety was significant for mother-child discussions of stressful events. Together, these results support and clarify the DP model of OGM by revealing that mother child-reminiscing and maternal attachment together influence child autobiographical memory specificity. This reinforces the salience of variables at multiple ecological levels in models of child AM and advances knowledge of predictors of specificity during childhood, as it appears that OGM develops within a social context and is multiply influenced by maternal characteristics (Hitchcock, Nixon, & Weber, 2013; Valentino, 2011). More broadly, these results support sociocultural models of memory development (e.g., Nelson & Fivush, 2004; Fivush, 2011) by demonstrating that child remembering abilities develop through mother-child interactions and are further influenced by a mother’s own attachment orientation, which involves her beliefs, behaviors, and emotional tendencies within interpersonal relationships.

Structural and Emotional Elaboration

This study is the first to demonstrate that the quantity of maternal emotional elaboration moderates the association between maternal quantity of structural elaboration during reminiscing and child autobiographical memory specificity, which is consistent with our first hypothesis. At high levels of emotional elaboration, mothers with high structural elaboration had children with more specific memory than mothers with low structural elaboration. This finding clarifies the lack of association between structural reminiscing quantity and child memory specificity in our previous report with this sample (Valentino et al., 2014) by demonstrating that the association between structural reminiscing and child memory depends on emotional elaborative elements of the discussion. Thus, highly structured and elaborated discussions of children’s emotions may facilitate children’s memory specificity by helping children to develop a coherent narrative of their past emotional experiences and how they fit into their developing autobiography (Fivush, 2011; Welch-Ross, 2001).

Emotional elaboration negatively predicted child memory specificity at low levels of structural elaboration. In other words, mothers who engaged in high levels of emotional elaboration (e.g., asking emotion questions, referencing emotions) but did not scaffold emotional discussions with high levels of structural support (e.g., providing additional elaborative details, asking questions or confirming child contributions) had children with less specific memory. These findings further underscore the importance of emotions and how they are discussed for the development of child AM specificity, which is consistent with associations between OGM and emotional disorders (Valentino, 2011). These results suggest that it is necessary to consider multiple indices of reminiscing in relation to child outcomes (e.g., Larkina & Bauer, 2010) and advance previous work showing main effects of structural and emotional elements of reminiscing on child functioning by demonstrating that emotional elements of reminiscing affect the manner in which structural elements of reminiscing relate to child variables (Fivush et al., 2006).

Post-hoc analyses demonstrated that there were no significant differences in structural elaboration between mothers high and low on emotional elaboration for three of the four event categories, suggesting that those high on emotional elaboration were not consistently higher in structural elaboration. Moreover, mothers high on emotional elaboration used a higher proportion of negative attributions and closed-ended questions and a lower proportion of elaborative statements that added new, non-emotional information to the dialogue. This suggests that mothers high on emotional elaboration may have been disproportionately discussing negative emotions without elaborating on structural event details or encouraging child contributions with open-ended questions. In other words, mothers high on emotional elaboration appeared to initiate discussion of negative emotions, in particular, but then did not explain them or encourage the child to participate in the conversation. This is consistent with our prediction that unstructured emotional elaboration which does not involve the child nor elaborate on causes and consequences of emotions would impede a child’s ability to organize emotional experiences into a coherent narrative that would facilitate specific memory retrieval.

Maternal Attachment, Elaboration, and Memory

Consistent with our second hypothesis, maternal attachment moderated the association of emotional elaboration with child specificity. Importantly, maternal attachment did not moderate associations between structural elaboration and child memory specificity as it appears that maternal attachment primarily influences the emotional nature of the mother-child discourse, which is consistent with prior work (Bost et al., 2006). Emotional elaboration predicted less specific child memory at high levels of both attachment avoidance and anxiety, a profile referred to as fearful avoidance. Individuals with fearful avoidance tend to be disorganized or incoherent with their own emotional responses, sense of self, and coping strategies, and often present with approach-avoidance conflicts wherein they simultaneously desire and avoid closeness (Simpson & Rholes, 2002). Thus, mothers with fearful avoidance may not adequately facilitate emotional reminiscing with their children because they are disorganized with respect to their own emotions.

Post-hoc analyses with the variables in the current investigation were unable to characterize the nature of reminiscing differences across attachment styles. It is possible that mothers high in attachment avoidance and anxiety may initiate emotional discussions with their children because they desire closeness, but subsequently avoid emotional resolution or closure. Therefore, these mothers may reminisce with their children in a way that is more incoherent or impoverished than mothers with either anxiety or avoidance alone, an idea which should be further examined in future work. Overall, these findings advance previous work which has linked maternal attachment insecurity with impoverished reminiscing (e.g., Bost et al., 2006; Coppola et al., 2013; Reese, 2008) by demonstrating that avoidance and anxiety together create a context of risk in which high levels of emotional elaboration predict less specific child memory. Moreover, this extends Williams’ (1996) theory of functional avoidance in relation to OGM by demonstrating that maternal attachment strategies involving avoidance predict child specificity. Finally, this study extends previous work which has shown higher levels of elaboration among securely attached mother-child dyads (e.g., Laible & Thompson, 2000; McCabe, Peterson, & Connors, 2006) by demonstrating that maternal adult attachment, separate from the mother-child relationship, is a relevant individual difference variable that affects the context of reminiscing.

Consistent with our third hypothesis, maternal attachment primarily influenced the association between child memory specificity and emotional elaboration of stressful events (e.g., trips to the emergency room), but not for discussions involving happy themes or experiences of conflict or separation. This finding is consistent with Fivush and Sales (2006), who demonstrated that maternal attachment anxiety was correlated with higher narrative engagement during discussions of stressful events and higher levels of child internalizing symptoms. The current study extends this work by demonstrating a role for attachment anxiety and attachment avoidance in relation to child memory within the context of stressful event discussions. Although attachment avoidance is associated with distancing from others in response to stress, attachment anxiety is associated with seeking closeness from others in response to stress (i.e., dependence, clinginess) (Shaver & Mikulincer, 2002). Therefore, during discussions of stressful events in particular, mothers high on both attachment dimensions may simultaneously seek comfort from and avoid emotional closeness with their child. Future work should test the idea that this attachment pattern may be associated with boundary dissolution and emotional flooding that negatively influences the child’s ability to organize experiences and emotions in a way that allows for specific memory retrieval.

Contrary to our hypothesis, the interaction between maternal attachment and elaboration was not significant for discussions of separation. It is possible that separation events may be more difficult to discuss, as by definition, they involved unshared experiences between mother and child once the separation occurred, wherein the mother would not have been present to observe details of the event or how the child felt or responded. Further, separation cues may be more predictive of child attachment rather than maternal attachment, as theoretically a child would be more sensitive to separation from an attachment figure (Bowlby, 1969). Overall, it appears that conversations about stressful events engage the maternal attachment system more strongly than other emotional cues.

Limitations and Future Directions

There are limitations to this investigation that should be addressed and considered in future research. First, although our sample size was comparable to much of the extant literature concerning mother-child reminiscing, future research with a larger number of participants will increase the power of statistical analyses to examine more complex models. Further, the current investigation was cross-sectional; therefore, it was not possible to obtain unbiased estimates of indirect effects among study variables. Future longitudinal research is necessary to determine the direction of relations among the study variables and to consider possible mechanisms through which maternal attachment may influence child memory over time. Previous work has suggested multiple processes that may underlie an indirect association between maternal attachment and child memory (see Alexander et al., 2002b for a review), including child attachment (Messina & Zavattini, 2013) and child coping skills (e.g., Abaied & Rudolph, 2010).

In addition, the current investigation utilized a high-risk community sample, which advances previous literature on mother-child reminiscing and may have allowed for observation of a broader range of attachment insecurity than would have otherwise been observed. However, it will be important to replicate these findings among more normative samples. Further, the attachment measure in the current investigation utilized maternal self-report to build on prior work assessing the ECR in relation to reminiscing (Fivush & Sales, 2006) and child memory (Alexander et al., 2002a; Chae et al., 2014). Although maternal self-report attachment measures have been extensively validated with adults, it is important for future work to replicate these findings using observational measures of maternal attachment organization (e.g., Adult Attachment Interview). Finally, given the large percentage of ‘no memory’ responses on the child AMT, further work is needed to unpack the meaning of this response and evaluate whether omissions should be considered as OGM (Griffith et al., 2012). In addition, although the AMT has been validated for use with preschool children (Nuttall et al., 2014), future work should evaluate how child responses on the AMT compare to other well established memory assessment protocols. Moreover, comparison of AM specificity with generalized event representations (scripts) versus episodic memories of specific events would advance knowledge of how OGM relates to normative AM and further integrate our understanding of memory in the context of both typical and atypical development, a pillar of the developmental psychopathology perspective.

Additional directions for further research include consideration of broader indices of maternal engagement in reminiscing that evaluate the overall quality of the emotional discourse, and direct examinations of how quantitative coding of structural and emotional elaboration relates to quality variables. Previous investigations have considered maternal warmth (Fivush & Vasudeva, 2002), maternal autonomy support (Cleveland & Reese, 2005), and maternal sensitive guidance and emotional support (Valentino et al., 2014). It is possible that high quality reminiscing, as assessed in our previous report (Valentino et al., 2014), is characterized by both high structural and high emotional elaboration. However, no empirical work has directly compared the two rating schemes with one another, which is another important direction for further research. Thus, inclusion of multiple measures of the reminiscing discourse will further clarify how elaboration relates to child outcomes, may explain why high levels of emotional elaboration without structural support predict less specific child memory, and may inform understanding of the nature of reminiscing among mothers with fearful avoidance. It would also be informative to integrate variable-centered with person-centered analyses, such as cluster or latent class analysis, in order to further clarify meaningful differences in maternal reminiscing.

Conclusions

The findings of the current investigation extend theoretical models of OGM involving functional avoidance (Williams, 1996) and support the DP model (Valentino, 2011). In particular, our work suggests that individual differences in child autobiographical memory specificity develop through parental interactions, and are influenced by adult risk factors. The current results further highlight the salience of sociocultural models of memory development by revealing that child remembering develops, and can be both facilitated or impeded, within a social context involving child interactions with the adults in their lives. The present investigation also has important clinical implications for interventions targeting maternal reminiscing skills (see Wareham & Salmon, 2006 for a review). It is critical that mothers be trained to scaffold emotional exploration with high levels of structural support, as emotional elaboration without such scaffolding may negatively predict child outcomes. Further, maternal attachment may be a moderator of intervention effectiveness, as mothers with high levels of attachment avoidance and anxiety may reminisce in ways that negatively predict child functioning. Intervention outcome studies that utilize repeated measures at multiple levels of analysis will inform our understanding of the efficacy of reminiscing training in promoting child autobiographical memory specificity and reducing the onset of depressive and trauma-related symptoms associated with OGM. In addition, such longitudinal research designs will inform our understanding of the direction of associations between OGM and psychopathology, and may elucidate processes by which child AM specificity, maternal reminiscing, and/or parental risk factors such as maternal attachment may contribute to emotional maladjustment.