Maternal verbal co-construction skills: associations with maternal sensitivity and toddlers' attachment security

This study investigated whether organizational characteristics of mothers' verbalizations (i.e., co-construction skills) when creating attachment-related stories with their children are related to maternal behavioral sensitivity and child behavioral attachment security. Fifty-four 2.5-year-olds and their mothers from Mid-West USA participated in three sessions: lab, home, and playground. Three organizational characteristics of maternal verbalizations were assessed during a joint storytelling task in the lab: creating a co-constructive atmosphere, encouraging content elaboration, and supporting explanatory frameworks. Observations of maternal sensitivity and child attachment security were completed both at home and at the playground; composites of sensitivity and security across contexts were used in analyses. As expected, sensitivity and security were significantly associated. Co-construction skills were significantly associated with maternal sensitivity. Contrary to expectations, co-construction skills were not related to child security. Theoretical and methodological implications are discussed.

The quality of parental caregiving is key to the organization of child attachment behavior (Ainsworth et al., 1974, 1978/2015; Bowlby, 1969/1982). This has been supported through a wealth of empirical research demonstrating significant associations between sensitive maternal behavior and children's attachment security in infancy and early childhood (e.g., De Wolff et al., 1997; Hopkins et al., 2013; Verhage et al., 2016; Madigan et al., 2024). As young children's linguistic and communication skills expand, new lines of social interaction beyond those seen in infancy arise and may come to inform parent-child attachment security (Waters and Waters, 2024). As children develop increasing levels of behavioral and social competence, caregivers may come to increasingly rely on language to coordinate interactions with their children (e.g., Bretherton, 1990, 1993, 1995). Thus, although the behavioral building blocks established in infancy remain, child-parent attachment relationships involve a more salient verbal-representational component during early childhood (O'Neill et al., 2021; Posada et al., 2018; Waters and Waters, 2024). Those changes give rise to questions concerning the associations among characteristics of caregiver verbal communication, caregiver sensitive behavior, and child secure-base behavior.

Mother-child verbal communication and attachment relationships during early childhood

Both the quality of caregiving (i.e., caregiving sensitivity; Ainsworth et al., 1974, 1978/2015) and parent-child communication quality (e.g., emotional openness, responsiveness; Bowlby, 1969/1982, 1973; Bretherton, 1990, 1991, 1993) have been theorized as key factors in the development of secure child-parent attachment relationships and internal working models. While caregiving sensitivity has been widely linked to attachment security during infancy and early childhood (Ainsworth et al., 1974, 1978/2015; De Wolff et al., 1997; Verhage et al., 2016), fewer studies have examined the role of parent-child verbal communication during the early childhood period (O'Neill et al., 2021; Posada et al., 2018).

As children's language abilities expand during the second year of life, parent-child open communication about emotions and interactions are likely to be related to and foster secure attachment relationships (Bowlby, 1988; Thompson, 2000). Indeed, children's expanding language skills during early childhood offer opportunities for verbal exchanges with caregivers that provide a means of influencing attachment-related experiences as well as the organization of knowledge about attachment relationships (Guttmann-Steinmetz et al., 2003; Lu et al., 2018). During toddlerhood and the early preschool years parents may initially serve as the organizers of verbal interactions with young children, providing most of the content and structure for these conversations as a means of scaffolding (Eisenberg, 1985; Fivush, 2001; Hudson, 1990). Specifically, attachment figures may use language to help children understand and organize attachment-related information, regulate emotions, and organize attachment behavior in daily interactions with them (Thompson, 2000; Waters and Waters, 2024).

Indeed, prior research on the link between parental reminiscing style and child attachment security suggests a significant association between the two. Several studies on parent-initiated conversations about past events report that mothers of securely attached preschoolers are more elaborative (e.g., use open-ended questions), use more emotional evaluation statements and words, and adjust their reminiscing style as children grow (Bost et al., 2006; Coppola et al., 2014; Fivush and Vasudeva, 2002; Laible, 2004, 2011; Laible and Thompson, 2000; Newcombe and Reese, 2004; Reese et al., 2019). All those findings support the notion that mothers' conversation style when talking with their children about past events likely influences their relationship and a child's attachment security.

While studies on parent-child reminiscing support the general claim that some aspects of parental verbal style contribute to the quality of child-parent attachment relationships (e.g., Newcombe and Reese, 2004; Reese et al., 2019), many questions remain. First, past research has mainly focused on frequency counts of emotional words and/or elaborative strategies in reminiscing tasks (e.g., Fivush and Vasudeva, 2002; Laible, 2004; Newcombe and Reese, 2004). Less attention has been paid to the pragmatic organization of conversations (i.e., framework used for the causal explanation of events; cooperation between dyads). Looking at organizational features of parents' conversational input during exchanges with their children may shed light on an important scaffold through which children learn to put together attachment-related information and guide their behavior during interactions with their attachment figures. Thus, while it is well supported that the organization of parental behavior (i.e., sensitivity) is important in helping a child to organize her/his attachment behavior (e.g., De Wolff et al., 1997; Madigan et al., 2024), additional factors such as the verbal scaffoldings from parents may also play a role. As children move into the toddlerhood period and beyond, the organizational features of parental contributions to parent-child conversations may help children feel secure in their child-parent relationships.

Second, positive associations between maternal reminiscing styles and attachment security during both preschool and early childhood (i.e., age 2.5–5) have been reported (e.g., Bost et al., 2006; Coppola et al., 2014). However, the link between conversation organizational characteristics (i.e., balance or match of dyadic emotions when discussing events) and attachment security has been less studied, and contradictory results have been found. Significant associations were reported for older preschoolers (i.e., 4 years and older, Etzion-Carasso and Oppenheim, 2000; Gini et al., 2007), but not for younger preschool-aged children (i.e., 4 years old and below, Oppenheim and Koren-Karie, 2009). One possible explanation for this discrepancy concerns the cognitive and linguistic demands of reminiscing tasks, which require children to recall past events-an ability that is still developing in toddlerhood (Nelson and Fivush, 2004). Thus, alternative means of assessment that do not rely on memory recall may help us better understand whether there is truly a developmental shift during early childhood where characteristics of caregiver-child conversations begin to inform attachment security and whether children of different ages might benefit from verbal support from their caregivers in differentiated ways.

Along these lines, caregiver-child dyads not only discuss past events in their day-to-day lives but also talk about co-occurring events and expectations for future and hypothetical situations, including those concerning social relationships. Parents may share with their children what to anticipate, how to respond, and potential emotions they might experience as they encounter and explore novel contexts. And yet, most work on caregiver-child conversations focuses on discussions of the dyad's past experiences (i.e., the reminiscing task, Fivush and Vasudeva, 2002; Newcombe and Reese, 2004; Reese et al., 2019). While such studies are informative, they do not provide a holistic understanding of the varied ways parents may converse with their children that may add to their understanding of social interactions with others. Thus, in addition to reminiscing, discussion about co-occurring attachment-related events provides a structured yet flexible context that is more accessible to younger children, as it does not rely on their ability to recall past events, but instead allows them to actively engage in co-constructing a narrative with their caregiver in real-time, enabling a more immediate and interactive exchange.

Maternal co-construction skills and child-mother attachment relationships

In addressing pending questions about how parental verbal communication contributes to the development of children's security, Posada and Waters (2018a) suggested that parental verbal input provides children with information about secure base interactions. Notably, they argued that during conversations about attachment-related events, parents offer children information about the availability and responsiveness of attachment figures for secure base support (e.g., Lu et al., 2018). Furthermore, they suggested that these conversations offer children a context to learn about expectations for attachment relationships that includes discussions of children's emotions and the role of the attachment figure in navigating these emotions. As a result, as children start to develop a deeper understanding of a caregiver's role in navigating attachment relationship, the parent-led verbal exchanges may influence both the content and organization of children's attachment-related knowledge, and the organization of their secure base behavior and feelings of security.

To this end, H. Waters and colleagues (Guttmann-Steinmetz et al., 2003; Posada and Waters, 2018b) proposed three specific parental conversational skills that may play a significant role in children's organization of secure base behavior: (1) creating a collaborative conversation atmosphere with their child when talking about attachment-related issues; (2) promoting children's elaboration during these exchanges by encouraging greater depth in the details of their child's contributions to the conversation; and (3) supporting children's understanding of experiences by building causal links among critical aspects of attachment-related experiences. These three skills are collectively labeled “co-construction skills” and are concerned with how mothers frame and participate in verbal exchanges with their young children rather than with the quantity of maternal verbalizations.

Existing research on the relations between parent-child attachment relationships and the organizational characteristics of parental' verbalizations, understood as co-construction skills, during attachment-related conversations is limited (but see Apetroaia and Waters, 2018; Lu et al., 2018). Moreover, developmental considerations highlight the need to study how caregiving behaviors change with child age and encourage researchers to explore the concept beyond infancy (Waters and Waters, 2024). Studies of sensitivity during infancy are centered around a caregiver's ability to respond promptly and appropriately to a child's signals and communications (e.g., Ainsworth et al., 1974, 1978/2015; Verhage et al., 2016). As children develop and acquire more advanced language and cognitive abilities, caregiver-child interactional dynamics may change and verbal interaction may start to play an increasingly important role in child-caregiver exchanges (e.g., Guttmann-Steinmetz et al., 2003; Lu et al., 2018). Questions about quality of maternal care and its associations, if any, with organizational characteristics of mothers' verbalizations during the toddlerhood years need to be addressed to better understand age-related caregiving influences and their implications for child security that may extend beyond those comprised in customary assessments of sensitivity.

To date, only one study has examined the relations among parental co-construction skills, sensitive caregiving behavior, and child attachment security among preschool-aged children (Lu et al., 2018). The authors found that mothers' co-construction skills were significantly associated with maternal sensitivity when children were 3.5 years old. Furthermore, co-construction skills at 3.5 years significantly predicted the organization of children's secure base behavior (i.e., security) concurrently and longitudinally at 5.5 years of age, above and beyond the contributions of maternal sensitivity at each point in time. This suggests that while related, maternal co-construction skills and maternal sensitivity play unique roles in the prediction of children's attachment security among 3.5–5.5-year-old children. While stimulating, those results not only require replication, but also further exploration in different age groups and samples given the contradictory findings reported across studies using reminiscing tasks, especially the non-significant findings among younger age groups (e.g., Etzion-Carasso and Oppenheim, 2000; Oppenheim and Koren-Karie, 2009). These findings raise question about how early those associations are established, and whether maternal co-construction skills add significant information to our understanding of the development of attachment security during toddlerhood.

Current study

Based on consideration of developmental adaption of parent-child interaction and the role of verbal communication in child attachment security, we first investigated whether maternal verbal co-construction skills (“creating a collaborative/co-constructive atmosphere,” “encouraging content elaboration,” and “supporting an explanatory framework”) during a joint storytelling task about attachment-related events are associated with mothers' sensitive caregiving behavior during child-mother interactions in naturalistic settings. Evidence of a positive association would provide support to the notion that the ways mothers frame attachment-related conversations with their toddlers are related to maternal sensitive behavior, at an age when children increasingly signal and communicate by using verbalizations. Further, because verbal co-construction skills are concerned with how conversations are framed by caregivers rather than with the quantity of their verbal output, we investigated whether the amount of maternal verbalizations (measured as the number of words uttered in an independent, yet related story from the joint storytelling task; see procedures) was associated with maternal sensitivity. This allowed us to discern whether it is the structure of maternal verbalizations, as demonstrated by mothers' verbal co-construction skills, rather than the quantity of mothers' verbal output underlying the association, if any, between maternal language and sensitivity.

Second, we studied whether maternal verbal co-construction skills are positively associated with secure base behavior in 2–3-year-old toddlers, and if so, whether those skills contribute unique information to the prediction of toddlers' use of mother as a secure base beyond any information contributed by sensitive maternal behavior. Empirical evidence of a unique contribution of co-construction skills in predicting attachment security would back the notion that, in addition to maternal sensitive behavior, mothers' structuring of attachment-related conversations may be a relevant factor in the organization of children's secure base behavior soon after infancy. Based on past mixed findings regarding the association between maternal conversation style and child security (e.g., Lu et al., 2018; Oppenheim and Koren-Karie, 2009) in different age groups and using different methodologies, as well as the limited existing research examining the contributions of both maternal sensitivity and co-construction skills to children's secure base use among young preschool-aged children, this investigation was exploratory in nature.

Participants

Families with children 2–3 years old from a Midwestern town in the United States were enrolled in the project (IRB: 1505016087). We recruited participants by posting flyers in community preschools; recruitment information was also posted on local websites and shared via a university newsletter. Fifty-four mother-child dyads from a larger study on marital discourse and family interaction during early childhood completed three study visits: at home, playground, and university laboratory. Initially, 75 dyads were recruited, however, 13 dyads dropped out prior to the lab visit, and for 3 families, the child refused to participate in the Joint Storytelling Task. In addition, 5 dyads completed the Joint Storytelling Task using a native language other than that of the research team members, making it impossible to assess maternal verbal co-construction skills. Thus, complete data was obtained for 54 dyads.

Demographic information is presented in Table 1. There were 27 boys and 27 girls who participated in the study. Seventy percent of participating children were White (N = 38), followed by 5.6% Asian (N = 3), 1.9% African American (N = 1), and 1.9% Hispanic (N = 1). For 18.5% (N = 10), parents reported multiple ethnic-racial backgrounds. One family did not provide information. For 63% (N = 34) of the children, mothers were their primary caregiver while 33.3% (N = 18) children had both mothers and fathers as primary caregivers, and 3.7% (N = 2) children had their fathers as the primary caregiver. Most children were enrolled in childcare (N = 32). Most mothers were White (N = 41), 9.3% were Asian (N = 5), 7.4% were Hispanic (N = 4), 1.9% were African American (N = 1), and 5.6% reported multiple ethnic-racial backgrounds (N = 3). Twenty-four mothers did not work outside the home, 9 mothers worked half-time, and 21 mothers worked full-time. All mothers were married or cohabiting with their partners at the time of the study.

Procedures

Parents who were interested in the study contacted researchers via email or phone. Detailed information about the study and procedures were provided and if the family wanted to participate, three visits with mother-child dyads were scheduled: one home visit, one park visit, and one lab visit. Home and park visits lasted between 1.5–2 h each; the lab visit lasted around 2 h. The order of home and park visits varied, largely dependent upon the weather on the day of the first scheduled visit. At the start of the first visit, consent for participation was obtained, as was family demographic information. The lab visit was always the final of the three visits. Mothers were paid $20 after each visit and children were given a book at the end of the lab visit as a token of appreciation for their participation. The average between-visit time lag was 36 days between the first visit and the second visit, and 75 days between the second visit and the final lab visit.

During home visits, mothers and children were instructed to interact as they usually would. Mothers were not required to interact with their children. After about one hour, they were instructed to read a book together that was brought by the researchers to encourage more direct parent-child interaction. Two to three trained research assistants observed mother-child interactions across the entire visit; some observers focused on maternal behavior while others focused on child behavior. If mother and child separated during the visit, child observers followed the child while those observing the mother followed her. After the visit, both maternal and child behavior were described independently by observers using the Maternal Behavior with Preschoolers Q-set (MBPQS; Posada and Waters, 2018b) and the Attachment Q-Set 3.0 (AQS; Waters, 1995), respectively. During the observation, observers were free to interact (e.g., play with the child) and ask questions for the mother to clarify observed behaviors, as both MBPQS and AQS ask for child and maternal behaviors when being interrupted or being approached by others (e.g., mother monitors child behavior when occupied with other tasks; child readily lets new adults hold or share things he/she has, if they ask to). However, minimal observer-child and observer-mother interaction was encouraged in order to maintain natural mother-child interaction.

For park visits, families were asked to choose a familiar playground. The visits were unstructured and dyads were again instructed to behave as they normally would in this context. They were told that the research team would follow along if they moved from place to place. As in the home visit, if separated, child observers followed the child and mother observers followed the mother. Independent observers used the MBPQS and AQS to describe behavior following the playground visit.

To reduce the discomfort of being observed and to encourage mother-child dyads to interact naturally, neither of the visits were recorded and notes were not taken either (Posada et al., 2021; Vaughn et al., 2021). It should also be noted that across all visits, maternal and child observers were independent from each other (i.e., no observer reported on both maternal and child behavior for the same dyad).

During the laboratory visit, mother-child dyads were instructed to complete a joint storytelling task adapted from Posada and Waters (2018b) to accommodate toddlers/younger preschoolers' developmental capacities. While the original task includes 4 separate panels with 3–4 pictures in each panel that need to be changed and organized in a particular order for each scenario every time that a dyad finishes with a story, the modified task used the same picture prompts but organized them in a picture-book format making it more familiar to both children and mothers and easier to transition from story to story. Mother-child dyads were instructed to sit side by side on a couch, then the experimenter familiarized the child with the storytelling task by talking to the child and asking whether they “tell stories with mom at home.” When the child felt at ease, s/he was invited to create some stories with the mother.

Each dyad was given four separate wordless books (one at a time) with simple picture prompts designed for younger preschoolers. They were instructed to make up stories together using those picture books. Each book had a title and depicted a typical child-mother interaction scenario. Three of the stories depicted events including a stressful component that is thought to activate the attachment system: “Mommy comes home from the city” (separation and reunion), “Can't sit on mommy's lap” (child seeks comfort but is refused by mother), and “Finger stuck in bottle” (child is hurt). The fourth story (“Mommy and child go to the beach”) depicted a mother-child interactive event without a stressful component (i.e., play) and was used as a control story to assess the frequency of maternal language input. The order of the stories was counterbalanced across participants. In each book, there is a mother and a child character; the child character's sex was matched with the participating child's sex. Pictures in each wordless book imply a simple storyline, framing a beginning, middle, and end. Details and resolutions of the storyline were open-ended. The title of each story was given on the first page of the book which suggested the general story outline (e.g., “Mommy comes home from the city”). Mothers were instructed to help their children come up with the best story they could using the pictures in each book. Mother-child interactions during the task were videotaped and then transcribed verbatim.

Assessment

Maternal verbal co-construction skills

Mothers' co-construction skills were assessed using the three stories that included a stressful child-mother scenario (“Mommy comes home from the city,” “Can't sit on mommy's lap,” and “Finger stuck in bottle”) from the Joint Storytelling Task (Posada and Waters, 2018b). Three rating scales (“ creating a co-constructive atmosphere [CCA],” “ encouraging content elaboration [ECE],” and supporting an explanatory framework [SEF]) were adapted from the scales proposed by H. Waters. The original configurative scales from H. Waters were deconstructed and specified into several items. In our study, the CCA subscale consisted of 11 items (e.g., “tone of voice when talking is inviting,” “allows child to decide when to go to the next part or when the story is finished”); the ECE subscale consisted of 7 items (e.g., “encourages child to take the initiative in filling in details,” “uses open-ended prompts”); and the SEF subscale consisted of 5 items (e.g., “relates the storyline to some experiences the child has had, bridging the story line to the child's own experiences,” “provides an ending to the story that helps the child “see” the overall storyline”). Coders rated each subscale item using a 7-point Likert scale ranging from 1 (uncharacteristic) to 7 (characteristic) of a mother's verbalizations during the storytelling task. It should be noted that one item (“answers child's questions”) was removed from the ECE subscale due to lack of information; 34 out of the 54 children did not ask any questions during the Joint Storytelling Task, thus there was no context to observe maternal responses specific to this item.

For each dyad, two trained coders with no knowledge of maternal or child behavior as assessed with the MBPQS and AQS rated mothers' co-construction skills during each story. Training for coders included 4 sessions. During the first session, coders were introduced to the concept of co-construction, and reviewed parent-child storytelling examples illustrating different levels of co-construction skills. The subsequent sessions consisted of practicing coding exercises using example transcripts and videos. To be considered reliable coders, trainees need to achieve agreement with expert coders on three consecutive coding exercises, with discrepancies of less than 2 points across three scales composite. While co-construction skills were rated separately for each individual story, all three stories from a given mother-child dyad were coded by the same pair of coders. First, coders individually read the transcription for each story and scored the items, then watched the recorded video and adjusted their scores if needed. After, raters compared their item scores and discrepancies of 2 points or greater for each were discussed to try and reach consensus. A score for each scale was calculated by averaging the two coders' revised scores for the CCA, ECE, and SEF subscales across all three stories. Intraclass correlation (ICC) coefficients, two-way random effect model, ranged from 0.77 to 0.85 for the three subscales. Internal consistency (Cronbach's alpha) for the CCA, ECE, and SEF subscales were 0.92, 0.93, and 0.87, respectively. A composite score for “maternal co-construction skills” was created by averaging scores for each mother across the three subscales; both composite scores and the three subscale scores were used in analyses.

Maternal sensitivity

Maternal sensitivity was assessed with the MBPQS (Posada et al., 2007, 2018) after both the home visit and park visit. The MBPQS is a 90-item q-set that allows researchers to describe caregiving behavior germane to security outcomes in naturalistic settings during early childhood. Empirical support for the reliability and validity of the MBPQS has been reported (Posada et al., 2007, 2016, 2018). For this study, immediately following each visit, the MBPQS was completed independently by trained observers who sorted the items along a continuum from least characteristic to most characteristic. Following Q methodology, observers first sorted the 90 items into three piles: uncharacteristic, neither characteristic nor uncharacteristic, or characteristic. The three piles were then subdivided into nine piles of 10 items each, ranging from 1 (least characteristic) to 9 (most characteristic). The pile number in which items are placed determined the score for each item. As is common in q-methodology (see Vaughn et al., 2021), observer disagreements of greater than 3 piles were discussed to try to reach consensus. Following discussion, descriptions from all observers were composited across the home and park contexts. Out of a total of 108 observations of maternal behavior, 37 (34.3%) were conducted by two observers. Inter-observer reliability (Spearman-Brown coefficient) ranged from.70 to.92. Observers' descriptions across park and home contexts were averaged into a Q-composite description. For each mother, a global sensitivity score was computed by correlating her composite Q-description with a criterion sort that describes the prototypically sensitive mother. The resulting Pearson correlation coefficient is a mother's sensitivity score (range: −1 to +1) and expresses the degree of correspondence between a mother's composite behavioral description and the MBPQS criterion sort (Posada et al., 2007); higher scores reflect more sensitive behavior. Those scores were used in analyses.

Child security

Child security was assessed using the AQS 3.0 (Waters, 1995) after both home visit and park visit. The AQS is a 90-item Q-set that allows researchers to describe children's secure base behavior in naturalistic settings. Validity and reliability of the AQS have been reported in various studies [e.g., see meta-analysis by Van IJzendoorn et al. (2004)]. The AQS was completed independently after each home and park visit by trained observers, different from those who described maternal behavior. Observers followed the same procedures as those described for the maternal sensitivity assessment. Out of the 108 observations, 48 (44.4%) were conducted by two observers. Inter-observer reliability (Spearman-Brown coefficient) ranged from 0.70 to 0.93. Observers' descriptions across park and home contexts were averaged into a Q-composite description. For each child, a security score was computed by correlating that composite Q-description with a criterion sort that describes the prototypically securely attached child who uses her/his mother as a secure base. A security score thus expresses the degree of correspondence (i.e., correlation) between a child's behavioral description and the AQS criterion sort (Waters, 1995); higher scores reflect greater secure base use. Those scores were used in analyses.

Quantity of maternal verbalizations

For purposes of discriminant validity mothers' verbal output quantity as indexed by the number of words uttered when co-constructing a child-mother story was calculated and examined as a covariate. Specifically, quantity of maternal verbalization during the “Mommy and child go to the beach” story was used as a covariate in all analyses. Transcripts were first separated by speakers, and only maternal utterances during the storytelling were included. Mothers' total number of words during “Mommy and child go to the beach” story transcript was calculated using the Natural Language Process Toolkit in Python as length of the text corpus.

Analytical plan

Preliminary analyses were conducted to examine descriptive statistics, identify potential outliers, and evaluate demographic variables as potential covariates. Group differences were examined using t -tests and analyses of variance (ANOVA), and correlations were used to evaluate associations between continuous demographic variables and the primary study variables. Variables significantly associated with maternal sensitivity, child attachment security, or maternal co-construction skills were retained as covariates in subsequent analyses.

Main analyses were conducted to address the study's research questions concerning associations among maternal co-construction skills, maternal sensitivity, and child attachment security. Accordingly, partial correlations controlling for relevant covariates were conducted. One-tailed tests were used for analyses examining associations between maternal sensitivity and maternal co-construction skills as both prior theoretical and empirical work supported a positive directional hypothesis. In contrast, analyses involving child attachment security with maternal co-construction skills were considered exploratory due to mixed findings in the literature and were therefore evaluated using two-tailed tests.

Preliminary analyses

Dyads who completed participation in the Joint Storytelling Task (n = 54) and those who did not (n = 21) were compared on demographic and central variables using t -tests and analysis of variance (ANOVA). No significant differences were found for child age or biological sex, maternal age, maternal education, maternal occupation (e.g., full-time, part-time, stay-at-home), family annual income, race/ethnicity, primary caregiver status (e.g., mother, father, both mother and father), childcare arrangements, or years of marriage. Regarding the main variables of interest, no significant differences were found for maternal sensitivity or child attachment security scores between those completing the Joint Storytelling Task and those who did not. Descriptive statistics for child attachment security, maternal sensitivity, maternal co-construction skills (including the three subscales and overall composite), and frequency of maternal verbalizations are presented in Table 2. One child security score, 2 maternal sensitivity scores, and 1 maternal ECE score were found to be outliers, as they exceed three standard deviations below the mean of the respective variables. To minimize the influence of these extreme values while retaining available samples, winsorization was applied such that the innermost outlier was assigned the value three standard deviations below the mean of the respective variable, and for the second outlier in the case of maternal sensitivity, 0.01 was subtracted to maintain rank order within the data. Maternal sensitivity and verbalization amount were found to be negatively skewed. A log transformation was performed on both variables, but the transformation did not improve normality (i.e., skewness was not improved) thus, the original data were used in the analysis.

Covariate analyses were performed using t -tests to determine if maternal sensitivity, child attachment security, composite maternal co-construction skills, or the three maternal co-construction subscales varied by child biological sex, race/ethnicity (White vs. non-White), childcare (any vs. none), and child's main caregiver (only mother vs. mother and father). No significant associations were found. Correlational analyses were then performed to examine the relation between continuous demographic variables (child age, weekly hours in childcare, number of siblings, maternal age, education level, years of marriage, and family annual income) and the variables of interest. Findings revealed that child security was positively associated with child age [ r (52) = 0.313, p = 0.02], and maternal sensitivity was positively associated with maternal education [ r (52) = 0.348, p = 0.01]. No other significant associations were found. Thus, only child age and maternal education were included as covariates in main analyses.

Next, we examined the associations between quantity of maternal verbalizations and maternal co-construction skills, maternal sensitivity, and child attachment security. None of the main variables were significantly associated with the amount of maternal verbalizations (see Table 3); thus, it was not included as a covariate in the main analyses.

Main analyses

Our first research question addressed the association between maternal sensitivity and co-construction skills. Consistent with the prior directional hypothesis, a one-tailed test was applied while controlling for maternal education. We found that the maternal co-construction composite was significantly related to maternal sensitivity [ r (51) = 0.23, p = 0.047]. We further examined the association between each of the components of the co-construction composite and maternal sensitive behavior, controlling for education. Among the three subscales, creating a co-construction atmosphere [ r (51) = 0.25, p = 0.04] and supporting an explanatory framework [ r (51) = 0.30, p = 0.01] during the Joint Storytelling Task were both significantly related to maternal sensitivity. On the other hand, maternal encouragement of content elaboration was not [ r (51) =.11, p = 0.23].

Our second question explored the association between maternal verbal co-construction skills and toddlers' attachment security. Given the exploratory nature of these analyses, two-tailed tests were performed and revealed that after controlling for child age and maternal education as covariates, though maternal sensitivity and child attachment security were positively associated with each other [ r (50) = 0.47, p < 0.001], maternal verbal co-construction skills and child attachment security were not significantly related [ r (51) = 0.10, p = 0.49]. Similarly, none of the co-construction subscales were significantly related to toddlers' security after controlling for covariates [ rs ranged from 0.04 to 0.12, df = 51, p's > 0.10). Thus, while maternal sensitivity was related to child attachment security, maternal verbal co-construction skills did not contribute to the explanation of variance in child security in the current sample.

Attachment theory suggests that open emotional communication plays a key role in child-parent relationships, as parents provide children with information about their availability and responsiveness during conversations (Bowlby, 1988). Embedded in child-mother exchanges, maternal verbalizations provide children with information about expectations for child-parent interactions, emotions they may experience, and the role of attachment figures and children in those interactions. The purpose of the current study was to explore the associations among maternal verbal co-construction skills, maternal sensitivity, and child attachment security in a sample of 2–3-year-old children and their mothers.

Overall, maternal verbal co-construction skills were significantly associated with mothers' sensitivity. Specifically, we found that maternal co-construction skills concerned with creating a co-constructive atmosphere and supporting an explanatory framework during a joint storytelling task were significantly associated with maternal sensitive behavior during child-mother interactions in naturalistic settings, highlighting the relation between maternal caregiving behavior and pragmatic elements of parent-child conversations about attachment-related events during toddlerhood. Importantly, the amount of maternal verbalizations during the storytelling task was not associated with maternal co-construction skills or maternal sensitivity, suggesting that how mothers structure attachment-related conversations may be more informative than how much they talk. However, neither the overall composite score for maternal verbal co-construction skills nor any of the three co-construction subscales studied were related to child attachment security. Thus, while maternal sensitivity was significantly associated with children's security as expected, no evidence was found to support the idea that maternal verbal co-construction skills are related to attachment security in this sample of 2–3-year-old toddlers and their mothers.

Results extend previous research on maternal co-construction skills and their associations with child-mother attachment relationships, demonstrating that several aspects related to the organization, but not the quantity, of maternal language are significantly related to maternal sensitivity as early as toddlerhood. Results also suggest, in conjunction with previous research (e.g., Oppenheim and Koren-Karie, 2009), that verbal co-construction skills about attachment-related events may not play a significant role in the organization of children's secure base behavior (i.e., security) until later in early childhood (e.g., 3.5 years; Lu et al., 2018).

Maternal co-construction skills and maternal sensitivity

Mothers who showed more sensitive caregiving behavior while interacting with children in naturalistic settings (e.g., home and playground) were found to score higher on their verbal co-construction skills, specifically, on their ability to create a co-constructive atmosphere, a shared environment, and offer an explanatory framework when talking to their children about attachment-related events during a joint storytelling task. That is, sensitive mothers were attentive and skillful in creating a collaborative atmosphere for conversation when engaging their child in attachment-related storytelling. Further, those mothers were also more skillful in providing emotional references and an overall causal framework to facilitate their children's story comprehension. Such findings, in conjunction with those reported by Lu et al. (2018) for 3.5-year-olds, may suggest that verbal co-construction skills may represent one manifestation of sensitive caregiving within verbally mediated interactions during early childhood. At the same time, the fact that the associations between the constructs was modest implies that a significant amount of variance in maternal sensitivity remains unaccounted for by co-constructive skills. Thus, co-construction does not appear to be reducible to sensitivity alone but rather may capture a related but distinct aspect of sensitive caregiving. This extends attachment research beyond traditional behavioral indicators of sensitivity and highlights the importance of examining caregiver-child verbal interactions during early preschool years.

However, the fact that the associations between the constructs was modest implies that a significant amount of variance in maternal sensitivity remains unaccounted for by co-constructive skills. Thus, while creating a co-constructive atmosphere and providing an exploratory framework may indeed be important characteristics of sensitive caregiving when conversing with toddlers, there are other factors at play that encompass the broader sensitivity construct.

Interestingly, our assessment of mothers' ability to encourage content elaboration during the joint storytelling task was not related to sensitive maternal behavior assessed in naturalistic settings. This is at odds with findings reported in other existing studies on the relation between the two constructs (Lu et al., 2018; Reese et al., 2019). This discrepancy may be due to age differences in the samples studied. Alternatively, the difference may be due to our operationalization of “ encouraging content elaboration ” being distinct from that used in research on elaborative reminiscing, such as the work done by Reese et al. (2019). Elaboration in the context of reminiscing has been defined as the extent to which mothers support children's recall during ongoing parent-child conversation through use of open-ended questions, added details, and/or confirmations and feedback (Fivush et al., 2006; Reese et al., 2019). Our assessment of ECE focused on more specific strategies adopted by mothers, including the use of open-ended questions, follow-up questions, and confirmation of a child's statements (rather than the mother's own statements).

Maternal co-construction and child attachment security

Notably, this study did not reveal a significant association between maternal co-construction skills and the organization of children's secure base behavior (i.e., security) among 2–3-year-old children. This is important information as prior research has found a significant association between maternal reminiscing styles or maternal co-construction skills (Lu et al., 2018) with children's attachment security (Etzion-Carasso and Oppenheim, 2000; Gini et al., 2007; Lu et al., 2018). To our knowledge, the only published study that did not reveal a significant association between maternal language input and child attachment security was conducted among younger preschoolers (Reese et al., 2019). In the longitudinal study by Reese and colleagues, child security was assessed at 15 months of age using the Strange Situation procedure, and maternal reminiscing style was assessed when children were 26 months old, similar to the age group investigated in the current study.

Thus, in light of past work, our null result would be consistent with an age-related process of maternal language input and child attachment behavior. It is possible that the association between maternal co-construction and child attachment security does not become more pronounced until later in early childhood. During the early preschool years, the integration between verbally mediated mental representations of attachment and behavior may not yet have occurred. It is likely that as children develop, they become better at utilizing verbal cues from parents and, thus, may benefit more from verbal exchanges (e.g., Hudson and Shapiro, 1991; Hudson and Sheffield, 1998). Older preschoolers with more advanced representational and language capacities are likely to increasingly partake in social communicative verbal exchanges, have more opportunities to actively participate in attachment-related conversations with caregivers, and develop a deeper understanding of relationships and social interactions, which in turn may influence their attachment behavior with caregivers. However, given the relatively small sample size and limited statistical power of the present study, this interpretation remains tentative. Replication with longitudinal design and among larger samples is needed to determine whether developmental processes, methodological factors, or other influences account for the pattern of findings observed here.

Limitations and future directions

Several limitations of the current study should be noted. Findings are based on a small sample recruited in a university town in the Midwest, which led to a sample with overall high levels of maternal education and relatively low diversity that likely impacted findings. For example, although the current study did not reveal child biological sex differences or race/ethnicity differences in maternal verbal co-construction skills, cultural and contextual differences have been related to parental verbal input in previous studies (e.g., Aznar and Tenenbaum, 2013, 2015; Burumlu-Kisa and Shin-Acar, 2020; Fivush et al., 2000, 2003). In addition, the relatively small sample size may have limited statistical power to detect small effects. Consequently, the absence of a significant association between maternal co-construction skills and child attachment security should be interpreted cautiously. Future research with larger and more diverse (ethnically, economically, and culturally) samples is needed to determine whether the null findings observed in the present study reflect developmental processes, limited statistical power, or other factors.

Another important consideration concerns children's language abilities. Emerging expressive and receptive language skills likely influence both children's participation in attachment-related storytelling interactions and mothers' conversational scaffolding strategies during co-construction. Although child age was included as a covariate to partially account for developmental variation, no direct standardized assessment of child language ability was available in the current study. Furthermore, many children in this young sample produced minimal verbalizations during the storytelling task, making it difficult to derive a reliable proxy measure of expressive language from the observational data. Future studies, particularly those involving older preschool-aged children, should include developmentally appropriate assessments of child language ability to better disentangle the role of children's linguistic competence in maternal co-construction processes and attachment-related outcomes.

The assessment of maternal co-construction skills employed here also requires further study. Both the Joint Storytelling Task and the co-construction coding system were modified in this study, compared with the study design of Lu et al. (2018). In the original Joint Storytelling Task (Posada and Waters, 2018b), the attachment-related stories were presented on panels placed facing the mother-child dyad and they only had to co-construct two stories. In this study, the stories were presented as wordless books that were given to the mother-child dyad directly and dyads created 4 stories. Further, the setting where information about maternal co-construction skills was collected is another potential factor that needs to be considered. Both maternal sensitivity and child security were assessed concurrently in naturalistic settings, while information about maternal co-construction skills was collected in the laboratory. Those modifications could have affected our findings compared to those of Lu et al. (2018), who used both the original Joint Storytelling Task and coding scheme on information gathered during a home visit (along with assessments of sensitivity and security concurrently during that visit). In addition, the lab-based storytelling task may have captured mothers' narrative scaffolding abilities somewhat differently than attachment-related conversations occurring spontaneously in naturalistic settings. Verbal exchanges between mother and child occur daily in naturalistic settings, rather than in a lab; future studies should consider collecting maternal verbal input information during real-time interactions in naturalistic settings as well.

It is also possible that individual characteristics influence the ways mothers and children co-construct narratives and future work should account for this. For example, child temperament was found to be related to maternal elaborative reminiscing in past work (e.g., Bird et al., 2006; Lewis, 1999). Mothers may show more effort and adopt different strategies to co-construct with children who are more interested, persistent, sociable, and active, and thus child characteristics may moderate how maternal verbal skills relate to child security. Mothers' own scripted attachment representations have also been found to relate to their parenting practices and behaviors with children (e.g., Waters et al., 2018; Huth-Bocks et al., 2014; Groh et al., 2015; Groh and Haydon, 2018). Further, child's own language proficiency may also influence how mothers co-construct narratives with their children, as they may be more verbal when narrating with children who are more verbal. Future studies should include such variables and examine their potential influences.

The main aim of this study was to examine the role of maternal co-construction skills and associations with the quality of early child-mother attachment relationships in toddlerhood. Results suggest that skills such as creating a co-construction atmosphere and providing explanatory frameworks when having attachment-related conversations with their toddlers are significantly associated with maternal sensitivity. However, maternal co-construction skills were not found to be related to toddlers' attachment security as inferred from children's secure base behavior at this age. Findings highlight that a mother's ability to create a co-construction atmosphere and provide explanatory frameworks during attachment-related conversations are relevant factors that sensitive mothers engage in when scaffolding their children's understanding of attachment relationships, even though such skills didn't translate into children's use of their mother as a secure base. The use of longitudinal designs and inclusion of larger and more diverse groups would help to further elucidate features of maternal co-construction skills that may be particularly germane to understanding children's attachment security outcomes across childhood.