Studying family-emotion dynamics requires a research design that balances ecological validity, multilevel emotion measurement, and advanced analytical methods. Each of these components is essential for capturing the complex and interconnected emotional interactions that occur within whole-family systems.

Ecological validity refers to the extent to which research findings can be generalized to real-world settings [21]. Ecological validity is essential for understanding family-emotion dynamics because emotional interactions within families are influenced by complex and context-specific factors that may not manifest in artificial or controlled environments such as a research laboratory. Knowing that emotional processes are shaped not only by the members themselves but also by the surrounding environmental context, studying families in their real everyday environments yields meaningful information. EMA strengthens ecological validity by facilitating real-time data collection of participants’ experiences in their natural environments, such as at home or during daily routines [21,22]. The real-time assessment of emotions reduces retrospective bias and enhances the accuracy of reported emotions [23,24]. Furthermore, EMA can offer an additional layer of context by incorporating questions about participants’ current activities, social interactions, and environmental factors, offering deeper insight into emotional experiences.

Emotions are complex responses to specific events or stimuli, encompassing subjective experiences, physiological reactions, and behavioral responses [25]. Thus, to fully understand family-wide emotion dynamics, it is essential to examine emotional experiences across multiple levels. Many studies focus on the experiential aspects of emotion through self-report measures, which provide valuable insights into individuals’ subjective emotional states. However, self-reports alone cannot capture other critical components of the emotional experience, such as physiological responses. Physiological aspects of emotion reflect internal processes regulated by the autonomic nervous system, including changes in HR and EDA (ie, bodily perspiration), both of which are strongly linked to emotional arousal, making them essential metrics for assessing emotional responses [26-29]. Measuring physiology in response to emotional experience necessitates the use of equipment capable of capturing the subtle, dynamic changes regulated by the autonomic nervous system. In-laboratory studies permit a wide range of tools used to collect physiological signals through sensors or electrodes placed on the body. However, capturing physiological responses in everyday life requires wearable ambulatory equipment that is both portable and nonintrusive. These devices must allow participants to carry out their daily activities while continuously recording physiological data. Examples of these types of technologies include wearable wristbands that either measure HR and EDA or just HR, as well as chest straps that measure HR. When studying emotion across families, it is crucial that assessment tools and equipment are suitable for individuals at different developmental stages, with equipment needing to be validated for use across age groups and appropriately sized for children [30]. A key advantage of ambulatory physiological monitoring is that it can be combined with EMA surveys, enabling the concurrent measurement of subjective emotional experiences and physiological arousal. In a family-wide context, fully understanding emotional processes requires the simultaneous collection of data from all family members to capture dynamic, real-time interactions. This approach allows researchers to investigate how emotions influence or align across family members. To achieve this, active participation from each family member is crucial, including completing emotional self-reports and having their physiological responses monitored.

The approaches described above yield numerous data points (eg, repeated surveys each day for several days), at multiple levels of measurement (eg, emotion self-report and physiology), for several family members, resulting in multiple levels of nested data captured repeatedly over an extended time frame. Thus, the analytic approach used for this type of data must account for the nested and interdependent nature of the data, as well as the longitudinal, repeated measurement design. Further, as with any intensive data collection design, approaches to missing data should be considered carefully. Other factors to consider include the synchrony between different family members’ reports, synchrony in emotion reports (self-reported) and physiology obtained passively through a wearable device, and the directionality of effects. Finally, a whole-family analysis needs an approach that can generate family-level and person-specific values to allow for further testing of individual differences in families (eg, how emotionally cohesive a family is) and family members (eg, the extent to which fathers influence the family unit). Taking all of these factors into account, we recommend a network analytic approach, specifically multilevel vector autoregressive modeling [31-33]. As will be described further in the analytic section, resulting family-emotion networks will provide a measure of density, or how connected the family emotion network is, and a measure of centrality for each family member, measuring how influential the member is within the whole family unit. By capturing the dynamic interplay of emotional and physiological processes within family systems, these networks provide a deeper understanding of how emotions are expressed and regulated during everyday family interactions. This approach offers unique insight into the emotional climate of the family and the roles individual family members play in shaping shared emotional dynamics. Given that emotion regulation is a key predictor of mental health and well-being [4], understanding how families influence each other’s regulation and emotion expression may offer insights into how family dynamics relate to psychological health and well-being.

Families begin by completing web-based questionnaires on Qualtrics (Qualtrics International Inc) sent via email, with a separate survey link provided for each family member. Once all participating family members have submitted the initial questionnaires, a research assistant drops off the study equipment (ie, phones, wristbands, chargers, and a laptop if needed) at the families’ home. To ensure proper use and care of the study equipment, families attend a training session with a research assistant via Teams (Microsoft Corp). During this session, children also complete additional tasks on the Teams call. The day following the training session, families begin the 7-day EMA and ambulatory monitoring period using the Time2Feel app (Troon Technologies) and Empatica E4 wristbands [37]. At the end of the study period, a research assistant picks up the equipment from the families’ homes, and family members complete final web-based questionnaires. In each family, one child aged 8 to 12 years is randomly selected to be the “focus child.” Figure 1 gives an overview of the participant timeline.

Each family member completes their own survey based on their role in the family (eg, one survey for each parent, one for the focus child, and one for the sibling). Separate links to these surveys are sent to each family member by email so they can complete their surveys independently. One of the parents provides demographic information about themselves, the other parent, and their children, including details such as age, gender, ethnicity, education level, occupation, medical diagnoses (physical and psychological), and household income. Both parents complete questionnaires that evaluate parenting practices, as well as their own and the focus child’s emotion regulation. Children complete questionnaires that assess emotional functioning, including emotion-regulation abilities. All survey items are presented in a fixed order, with no randomization or adaptive questioning. Page breaks are inserted after each measure, and the number of questions per page varies depending on the length of the measure. If a participant leaves a question blank, a prompt appears notifying them that some questions have not been answered, giving them the option to return and complete them or continue regardless. Once a participant advances to the next page, they cannot return to previous pages to change their responses. Completing parent questionnaires takes approximately 40 minutes, while the child questionnaires take about 20 minutes. A member of the research team reviews each survey for completeness before scheduling the training session. If entire scales are missing from a survey, the research team follows up by emailing the family to confirm whether the missing responses were intentional or due to an error. Duplicate survey entries are identified and manually reviewed using the participant ID. If a duplicate survey contains missing sections that are completed in another entry, the responses are combined. Otherwise, the first complete entry is used for analysis. Both completed and partially completed questionnaires are included in analyses, depending on the extent and pattern of missing data. If participants complete only part of the survey but provide responses to key measures relevant to specific research questions, their data are retained for those analyses. Completion time is monitored through Qualtrics, but no strict cut-off is used to exclude responses based on duration alone. Instead, potentially invalid submissions (eg, extremely short completion times coupled with missing data) are reviewed on a case-by-case basis to determine inclusion.

Parents and children participate in a training session via Teams, led by a graduate or senior undergraduate research assistant. During this session, the research assistant explains the EMA period, including how to use the Time2Feel phone app, and provides step-by-step instructions for setting up and pairing the Empatica E4 wristbands with the app. Participants also calibrate their wristbands with the app to establish baseline measurements for HR rate and EDA. The focus child then completes a computer-based cognitive reappraisal task (CRT) [38].

To assess the feasibility and usability of the EMA protocol, engagement and compliance with Time2Feel for each family member is calculated as the percentage of completed surveys out of the total survey prompts received during the 7-day EMA and ambulatory monitoring period. A survey is considered complete if all questions are answered, while incomplete responses have at least one missing answer. Separate response rates are calculated for both random and deviation surveys.

After the 7-day EMA and ambulatory monitoring period, each family member completes the satisfaction and ease-of-use and subscales from the Usefulness, Satisfaction, and Ease of Use questionnaire to assess their experiences using the Time2Feel app [45]. The satisfaction subscale includes 4 items (eg, “I would recommend it to a friend”), while the ease-of-use scale consists of 9 items. Originally, the ease-of-use subscale contained 11 items (eg, “Using it is effortless”); however, items 7 and 8 were excluded due to low factor loadings (≤0.20) [45]. Gao et al [46] reported that the satisfaction and ease-of-use subscales had high internal consistency (αs from 0.88 to 0.95). Participants are also asked one open-ended question, inviting them to share any additional thoughts or experiences with the Time2Feel app.

During the 7 days of the EMA and ambulatory monitoring period, families report on their emotion experiences using the Time2Feel app. Resulting variables include the specific emotion experience, the intensity with which the emotion is experienced, the activity the family member is engaged in at the time, and any individuals the family member was with at the time of the survey. Over the 7-day period, each family member’s reported emotions are categorized as positive or negative, and their intensity ratings are used to calculate overall positive and negative affect scores. These scores are used to estimate family-wide emotion dynamics.

HR and EDA are continuously recorded during study hours throughout the 7-day EMA period using the Empatica E4 wristband. Changes in blood volume pulse are used to estimate HR, while skin conductance levels are derived from EDA. Data are imported into AcqKnowledge version 5.0, and guidelines for quantifying HR and EDA are followed [36,47]. Physiological data are synchronized with EMA survey responses to examine physiological states during self-reported emotional experiences. For both HR and EDA, time-series data are segmented into 10-second increments using AcqKnowledge 5.0. Intervals within ±5 minutes of survey responses are averaged. This procedure results in 28-56 HR and EDA data points per family member, which are used to estimate family-wide emotion dynamics. These physiological indicators contribute to our understanding of how emotional states are experienced and coregulated across family members in daily life, addressing the study’s primary aim of capturing the emotional processes that influence children’s psychological functioning.

Family-wide emotion dynamics are indexed through two primary variables derived from the family emotion networks (see analytic plan): density and centrality [31].

Density is a measure of the overall connectivity of the network, reflecting how closely family members are linked through emotional and physiological states. It is calculated by averaging the strengths of all connections between family members. Each connection, or “edge,” represents the strength of the shared emotional and physiological state between two members. Each family will receive a positive and negative emotion density score, representing the intensity of these emotions and their physiological states. Research highlights the significance of density, showing that denser networks are more ingrained and may therefore be more resistant to change [48]. Higher density scores indicate stronger and more consistent connections between family members’ emotions.

Centrality is a measure of how central, or important, each family member is within the broader family network. It is calculated by summing the in-strength (ie, the influence from other family members toward the individual) and the out-strength edges (ie, the influence exerted by the individual toward other family members). Each family member will have a centrality score, which will indicate how impactful this family member is within the whole family. A higher centrality score suggests greater importance.

Parents’ beliefs about their families’ strengths, abilities, and competencies in managing challenges and maintaining healthy functioning are assessed using the Family Functioning Styles Scale (FFSS) as part of the pre-EMA questionnaires [49]. The FFSS asks parents to indicate to what extent their family is characterized by 26 different qualities on a 5-point scale ranging from not at all like my family to almost always like my family (eg, “we generally agree about how family members are expected to behave”). The FFSS has been reported to have high internal consistency [49]. This measure is included to assess patterns of family functioning that may shape family-emotion networks, directly supporting our secondary objective of identifying family-level characteristics that influence the emotional dynamics observed during the EMA period.

Parents complete the Coping with Children’s Negative Emotions Scale (CCNES) during the pre-EMA questionnaires [50]. This is a self-report measure that evaluates how parents respond to their children’s negative emotions. The CCNES presents 12 vignettes describing hypothetical situations in which their child expresses negative emotions (eg, “if my child falls off his/her bike and breaks it, and then gets upset and cries, I would...”). Parents rate the likelihood of each response on a 7-point scale (1=very unlikely to 7=very likely) across six response types. These responses are categorized into six subscales: Minimizing Reactions, Punitive Reactions, Distress Reactions, Expressive Encouragement, Problem-Focused Reactions, and Emotion-Focused Reactions. Consistent with previous research, we use a version with the six response types combined into two overall scores: Supportive practices (Expressive Encouragement, Problem-Focused, and Emotion-Focused reactions) and Unsupportive practices (Punitive, Minimizing, and Distress reactions) [51]. This measure demonstrates good internal consistency and test-retest reliability, along with sufficient construct validity [50]. The CCNES is used to assess whether parent responses to children’s negative emotions contribute to family emotion dynamics and influence child outcomes.

Parents complete the Parents’ Gendered Emotion Beliefs Scale (PGEB) during the pre-EMA questionnaires [52]. The PGEB is a 15-item questionnaire that assesses parents’ gendered beliefs about emotion expression, gender-neutral beliefs, and gendered emotion socialization (eg, “I don’t think boys should cry because it shows weakness”). Responses are rated on a 4-point scale, from strongly disagree to strongly agree. The PGEB has shown good internal consistency for both the overall scale and each of the three subscales [52]. The PGEB is used to examine how parents’ gendered emotion beliefs may shape family emotion patterns.

Parents self-report on how often they engage in six emotion regulation strategies to manage negative emotions on the Regulation of Emotion Systems Survey (RESS) during the pre-EMA questionnaire [53]. This measure consists of 48 items using a 5-point scale ranging from never to always. Each of the six strategies—distraction, rumination, cognitive reappraisal, suppression, engagement, and relaxation—is assessed by 8 items each. Each subscale demonstrates high internal reliability [53]. Understanding parents’ regulatory practices supports the objective of identifying individual and family-level factors that contribute to family emotion dynamics and child psychological outcomes.

In the pre-EMA questionnaire, parents report on their tendency to react frequently and intensely to negative emotions, adapted from the 7-item Child and Adolescent Disposition Scale (CADS) [54]. The measure is altered to have parents reflect on their own emotional behavior over the past 12 months and indicate how often negative emotionality occurred (eg, “moods change unpredictably”), using a scale from 4-point scale (not at all to very much/very often). The CADS questionnaire shows high internal consistency [54]. This measure is used to identify how parent emotionality influences child outcomes (eg, child emotionality and psychopathology).

In the demographics section of the pre-EMA questionnaire, parents are asked if they currently have a diagnosis for a psychological condition, and if so, what they have been diagnosed with, and at what age they were diagnosed. They are then asked to rate the severity of their condition from not severe—I rarely or do not experience any symptoms to extremely severe—I always experience symptoms. Additionally, parents fill out the Kessler 10 (K-10) during the pre-EMA questionnaire, which assesses 10 symptoms of anxiety and depression (eg, “how often did you feel tired out for no good reason?”) on a 5-point Likert scale (none of the time to all of the time) [55]. The K-10 had been found to be a reliable measure in assessing symptomology in adults [56]. Parent psychopathology is assessed to support the study’s goal of identifying parent-level factors that shape family emotion networks and child outcomes.

Parents respond to the 17-item Parental Stress Scale (PSS) during the pre-EMA questionnaires [57]. The PSS measures perceived stress related to parenting (eg, “In the last month, how often have you been upset because of something that happened unexpectedly?”) on a 5-point scale (strongly disagree to strongly agree). Adequate internal consistency of the PSS was found across a range of samples [57]. Assessing parental stress supports the study’s aim of identifying parent-level factors that may shape family-emotion networks and influence children’s emotional functioning.

Child emotion regulation is measured through parent-report, child self-report, and a computerized CRT. These measures are described below.

Parents complete the RESS questionnaire for the focus child during the pre-EMA questionnaire [53]. Focus children also self-report their own strategy use during the pre- and post-EMA questionnaires.

Parents report on the focus child’s self-regulation and ability to manage their emotions using the Emotion Regulation Checklist (ERC) at pre- and post-EMA. The ERC is a 24-item parent-report measure consisting of two subscales: emotion regulation and emotional lability and negativity [58]. Items are rated on a 4-point scale (rarely or never to almost always). The ERC has been validated for children aged 6 to 12 years and demonstrates acceptable internal consistency for both subscales [58].

The focus children’s effectiveness in using cognitive reappraisal, an emotion regulation strategy known to reduce negative affect, is measured by their scores on the CRT [38]. This task is completed on Inquisit Web (Millisecond Software), a web-based platform, during the training session. In this task, children are trained to look and think about negative and neutral images of people and objects in different ways. When shown the “look” cue, children are instructed to observe the images naturally, without attempting to reappraise their emotions. When shown the “far” cue, children are trained to reappraise their emotions by either imagining the image as being far away from them or by imagining the image as being unreal. Following the cue and the image, children rate their negative emotion intensity on a feelings thermometer ranging from 0=felt no negative emotion to 3=felt a big amount of negative emotion. This task is comprised of a practice phase followed by a test phase consisting of 45 negative or neutral images, divided into three blocks with breaks in between. Image and cue types are counterbalanced and randomized across trials, resulting in three trial conditions: far negative (reappraisal), look negative (nonregulation), and look neutral (nonemotional). Cognitive reappraisal effectiveness is calculated as the percentage of negative affect reduced by reappraisal on far-negative trials compared to look-negative trials ([Look Negative Rating–Far Negative Rating]/Look Negative Rating)×100).

Child emotion regulation measures, including the RESS, ERC, and CRT, are used to examine how aspects of family functioning and parent characteristics (eg, psychopathology, emotionality, and perceived stress) relate to children’s emotion regulation, a key component of psychological functioning and well-being.

During the pre- and post-EMA questionnaires, both parents and the focus child report on the focus child’s propensity to react frequently and intensely to negative emotions on the 7-item CADS [54]. They are asked to reflect on the child’s emotional behavior and indicate how often these behaviors occurred in the past 12 months (eg, “moods change unpredictably”) using a scale from 1=not at all to 4=very much/very often. The CADS presents acceptable internal consistency for both parent-report and child self-report [54]. Additionally, the CADS is used as an outcome measure to examine whether family emotion dynamics are associated with changes in children’s emotion reactivity over time, supporting the study’s aim of understanding how family-level processes relate to children’s emotional functioning.

In the demographic section of the pre-EMA questionnaire, parents are asked if either of their children that are participating in the study currently have a diagnosis for a psychological condition and if so, what have they been diagnosed with, at what age were they diagnosed and then asked to rate the severity of their condition from 0=not severe—he/she rarely or does not experience any symptoms to 3=extremely severe—he/she always experiences symptoms.

Parents report on the focus child’s behavior and the focus child self-reports on their own behavior using the Brief Problems Monitor (BPM) as part of the pre- and post-EMA questionnaires [58]. The BPM includes 19 items rated on a scale from 0=not true to 3=very true, which measures internalizing problems (eg, “feels worthless or inferior”), attention problems (eg, “acts too young for his/her age”), and externalizing problems (eg, “argues a lot”). The internalizing and attention subscales each consist of six items, while the externalizing subscale contains seven items. Subscale scores are calculated by summing item ratings within each domain. A total problems score is computed by summing all 19 items. Both subscale and total scores are converted to standard T scores based on age- and gender-specific norms. T scores below 65 are considered within the normal range, while scores above 65 indicate clinically elevated concerns. Each subscale has been shown to display acceptable test-retest reliability correlations and internal consistencies [59]. Measures of psychopathology are included as outcome variables to examine how family emotion dynamics relate to children’s emotional functioning, directly addressing the study’s primary objective.

Simulations in triadic family network models with as many as 12 nodes demonstrate reliable network estimation with at least 59 families and as few as 56 time points [33]. Nodes are observed variables in the network, such as mother positive emotion, father positive emotion, child positive emotion, sibling positive emotion, mother HR, and father HR. Fewer nodes would require fewer families and timepoints. Published benchmarks indicate that temporal networks require more than 75 time points to achieve reliable precision and sensitivity and can accommodate up to 25% missing data [60]. The FACES study design provides between 28 and 56 time points per primary outcome for each family member with an estimated sample size of 70 families, or 280 individual family members. This design considers both data requirements and the need to minimize participant burden for children as young as 7 years [60].

To address study aims, the analytic plan involves a network analytic approach, specifically multilevel vector autoregressive modeling using the mlVAR package in R. The analytic plan includes three stages: (1) preparing the data, (2) estimating networks and calculating the primary outcomes (ie, network density and centrality outcomes), and (3) examining their relation with secondary outcomes (eg, individual emotion-regulation outcomes). This plan aligns with analytical approaches recommended by the developers of these methods and prior applications of network analysis to dyads and triads [33,61-63].

Descriptive statistics for all variables, compliance with EMA surveys and ambulatory physiology wear times, and mean satisfaction with the Time2Feel app (assessed via the Usefulness, Satisfaction, and Ease of Use questionnaire) are to be reported.

In terms of family emotion networks, mlVAR assumptions include equal time spans between timepoints (eg, EMA surveys), multivariate normality, and stationarity. To accommodate our pediatric sample, the FACES protocol features equal spacing within designated periods (eg, hourly from 4 PM to 8 PM on weekdays and 10 AM to 6 PM on weekends) but a longer overnight interval between the last survey of one day (eg, 8 PM) and the first survey of the next (eg, 4 PM). To address this, the network model excludes estimates of relationships between evening and morning surveys. Normality is assessed using the Kolmogorov-Smirnov test. Data are detrended to ensure stable means, variances, and autocorrelations over time. Then the Kwiatkowski-Phillips-Schmidt-Shin unit root test is used to examine stationarity [31]. In instances that stationarity assumptions are not met, a semiparametric time-varying vector-autoregressive model is used instead [64].

Secondary outcomes (ie, RESS and CADS) are analyzed using multiple regression. Their distribution and normality are assessed by examining skewness and kurtosis, and outliers are identified as data points that deviate more than 3 SDs from the sample mean. Identified outliers are winsorized [65].

Temporal nomothetic (group-level) and idiographic (family-specific) networks are estimated using the mlVAR package and visualized with the graphicalVAR package in R [66,67]. These analyses produce Family-Wide Emotion Networks as shown in Figures 2 and 3. Figure 2 represents positive emotion-physiology dynamics within families, and Figure 3 represents negative emotion-physiology dynamics. Each node (circle) represents one of three indicators—emotion experience (positive or negative), HR, or EDA—for each of the four family members, resulting in 12 nodes per family (3 measures×4 individuals). The edges, or lines connecting the nodes, represent the effect of one family member’s state on another while accounting for the influence of other family members in the network [31]. Thicker edges indicate stronger effects, which can be excitatory (green) or inhibitory (red). Nomothetic networks show average temporal associations across all families and are used to describe group-level patterns [35]. Idiographic networks, which represent how emotions and physiological states interact within individual families over time, are used to calculate metrics of density and centrality.

Two family emotion density scores are calculated: one for positive emotion and one for negative emotion. The model also calculates in-strength and out-strength edges for each node. In-strength shows how much a node is influenced by others, while out-strength indicates how much it influences other nodes. Using these measures, eight centrality scores are calculated—one for each family member in both the positive and negative emotion networks. The reliability and significance of edges and centrality measures are assessed using CIs and bootstrap methods [66].

To test the primary objective of this protocol—how family emotion networks contribute to children’s functioning—multiple regression analyses are conducted in which child emotion regulation (ie, RESS, ERC, and CRT) and child psychopathology (ie, BPM) are regressed on family-wide density scores. To examine the secondary aim—identifying which family, parent, and child factors may influence family emotion networks—additional regression analyses are conducted in which family-wide density scores are regressed on family (ie, FFSS), parent (ie, CCNES, PGEB, RESS, CADS, K-10, and PSS), and child (eg, CADS) factors. To disentangle the contributions of specific family members to child outcomes, regression analyses are conducted in which child emotion regulation (ie, RESS, ERC, and CRT) and child psychopathology (ie, BPM) are regressed on each family member’s centrality score. This approach will clarify the specific contributions of individual family members to these child outcomes, accounting for their relative prominence and connectivity within the family emotion network.

All protocol amendments are submitted to the University’s Research Ethics Board for review and approval prior to implementation.

Any adverse events are to be reported to the University’s Research Ethics Board as per ethical guidelines.

Audits are conducted as required by the sponsoring institution’s Research Ethics Board.

Authorship eligibility will be determined based on substantial contributions to study conceptualization, design, analysis, and writing, based on the International Committee of Medical Journal Editors. Contributors who do not meet authorship criteria will be acknowledged. No professional medical writers will be used in the preparation of study-related publications.

The findings from this study will be disseminated through multiple channels to maximize visibility and impact within the scientific and broader public communities. Key dissemination methods will be through peer-reviewed publications, presentations are national and international conferences, and media outlets (eg, the lab’s official website, news interviews, and magazines).