Eye of The Beholder: Neural Synchrony of Dynamically Changing Relations Between Parent Praise and Child Affect

Abstract

In this study, we aimed to determine the role of parental praise and child affect in the neural processes underlying parent-child interactions, utilizing functional near-infrared spectroscopy (fNIRS) hyperscanning. We characterized the dynamic changes in interpersonal neural synchrony (INS) between parents and children (4 to 6 years old, n = 40 dyads) during a cognitively challenging task. We then examined how changes in parent-child INS are influenced by parental feedback and child affect. Parent-child INS showed a quadratic change over time, indicating a decelerated decline during the interaction period. The relationship of parental praise, in the form of positive feedback, to change in INS was contingent upon the child’s positive affect during the task. The highest levels of INS were observed when praise was present and child affect was positive. The left temporo-parietal regions of the child and the right dorsolateral prefrontal and right temporo-parietal regions of the parent demonstrated the strongest INS. The dynamic change in INS during the interaction was associated with children’s independent performance on a standardized test of visuospatial processing. This research, leveraging fNIRS hyperscanning, elucidates the neural dynamics underlying the interaction between parent praise and child positive affect, thereby contributing to our broader understanding of parent-child dynamics.

Introduction

Warm, responsive, and loving parent-child relationships play a pivotal role in children’s development, as demonstrated by both behavioral (e.g., Harlow, 1958; Groh et al., 2017; Ginsburg, 2007) and neuroimaging studies (e.g., Bick et al., 2016; Kopala-Sibley, 2022; Whittle et al., 2014). While the prior literature on the role of loving relationships often focuses on socio-emotional outcomes, these interactions also form the very fabric of children’s early learning experiences, serving as vital conduits for parental support in children’s problem-solving and exploration. Loving relationships are multifaceted, and include multiple components, one of which is praise. Praise, in the form of positive feedback, serves as a crucial link between parents and their children during learning interactions and has the potential contribute to children’s learning (Hattie & Timperley, 2007). However, despite the general positive perception of praise, the underlying mechanisms for why/or how the effects of praise are moderated in varied contexts or for specific individuals remain opaque. Here, we utilize fNIRS-based hyperscanning to elucidate the neural basis of parent-child interactions during a challenging problem-solving task. We investigate how the underlying neural dynamics unfold and serve as a function of interactive relations between parental praise and child factors.

Praise, as defined by Kanouse et al. (1981), encompasses the provision of positive verbal assessments regarding an individual’s accomplishments, actions, or attributes, based on the evaluator’s subjective standards. Multiple traditions, including theories on attachment, behaviorism, and cognitive traditions collectively emphasize the role of praise in children’s development (Troutman, 2015). Although some associate praise more strongly with a behavioral perspective and differentiate it from warm, responsive, and loving parent-child relationships that attachment theory emphasizes, both behaviorist and attachment traditions endorse positive parenting skills. These positive parenting skills include parents closely observing their children’s behavior (e.g., parents reflecting on child behavior) to create a supportive environment for the child, which often involves praise (Troutman, 2015). According to attachment theory, parents’ sensitive and responsive interactions allow the child to feel comfortable exploring independently, where the child knows that the secure base will be there if needed. (Bowlby, 1973). In this context, parental praise for children’s efforts to explore and problem-solve contributes to the secure base, a characteristic of secure attachment. Because praise communicates acceptance, it potentially plays a pivotal role in shaping a positive working model of secure attachment between child and caregiver (Bowlby, 1973; Stipek et al., 1992). These views overall align with a reinforcement perspective where praise may correspond to positive affect in the child and in turn be associated with positive developmental outcomes (Blumenfeld et al., 1982; O’Leary & O’Leary, 1977).

Some empirical evidence supports the role of praise and its’ positive effects on children. The impact of parental praise or other forms of responsive feedback is visible in the immediate context of parent-child interactions, but also extend to influence children’s motivation and behaviors both in the short term and over an extended duration, particularly during cognitively demanding tasks (Cimpian et al., 2007; Gunderson et al., 2013; 2018; Henderlong & Lepper, 2002; Kamins et al., 1999; Pomerantz & Kempner, 2013; Willingham, 2005). However, more recent evidence draws a more complex picture on the role of praise and better clarifies when, why and for whom praise is beneficial versus detrimental for children. More specifically, the role of praise might vary depending on various factors, including but not limited to the type of the praise given, excessiveness or sincerity of the praise, and children’s self-esteem (Brummelman et al., 2016; Gunderson et al., 2013; Henderlong Lepper, 2002).

Praise can indeed backfire. Referred to as the praise paradox, praise possesses the potential to lower children’s motivation in different contexts (Brummelman et al., 2016; Henderlong & Lepper, 2002; Mueller & Dweck, 1998). It is essential to recognize that the children themselves actively participate in the reception and interpretation of praise, indicating their agency in the process. While children commonly exhibit positive responses to praise, this association exhibits only modest strength (Stipek et al., 1992). Under specific circumstances, praise may evoke negative affect, for example when recipients interpret praise as an indication of low performance ability by the evaluator (Meyer, 1992). These findings are overall inconsistent with the reinforcement model of praise, which suggests that positive verbal feedback corresponds to positive affect.

In general, children’s subjective interpretation of praise may moderate its impact on their behavioral outcomes. This aligns with the transactional view of development, emphasizing the mutual influence between adults and children, where individual behaviors contribute to shaping their developmental context (Sameroff, 2010). Similarly, according to cognitive evaluation theory, the meaning given to praise by the receiver is a strong factor (Deci & Ryan, 1980). However, the precise processes underlying these transactional relationships within parent-child interactions remain unclear. Thus, the underlying mechanisms for why the effects of praise are moderated in specific contexts or for specific individuals remain opaque. Our main goal here is to determine how the role of praise might be moderated by child affect during a cognitive challenging task.

One barrier to our understanding of how praise works, specifically with respect to interactions between parent praise and child affect lies in the measurement of whether and how children process such praise. Previous behavioral studies predominantly relied on indirect measures to gauge children’s cognitive and affective responses to praise (Stipek et al., 1992), such as an independent self-report about self-esteem after receiving praise (Skipper & Douglas, 2012) or the observations of engaging time, enjoyment, or willingness to complete a following similar task (Mueller & Dweck, 1998). These studies typically assessed praise and children’s outcomes independently and conducted subsequent correlational analyses of separately collected measures. Thus, a gap in the literature exists regarding the interplay between children and parents, particularly in the context of praise, as these interactions dynamically unfold. The role of parental praise in real-time processes remains unclear. We aim to address this gap by utilizing neuroimaging measures to pinpoint the how parent praise and child factors jointly interact within the dynamic, moment-by-moment relations unfolding between parents and children.

Although behavioral approaches are powerful in highlighting significant associations between parent praise, child affect and child outcomes, neuroimaging studies can enhance these strengths further. Specifically, a relatively novel set of neuroimaging approaches that measures interactions between multiple partners holds promise to elucidate dynamic processes underlying the role of parental praise and child affect. Measuring two brains simultaneously, termed hyperscanning, has been successfully developed first using fMRI (Montague et al., 2002). Hyperscanning allows researchers to explore the synchronization of neural oscillations between multiple individuals engaged in various interactions – which is typically referred to as interpersonal neural synchrony (INS) (Czeszumski et al., 2020; Dumas et al., 2011; Wass et al., 2020). INS is measured by the coherence in time, space, and/or frequency dimensions between brain activity in two or more people (Cui et al., 2012; Jiang et al., 2012). In our study, our goal is to use functional near-infrared spectroscopy (fNIRS) hyperscanning, a non-invasive, portable, relatively low-cost neuroimaging method, to measure interpersonal neural synchrony (INS) between parents and children during a cognitively challenging task, i.e. tangram puzzle task.

While not considered a mechanism itself, INS is a measurable reflection of the neural alignment between partners (Wass et al., 2020). Recent theoretical frameworks characterize INS in terms of a mutual prediction hypothesis, where interacting partners continuously engage in social prediction to anticipate one another’s actions and mental states, which in turn increases neural synchronization. Mutual predictions are facilitated by subjective socio-emotional interpretations and behavioral communicative rhythms (Hamilton, 2021; Hoehl et al., 2021; Wass et al., 2020). The behavioral communicative rhythms are transmitted through the environment (via verbal and nonverbal cues or communicative signals) and these lead to the sensory system of one individual coupling with the motor system of another. Ultimately, this leads to synchronization of neural oscillations. This coupling occurs because the receiver reset their phases to the incoming oscillations from the sender; the sender’s and receiver’s brains entrain to the rhythm of the transmitted signal. Both the ostensive cues present during communication and the associated higher-order cognitive processes contribute to the underlying INS (Wass et al., 2020). In the current study, by utilizing hyperscanning, we aim to measure parent-child INS, i.e. the extent to which the neural activity of interacting individuals becomes synchronized, and test how this INS varies as a function of interactive relations between parental praise and child affect.

A growing body of work in adults supports the idea that neural activity between partners synchronizes during social interactions. For example, INS significantly increases in the frontal and parietal cortices during a cooperation condition compared to other conditions where partners work independently (Cui et al., 2012; Liu et al., 2016; Reindl et al., 2018). Two key brain regions underlying social interactions include the dorsal lateral prefrontal cortex (dlPFC) associated with problem-solving processes (Metuki & Lavidor, 2012; Ruocco et al., 2014), and the temporo-parietal junction (TPJ) involved in social mentalization processes (Li et al., 2014; Van & Mariën, 2016). Hyperscanning has been employed in studies of other forms of loving relationships in adults, like romantic relationships (Long et al., 2021). For example, the INS of romantic lovers increase in the right superior frontal cortex in a cooperation task when compared to a cooperation task with friends (Pan et al., 2017). This highlights the versatility of hyperscanning in capturing neural synchrony within diverse types of loving relationships.

Recent work leveraged hyperscanning to investigate the neural dynamics underlying parent-child interactions, but the primary focus has been on the socio-emotional aspects of these interactions. For example, relationships between different components such as INS, parental stress, child emotion regulation, and parent-child attachment have received attention (Miller et al., 2019; Nguyen et al., 2020; Reindl et al., 2018; Wass et al., 2020). A smaller number of recent studies successfully leveraged hyperscanning paradigms and highlighted children’s contribution in in learning contexts as well, such as in children’s observational learning (Zhao et al., 2021; Zhao et al., 2023). While earlier studies primarily focused on session-wide, averaged measures of interpersonal neural synchrony across interactions (Nguyen et al., 2020), emerging research suggests potential fluctuations in INS throughout interactions by segmenting the interaction session (Nguyen et al., 2021b). Finally, a growing body of work also indicated that INS has implications for the outcomes of interactions; for example, higher levels of INS during parent-child cooperative tasks predict improved task performance (Hoehl, Fairhurst, & Schirmer, 2021; Nguyen et al., 2020, 2021c; Reindl et al., 2018). Overall, hyperscanning offers valuable insights into the neural mechanisms underpinning parent-child interactions. Yet, our understanding of dynamic changes in parent-child neural synchrony during cognitive tasks, as well as the factors influencing these fluctuations, remains limited.

To the best of our knowledge, this rich emerging body of work on neural synchrony between parents and children remains largely disconnected from the extensive behavioral literature on the role of parental praise. Our study seeks to pinpoint the neural basis of the interaction of parental praise and children’s affective responses as the interactions unfold. Here, our primary objective is to synthesize research on parental praise, its’ possible interactions with child affect, and the growing field of parent-child hyperscanning. We examine the dynamic nature of interpersonal neural synchrony and its associations with multiple sources of variability during a challenging cognitive task. By integrating detailed behavioral coding of parental praise and child affect with fNIRS hyperscanning approach, we aim to identify the neurobiological underpinnings of interactive relations between parental praise and children’s affective responses as they ebb and flow during a cognitively challenging task. We used a cooperative tangram puzzle task to address our question. We picked the tangram puzzle task because it has been validated in both prior hyperscanning literature as well as in prior literature focusing on parent-child praise (e.g. Nguyen et al., 2020; Nguyen et al., 2021b, c; Ren et al., 2022).

Specifically, our goals are to examine: 1) how interpersonal synchrony (INS) during parent-child interactions dynamically changes during a cognitive challenging (i.e., tangram puzzle) task, 2) how the dynamically changing INS varies as a function of parental feedback and child affect, and 3) how the change in INS relates to children’s task performance. Based on recent work (Nguyen et al., 2021b), we expected the interpersonal neural synchrony to fluctuate throughout the interaction. We hypothesized that parental praise and positive child affect will predict both higher synchronization of parent-child interpersonal neural synchrony and possible dynamic changes in interpersonal neural synchrony (Cimpian et al., 2007; Gunderson et al., 2013; & Henderlong Lepper, 2002; Kamins & Dweck, 1999; Pomerantz & Kempner, 2013). Because prior literature suggests that children are not passive recipients of praise (Brummelman et al., 2016; Henderlong & Lepper., 2002), we also examined possible interactions between parental praise and child positive affect. Given the prior literature emphasizing the role of child affect (Blumenfeld et al., 1982; O’Leary & O’Leary, 1977), we predicted that relationships between praise and INS would be moderated by child affect. Specifically, we predicted that praise would be more strongly and positively related to INS when child affect is also positive. Finally, given prior work highlighting relations of synchrony for learning (e.g., Nguyen et al., 2020, 2021; Reindl et al., 2018), we expected higher interpersonal neural synchrony values overall to predict better performance on the cooperative tangram task with the parent and better performance on a standardized spatial task completed independently by the child.

Method

Participants

A total of 43 families enrolled as participants in the study, 40 dyads (80 participants) included in the subsequent data analysis. Three dyads were excluded from the analysis due to either incomplete session participation or technical issues. The final sample size is similar to other parent-child fNIRS hyperscanning studies in the literature (Miller et al., 2019; Nguyen et al., 2020; Reindl et al., 2018; 2022; Zhao et al., 2021; Zhou et al., 2023). Please see supplementary materials for further information on sample size and power analysis. The children involved in the study had an average age of 5.92 years (Range = 4.32 – 6.90 years, SD = 0.52, 22 males), while their parents had an average age of 38.75 years (Range = 25.21 – 53.74 years, SD = 5.47, 4 males). All parents who participated were primary caregivers based on parental report. All participants included in the study were typically developing, with no history of preterm birth, neurological disorders, developmental delays, or speech-language therapy. All the parents who participated in this study were white, and 97.56% were not Hispanic/Latino. 90.24% of the children who participated were white, and the remaining children were of mixed race. 90.24% of the children were non-Hispanic/Latino. Mean family income was $119,043, but exhibited variability (range= $18,000 – $340,000, SD = $53,506). The mean educational attainment of the parents was 16 years, corresponding to a college degree (range = 14 −18, SD = 1.18). We recruited parent-child dyads using the university hospital’s electronic health records, university mass emailing, social media, and word of mouth. This study was approved by the university’s institutional review board. All caregivers gave written informed consent for both themselves and their children. Children provided verbal assent before participating in the study. All families received a gift card for participating.

Procedures

The parent-child dyads participated in a two-hour laboratory session, where they engaged in a series of tasks while undergoing functional near-infrared spectroscopy (fNIRS) imaging. Subsequently, the children underwent standardized tests, while the parents completed multiple questionnaires about their child and family demographics.

fNIRS Task Design

During the fNIRS brain imaging, the parent and child sat on opposite sides of a table, face-to-face (see Figure 1). The participants were guided through five different tasks: a free conversation task (240 sec), the first individual tangram puzzle task (120 sec), a cooperative tangram puzzle task (240 sec), the second individual tangram puzzle task (120 sec), and a problem-relevant conversation task (240 sec). Our design was adapted from prior studies examining parent-child interactions (e.g., Nguyen et al., 2020). The order of tasks was consistent across participants. During the free conversation task, participants were instructed to engage in open dialogue discussing subjects of their choosing. During the individual problem-solving task, parents and their children were instructed to focus on their respective tangram tasks without engaging in verbal communication with one another and were separated from each other with a divider. For the cooperative tangram task, which is the main focus of the current paper, both caregivers and children were involved in the completion of a tangram puzzle. Specifically, they received instructions to collaboratively manipulate seven geometric pieces, with the aim of reconstructing seven predetermined templates depicting abstract forms, objects, and animals. During this cooperative section, participants were allowed to freely interact, pooling their efforts to solve the problems jointly (see Figure 1). During the problem-relevant conversation task, they were instructed to discuss the tangram tasks they previously completed. Given the current study’s purpose on the role of parental feedback and child affect during cognitively challenging tasks, we focused on the cooperative task and excluded other tasks from the current paper (see Figure 2). All sessions were video recorded utilizing a dual-angle camera system, enabling recording from two different perspectives.

Figure 1.

Illustration of the five tasks the dyads completed during the laboratory visit. a. Free conversation (240 seconds); b. First individual tangram task (120 seconds); c. Cooperation tangram task (240 seconds); d. Second individual tangram task (120 seconds); e. Task conversation (240 seconds). The focus of the current study is on Cooperation task.

Figure 2.

Illustration of cooperation during tangram task. The dyad shared one tangram booklet and worked together complete the picture shown on the booklet using 7 tangram pieces.

fNIRS Data Acquisition

Data was collected from both participants simultaneously using two NIRSport2 optical topography systems (manufactured by NIRx Medizintechnik GmbH, Germany) with one device allocated to the parent and one to the child. These systems were employed to capture imaging data in real time. Eight light sources and eight detectors were grouped into four 2 × 2 probe sets, which resulted in a total of 16 channels covering the prefrontal cortex and temporoparietal junction of the left and right hemispheres. The distance of each channel, comprised of one source and one detector, was equal to 3 centimeters. The fiber setup was securely attached to standardized EEG caps with 10–20 configuration system (see Figure 3). The probe localization was established and applied consistently for each participant using the international 10–20 system positioning; Fz, Cz, and pre-auricular were measured for each participant. Head circumference, ear to ear, and inion to nasion measurements were leveraged for consistency of cap placement (Jurcak et al., 2007). The data was acquired using the Hyperscanning software (v2014 NIRx Medical Technologies LLC) at two near infra-red light wavelengths (760 and 850 nm), at a sampling rate of 10.17 Hz. The beginning and end of each task were identified using triggers during the data collection process which were connected to fNIRS data acquisition software.

Figure 3.

Illustration of fNIRS probe configuration. a. Probe placement on 10–20 system. Regions of interest (ROIs). S represents the sources, and D represents the detectors. The number after the capital letter represents the order of the fibers. Lines between fibers represent the channels. Each region of interest has 4 channels. The brain regions corresponding to each ROI are marked below ROI number. b. Frontal view with probe configuration. The red dots represent the sources, and the blue dots represent the detectors. Green lines between fibers represent channels of signal. Each channel contains one source and one detector. in c (left hemisphere) & d (right hemisphere).

Based on the cranio-cerebral correlation demonstrated by Okamoto et al. (2004), when the fNIRS sensors are placed according to the international 10–20 system, the locations of 10–20 cortical projection points in the standard Montreal Neurological Institute (MNI) or Talairach space fluctuate with an average standard deviation of 8 mm, allowing reasonable spatial resolution at a gyrus level. The selection of brain areas covered by the cap in this study was informed by previous fNIRS hyperscanning research on interpersonal communication and problem solving (Jiang et al., 2012, 2015; Nguyen et al., 2020, 2021a, 2021b). Probe sensitivity profile was computed with the Atlas Viewer toolbox (Aasted et al., 2015) of Homer2 Software (Huppert et al., 2009). AtlasViewer provides wavelength-specific sensitivity maps of photon propagation for each optode pair that form a channel. In Figure 4, the optical sensitivity profile of the probe is mapped on a standard brain template. The sensitivity of the probe for detecting brain hemodynamics is calculated with the method explained in Aasted et al. (2015) and is presented as a heat map ranging from 0 to −2 in log10 units. The channel positions in the standard MNI coordinate space were determined also using AtlasViewer and are presented in supplementary materials, including information on the location of light sources and detectors according to the international EEG 10–20 system for electrode placement. The areas identified overlap with the areas identified in recent neuroimaging studies focusing parent-child hyperscanning (Nguyen et al., 2020; 2021b).

Figure 4.

Probe configuration and photon sensitivity profile for a representative subject scan. Channels are formed between adjacent source-detector pairs and represented with yellow lines and labeled accordingly. Sensitivity profile for photon propagation depth is presented as a heat map in log10 (mm⁻¹). units.

fNIRS Data Analysis

Preprocessing and first-level analyses.

The collected data underwent pre-processing using the AnalyzlR toolbox, an open-source MATLAB-based software developed by Santosa and colleagues (2018). The initial raw data acquired from the optical topography systems was in the form of optical density values. To convert the raw data into meaningful hemoglobin values, the modified Beer-Lambert Law was applied. For path-length selection, we used the default value of AnalyzIR toolbox (Santosa et al., 2018), since prior developmental studies including children as young as 1 year of age similarly used default values (Morgan et al., 2023; Kerr-German et al., 2022), and because wanted to keep the same value for both participants prior to dyadic neural synchrony analysis. Subsequently, the converted data were subjected to a wavelet filter, which effectively eliminated outliers and low-frequency drifts. The pre-processed data then underwent principal component analysis (PCA) with an 80% threshold of variance, as outlined by Zhao et al. (2021). This step aimed to remove motion-related artifacts and global physiological noise by extracting principal components associated with the majority of the data’s variance. These steps sought to mitigate both false-discovery rates resulting from serially correlated noise induced by physiological processes in NIRS and outliers associated with motion artifacts. For subsequent analysis, we focused on HbO values, as they have been recognized as being more responsive to variations in the regional cerebral blood flow (Hoshi 2016; Miller et al., 2019; Reindl et al., 2018) and more frequently used in hyperscanning studies (Cui et al. 2021; Jiang et al., 2012).

Pair-level analyses - Wavelet Coherence Analysis.

Following individual pre-processing, the data from one partner in each dyad was paired with the data from the other partner for pair-level analyses. Interpersonal neural synchrony was quantified using the Wavelet Transform Coherence (WTC) algorithm, implemented in MATLAB through the ‘wcoherence’ function (Grinsted et al., 2004), considering prior literature leveraging WTC scores as indicators of interpersonal neural synchrony (Cui et al., 2012; Jiang et al., 2012). The WTC algorithm enabled the examination of the relationship between the fNIRS time series of individuals within each channel, considering both frequency and time. Each channel pair (16×16 channels in our dataset) yielded a 2-D matrix of coherence values, representing the coherence between the two participants. This matrix provided information for each time frame (reflecting time series data) and frequency (representing signal distribution). Consequently, the WTC analysis generated a 4-D matrix encompassing all channels from participant 1 (16), all channels from participant 2 (16), frames representing time (varying across different task sections), and frequencies (varying across different section durations).

Consequently, our analyses yielded a Wavelet Transform Coherence (WTC) matrix representing interpersonal neural synchrony during the cooperation section. We focused on 0.02–0.1 Hz interval as our frequency range. This decision was based on prior reviews and empirical papers in the literature, specifically in parent-child hyperscanning studies (Hakim et al., 2023; Nguyen et al., 2020), which considered this band as task relevant. This frequency band excluded high-frequency noise components associated with respiration (~0.3 Hz) and cardiac pulsation (~1 Hz) and low-frequency noise associated with drift (Pinti et al., 2019). Please see supplementary materials for more detailed information on frequency band selection. As a result, the WTC matrix was transformed into a 3-D matrix, where the dimensions represented the channels from both participants and the time series information. WTC scores for each dyad was then converted to Fisher’s z-scores to facilitate subsequent statistical analysis encompassing all dyads involved (Baker et al., 2016).

To mitigate the risk of spurious correlations and to ensure specificity of the WTC findings, we also conducted a shuffled random pair WTC analysis involving 40 dyads. Two participants from different families were selected for this analysis. By comparing the outcomes obtained from these shuffled random pairs with those of our regular family pairs, we aimed to pinpoint the role of synchronization during familiar dyads (Jiang et al., 2012; Zhao et al., 2021). Comparing synchrony during real partners to shuffled data has been emphasized in recent papers on parent-child synchrony. This approach is preferred to address multiple confounds that come into play when comparing cooperation conditions to typical rest, including differences in the visual set-up, absence of a shared task, or changes in autonomic nervous system (Nguyen et al., 2020; Nguyen et al., 2021a, 2021; Reindl et al., 2018; Tachtsidis & Scholkmann, 2016).

To investigate spatial distribution of WTC, we partitioned the 16 channels into four distinct regions of interest (ROI), based on their respective coverage areas (see Figure 3). These four ROIs were specifically identified as the left dorsolateral prefrontal cortex (left-dlPFC), left temporoparietal junction (left-TPJ), right dorsolateral prefrontal cortex (right-dlPFC), and right temporoparietal junction (right-TPJ). For ROI analysis, we computed the average WTC value of the four channels within each ROI, resulting in a single WTC value per ROI. This results a 4×4 ROI matrix representing the four ROIs at each frame – four values, one for each ROI, per individual for each frame in the temporal dimension.

Child Spatial Performance

To assess children’s independent spatial performance, the Wechsler Preschool and Primary Scale of Intelligence (WPPSI-IV) (Wechsler, 2012) Block Design subtest was administered. This subtest measures children’s visuospatial constructive processing. Children are presented blocks with surfaces of solid red, solid white, and surfaces that are half red and half white. Then they are asked to replicate a pattern the experimenter presents to them, first looking at a physical block model as a guide and later referencing a two-dimensional picture. Children’s scaled scores from this subtest were used for analysis.

Behavior Coding

Tangram Performance.

At the end of the cooperative task, the total number of templates completed by the dyad was tallied as a measure of task performance.

Parental Verbal Feedback.

Based on the video recording of tangram interactions, we coded parental verbal feedback. For verbal feedback coding, we first transcribed all parental speech. The unit of transcription was the utterance, defined as any sequence of words preceded or followed by a change in conversational turn, intonation, or a pause. We then systematically analyzed all task-relevant utterances initiated by parents. Task-irrelevant utterances included utterances that were not about the tangram activity, e.g. “it is hot in here”. The identified relevant feedback utterances were further categorized into three distinct feedback subcategories: positive feedback, neutral feedback, and negative feedback. Positive utterances (praise) were statements involving positive attitudes about children completing and progressing on the tasks as well as encouragements to continue. These includes positive feedback or reinforcement to the child (their verbalizations or actions), where the positive content could be explicit (e.g. “Good job!”, “You’re really good at this”) or positively affirm child’s behavior (e.g. “You rock!”) (Carver et al., 2022; Gunderson et al., 2013). Neutral consisted of general, objective feedback utterances without a specific praise component and did not influence the children’s progress in a positive or negative way (e.g. “Okay”). Negative feedback included utterances with negative valence or negative evaluation of child’s verbalization or action, resembling corrections (e.g., “That’s not right”, “Don’t do that”) (Carver et al., 2022). To have a complete picture of all task-relevant language parents provide during the interactions and to establish the specificity of the parental verbal feedback, specific comments about the task itself were also marked. These specific, task-relevant utterances were divided into two groups: probe utterances (e.g., “Do you think this piece goes here?”), and informative utterances (e.g., “The green one touches the edge). An independent coder coded 25% of the data and reached 89% agreement on the categories assigned. All disagreements were resolved through discussions. Please see supplementary materials (S4) for more detailed informed on feedback coding.

Child Affect.

Based on videos, we coded the expression of positive and negative affect during interaction. The coding of affect was based on facial expression, tone of voice, and/or body language. Positive affect referred to the expressions of joy, pleasurable engagement, high energy, and full concentration. Examples of child positive affect included a smile, laughter, or giggle, singing, or excited/enthusiastic tone of voice. Negative affect referred to the expression of emotional distress and unpleasurable engagement (Lunkenheimer et al., 2013; Watson et al., 1988). Examples of child negative affect included whining, frowning, a heavy sigh, slouching or slumped shoulders, a furrowed brow, or placing their head on their hands or on the table indicating that they are frustrated or not interested in the task. For a comprehensive description of the procedures employed for affect coding, please refer to the supplementary materials (S5). An independent coder coded 25% of the data and reached 91% agreement on the categories assigned. All disagreements were resolved through discussions.

Combination of the behavior coding data and fNIRS data

To analyze the dynamic changes in the WTC data and align the WTC data to behavioral coding, the WTC data during 240-second cooperative section was partitioned into 10-second fragments, resulting in a total of 24 10-second time windows per dyad. While the prior literature often averaged WTC across the time dimension (Nguyen et al., 2020, 2021; Miller et al., 2019; Reindl et al., 2018), we utilized multiple shorter time windows to examine dynamic changes in WTC and to pinpoint changing relations to behavior throughout the interaction. Given our frequency range and sampling rate, optimal estimated duration needed for WTC analyses is between 10 to 50 seconds (0.1 Hz – 0.02 Hz). We opted for the shorter epochs which provide better temporal resolution (Marzoratti & Evans, 2022; Percival & Walden, 2000). The 10-second time window was applied to the behavioral data, mirroring the fNIRS partitioning. The frequency of occurrence for each feedback subcategory (positive, negative, neutral, probe, informative) within the given time window was tallied. For example, we counted how many instances of positive feedback utterances were present in each 10 second window. Similarly, we tallied the occurrence for each affect category (positive, negative) within the same 10-second time window. Thus, each dyad possessed 24 distinct 10-second windows encompassing scores for interpersonal neural synchrony, the occurrence counts for each feedback subcategory and affect subcategory. Stated differently, for each 10-second time window, the WTC values, the number of utterances in each feedback subcategory, and the number of positive and negative affect instances were extracted for each subject.

Analysis plan

To address each research question, different multi-level mixed models were fit using the lmer function in the lme4 package (Kuznetsova et al., 2017) in R (R Core Team, 2018). All variables were centered. In all models, WTC values were the dependent variable. All multilevel models were within-dyad-level analyses. We employed a two-step approach to build the lmer models. We had a priori top-down hypothesis that we aimed to test for each of our hypotheses. This guided the inclusion of specific fixed effects in the models. We utilized a chi-square test to identify the best-fitting model based on the goodness-of-fit statistics. This approach allowed us to select the model that provided the best overall fit to the data. Different models were built for each hypothesis. To address our first hypothesis, we first examined how WTC changed throughout the interaction by including the time window as a fixed factor. To address our second hypothesis, we added the feedback category as a fixed factor to examine its role on WTC and its possible interactions with time window. We then added child affect as a fixed factor, and examined its possible interactions with feedback, and time window. Random intercepts were included for each dyad to account for the within-dyad correlation. To address our third hypothesis, we tested whether the random slope for effect of time window improved the model, we then extracted the subject-specific coefficient for the time window based on prior analyses and used these as predictors of task performance as measured by the number of tangram puzzles completed by the dyad.

Results

Descriptive statistics

Parental feedback.

We calculated the number of parental positive, neutral, and negative feedback utterances during the cooperative tangram task. On average, parents provided 14.6 utterances (range: 2–32) that included positive feedback, 16.8 (range: 4–32) utterances that included neutral feedback, and only 2.9 (range: 0–10) utterances that included negative feedback. Not only was negative feedback rare, but some parents never (6 out of 40) provided negative feedback. Thus, we narrowed our focus down to two variables for feedback: positive and neutral feedback. We also calculated the number of probe and informative utterances parents provided. On average, parents provided 13.9 (range: 2–45) probe utterances and 25.1 (range: 8–43) informative utterances during the interaction.

Child affect.

We calculated the number of child positive and negative affect instances during the cooperative tangram puzzle task. On average, children expressed 4.7 (range: 0–17) instances of positive affect and 1.7 (range: 0–17) instances of negative affect. Importantly, not only was negative affect rare, but 23 of the 40 children never expressed negative affect. Once again, this allowed us to restrict our focus to the positive affect of the child.

Interpersonal Neural Synchrony (INS)

First, we examined how the dyadic INS, as measured by WTC, changed during the interaction period. A linear-mixed model analysis revealed that WTC changed significantly during the interaction. We included both a linear and a quadratic term for time window as fixed factors. The change in the WTC was best represented by a quadratic model since it most closely mirrored the plot of the empirical data (see Figure 5). In addition, removing the quadratic term significantly reduced the model fit, χ²(1) = 889.4, p < .001. In this Model 1, the estimated fixed effects indicated a significant negative association of WTC to time window and a significant positive association of WTC to time window squared (see Table 1). The significant, non-zero coefficient for time window suggests a linear relationship between time and WTC (our measure of INS). The significant coefficient for the time window squared suggests a quadratic relationship. Crucially, WTC decreased during the tangram puzzle interaction, with the significant quadratic effect suggesting that the rate of WTC showed a significant deceleration across time windows. These findings underscore the dynamic nature of WTC during brief interactions.

Figure 5.

Variation in WTC by time window with smoothed Loess curve (95% confidence).

Table 1.

Models examining the relations of time and time window squared (Model 1), parental feedback added (Model 2) and child affect added (Model 3) on WTC in a linear-mixed effects model analysis.

		Model 1			Model 2			Model 3
	Estimate (Std Error)	t-value	p-value	Estimate (Std Error)	t-value	p-value	Estimate (Std Error)	t-value	p-value
(Intercept)	−0.260 (0.035)	−7.527	<0.001***	−0.258 (0.035)	−7.380	<0.001***	−0.263 (0.034)	−7.654	<0.001***
Time window	−0.006 (0.001)	−5.071	<0.001***	−0.006 (0.001)	−5.130	<0.001***	−0.006 (0.001)	−5.255	<0.001***
Time window sq	0.005 (0.001)	30.259	<0.001***	0.005 (0.001)	29.840	0.012*	0.005 (0.001)	29.916	<0.001***
Parent neutral feedback				0.023 (0.009)	2.580	<0.001***	0.019 (0.009)	2.126	0.034*
Parental positive feedback				0.035 (0.014)	2.514	0.012*	0.021 (0.014)	1.470	0.142
Time window * Parental positive feedback				−0.002 (0.001)	−1.660	0.097~	−0.002 (0.001)	−1.712	0.087~
Time window sq * Parental positive feedback				−0.001 (0.001)	−4.550	<0.001***	−0.001 (0.001)	−3.856	<0.001***
Child positive affect							0.174 (0.026)	6.565	<0.001***
Time window * Child positive affect							−0.001 (0.003)	−0.442	0.659
Time window sq * Child positive affect							−0.001 (0.001)	−2.966	0.003*
Parental positive feedback * Child positive affect							0.045 (0.018)	2.46	0.014

$$\text{WTC}\sim \text{timewindow}+\text{timewindowsq}+(1\mid \text{ID})$$ Model 1:

To ensure the validity of the WTC results, we conducted a comparison between INS values derived from family parent-child pairs and those calculated from shuffled random pairings of parents and children (Reindl et al., 2022). In other words, we randomly redistributed the data across different families and recalculated the WTC using parents and children from unrelated families. Figure 6 represents general WTC values for regular pairs versus shuffled random pairs. A linear mixed effect model was employed to compare the WTC values obtained from shuffled random pairs with those from family pairs. Our analysis revealed a significant difference, indicating that WTC was significantly greater in the case of family pairs compared to shuffled random pairs, β= 0.049, SE = 0.004, t= −11.2, p < .001. This comparison provides additional evidence for the validity and specificity of the INS measurements obtained from parent-child pairs from the same family.

Figure 6.

WTC scores in regular pairs and random pairs. X axis represents the 16 channels from the parents, and Y axis represents the 16 channels from the children. The number represents the channel ID for both X and Y axis. Figure a and b share the same color bar on the right of the figure b. a. WTC score for all regular pairs (range = 0.345 – 0.399, mean = 0.367, std = 0.010); b. WTC score for all shuffled random pairs (range = 0.296 – 0.351, mean = 0.321, std = 0.009).

Role of parental praise in INS

To address our second hypothesis, we examined how parental feedback and child affect predicted WTC. First, we examined how positive feedback and neutral feedback from the predicted levels of WTC, as well as the slope and acceleration of the WTC change. To do so, we added positive and neutral feedback as additional predictors building upon the baseline Model 1 describing change presented above. The inclusion of these variables significantly improved the model fit, χ²(2) = 9.071, p = .011. To assess whether positive and neutral feedback also predicted the slope and acceleration of the WTC change, we introduced interaction terms between each feedback type and the two time window variables. Including the interaction of positive feedback with time window and time window squared both resulted in a significant improvement in model fit compared to the original model with main effects only, with a chi-square test statistic of χ²( (2) = 22.812, p < .001. Including the interaction of neutral feedback with the time window and the time window squared to the original model with only main effects did not improve the model fit, χ²( (2) = 5.278, p = .07. Thus, these two interaction terms with neutral feedback were removed from the final model. In the final model (Model 2), shown in Table 1, positive and neutral feedback were positive predictors, the interaction of positive feedback with the time window was marginally significantly related to WTC and the interaction with the time window squared was significantly related to WTC. This indicates that higher levels of positive feedback were associated with a more attenuated slope and an enhanced acceleration of the WTC change. In simpler terms, when positive feedback was present, the decline in WTC during the interaction was less pronounced.

$$\text{WTC}\sim \text{timewindow}*\text{positive feedback}+\text{timewindowsq}*\text{positive feedback}+\text{neutral feedback}+(1\mid \text{ID})$$ Model 2:

Joint role of parental praise and child affect in INS

To continue examining our second hypotheses, we added child positive affect as an additional predictor, and its interaction with the rate of change variables. In this model, adding the child affect and its interaction with time window and time window squared significantly improved the model, χ²(3) = 70.479, p < .001. Finally, we added an interaction between child affect and parental positive feedback. This interaction term significantly improved the model as well, χ²(1) = 6.048, p = 0.014. In this final model, there was a significant main effect of neutral feedback and child affect. The role of positive feedback and child affect varied as a function of time variables (specifically the acceleration of WTC). Most importantly, in the final model (Model 3), there was a significant interaction between positive feedback and child positive affect on WTC. As seen in Figure 7, in the presence of positive child affect, parent positive feedback was associated with higher WTC throughout the interaction compared to when child positive affect was low. This final Model 3 is presented in Table 1. The conditional r-square for this final model was 0.10, indicating that 10% of the variance in WTC could be attributed to the fixed effects of the model.

Figure 7.

Variation in WTC by time window as a function of parental positive feedback and child positive affect. a. Raw data with smoothed Loess curve (95% confidence interval), b. Model predicted data. Groups determined based on median split.

$$\begin{gathered} \text{WTC}\sim \text{timewindow}*\text{positive feedback}+\text{timewindowsq}*\text{positive feedback}+ \\ \text{neutral feedback}+\text{timewindow}*\text{child positive affect}+\text{timewindowsq}*\text{child positive affect}+ \\ \text{positive feedback}*\text{child positive affect}+\left(1\right|\text{ID}) \end{gathered}$$ Model 3:

Follow-up analyses

We conducted follow-up analyses to ensure the precision of the results presented above. To ensure that synchrony patterns were specific to feedback and not to task-relevant information provided by the parent, we ran the model described above (Model 3) while controlling for parental informative utterances. In this model, parental informative utterances did not improve the model, χ²(1) = 0.515, p = 0.473, and did not emerge as a significant predictor of WTC, β = −0.001, SE = 0.001, t = −0.717, p = 0.474. Second, to ensure that synchrony patterns were not due to differences in children’s spatial ability, we ran the analyses above controlling for children’s performance on WPPSI Block Design task. In this model, WPPSI Block Design did not significantly improve the model fit, χ²(1) = 1.565, p = 0.211, and it did not emerge as a significant predictor of WTC, β = −0.001, SE = 0.001, t = −1.231, p = 0.226. Most importantly for our purposes, the pattern of results remained the same.

Role of INS in predicting task performance

To examine the relations between WTC and children’s spatial performance, we first tested whether random slope for linear effect of time improved the baseline model (WTC ~ timewindow + timewindowsq + (timewindow|ID)). This random slope indeed improved the model, χ²(2) = 389.55, p < 0.001. Having confirmed meaningful variability in slope, we extracted coefficients for the intercept and time window (slope) obtained from the baseline lmer model describing the change in WTC. Subsequently, these were included in the linear regression model to examine their association with two measures of spatial performance – a measure of dyad task performance variable assessed by the number of tangram puzzles completed and a standardized test of child’s spatial ability. Here neither the intercept of INS, F(1, 37) = 0.338, p = 0.564, nor the slope of WTC, F(1,37) = 0.978, p = 0.329, during the tangram task emerged as a significant predictor of the number of tangram puzzles completed. Although the WTC intercept did not emerge as a significant predictor of the WPPSI Block Design performance, F(1, 37) = 0.904, p = 0.348, WTC slope emerged as a significant negative predictor, F(1, 37) = 4.382, p = 0.042. The lower the slope of WTC change was (i.e., the more attenuated the change was in WTC), the higher children’s WPPSI Block Design performance was. The association of WTC slope to WPPSI performance remained robust, F(1, 35) = 4.202, p = 0.048, even after controlling for parental positive feedback and child positive affect in the same model.

Spatial distribution of INS

To better understand spatial distribution of WTC, we examined which ROI pairs between parents and children presented higher interpersonal neural synchrony. To examine this question, we reviewed our baseline model and included time window and time window squared as fixed factors. We added ROI pair (with 16 levels) as an additional fixed effect to this model. Adding the ROI pair variable significantly improved the model, χ²(15) = 44.27, p < 0.001. Bonferroni corrected post-hoc comparisons showed that INS between child left TPJ and parent right dlPFC, and child left TPJ and parent right TPJ was significantly higher than the average of all other ROI pairs (all p’s < 0.05). The channel pairs exhibiting the highest synchrony primarily comprised the left parietal ROIs of the child and the right prefrontal and parietal ROIs of the parent (see Figure 8).

Figure 8.

Illustration of brain areas which revealed highest neural synchrony. a. Left hemisphere from the children. Two ROIs shown in the left hemisphere are the left dorsal-lateral prefrontal cortex and left temporo-parietal junction. b. Right hemisphere from the parents. Two ROIs shown on the right hemisphere are right dorsal-lateral prefrontal cortex and right temporo-parietal junction. The orange lines across figure a and figure b represent the most synchronized brain areas during interaction, which are left TPJ from the children and right TPJ and right dlPFC from the parents.

Discussion

Parental praise, in the form of positive feedback, is considered foundational for strong parent-child relations and children’s developmental outcomes. However, recent empirical work showed that praise can also reduce children’s motivation and subsequent achievement, referred to as praise paradox. In this study, we used fNIRS hyperscanning to identify the dynamically changing neural mechanisms underlying interactions between parental praise and child affect during a cognitively challenging task. fNIRS hyperscanning is valuable in that it provides a unique and non-invasive method to study the neural dynamics and synchrony between parents and children during interactions as the interactions unfold. It allows for a more holistic understanding of the underlying real-time neural mechanisms that contribute to effective parent-child interactions, specifically in the context of parental feedback and child affect. Below we discuss each of our findings.

First, we contributed to the emerging literature on interpersonal neural synchrony (INS) as a measure of parent-child interactions (Nguyen et al., 2020, 2021a, 2021b). INS reflects a putative neural marker for social connection and coordination, facilitating rapport and communication between interacting partners. Specifically, it is considered to reflect mutual prediction processes between communication partners, which is based on both ostensive cues (such as verbal and nonverbal communicative signals) as well as higher-order cognitive processes/socio-emotional interpretations between the partners (Hamilton, 2021; Hoehl et al., 2021; Wass et al., 2020). The role of INS might also vary depending on the task at hand - for example, in the context of parent-child interactions neural synchrony might help co-regulate the child’s emotion or improve the cooperative communication. Here we aimed to pinpoint factors that dynamically contribute to parent-child INS during a problem-solving task. Prior studies of INS mainly assessed it over a task for the interacting partners, creating an average. Contributing to the more recent work that examined INS more dynamically (Nguyen et al., 2021b), we modeled the dynamic changes in INS using a similar approach to many studies assessing physiological synchrony (e.g., Woody et al., 2016).

We observed decreases in INS as the task progressed, combined with a slowing rate of decline. This finding suggested that the level of INS between parents and children evolves throughout the interaction. This dynamic pattern emphasizes the need to consider temporal dynamics in the study of parent-child neural synchrony and the complex nature of real-time interpersonal neural processes. Examining the change in interpersonal neural synchrony in addition to levels of synchrony provided unique information that would be lost in an approach that averages synchrony throughout the entire interaction. Indeed, several emerging studies support our emphasis on dynamic changes. In a similar problem-solving task with adult groups, Mayseless and colleagues (2019) also observed decreases in INS. A decrease in INS may be indicative of complimentary behavior types that emerge in complex problem solving and has been linked to creation of novel ideas in adults (Wallot et al., 2016). Consistent with this interpretation, we found that that the change in INS throughout the interaction, not the general level of INS, predicted children’s performance on an independent spatial task. These findings support the argument that INS can be considered indicative of alignment between parents and children during the cognitive task. The INS observed may reflect a more coordinated interaction between the parent and child and associate with more optimal outcomes.

Some prior work indicated that higher levels of INS between parents and children are associated with better performance outcomes in cooperative tasks, but others did not find this association (Nguyen et al., 2020, Nguyen et al., 2021c; Reindl et al., 2018). Combined with current findings, this suggests that not only the degree of INS during interactions, but also its dynamic change might also serve as a potential predictor. However, the direction of the change in synchrony might be task specific. One study by Nguyen and colleagues (2021b) leveraged a free-conversation task between parents and children devoid of a specific goal. They observed an increase in synchrony and argued that INS could affect parent-child conversation quality. This highlights the need to describe the dynamic changes in INS in a variety of tasks.

In addition, we further authenticated the validity and specificity of the INS measure by comparing family pairs to shuffled random pairs. Following more recent studies and guides in the literature for parent-child hyperscanning studies (Nguyen et al., 2021a), we did not compare synchrony during cooperation to a control condition such as a resting period or an individual condition. This was because resting phases are not considered ideal control conditions for young children and synchrony differences observed can also be due changes in the autonomic nervous system (ANS) (Tachtsidis and Scholkmann, 2016). Random pairs analyses are considered to better account for changes in ANS and other confounds, such as the visual set-up. Random pairs (i.e., shuffled data comparison) maintains the mutual experience of completing the tangram task across two sets of pairs in the face of other dynamically changing biological measures such as heart rate or breathing (Nguyen et al. 2021, 2021a; Reindl et al., 2018). At the same time, we did observe a similar shape of INS for the random pairs, which suggests that the patterns observed could also be at least in part driven by task characteristics as well. Overall, both comparisons provide valuable information. Comparison to rest might better remove the spontaneous brain activity from cooperation-induced activity and address whether activation is different for pairs versus individuals. Comparison to shuffled data might better address whether the observed synchrony during cooperation is higher than chance level and whether observed synchrony is specific to parent-child pair than any general pairs. This possibility can be addressed in future experimental studies systematically manipulating task characteristics and nature of the interaction between parents and children. Regardless, although shape of the change in INS was similar for real family versus shuffled pairs, the fact that we observed significantly greater INS within family pairs confirms that the observed interpersonal neural synchrony indeed reflects the nature of the specific parent-child relationships in the same family (Reindl et al., 2022).

In terms of spatial distribution of INS, we showed that the highest interpersonal neural synchrony was between the left TPJ of the child and right dlPFC and TPJ of the parent. TPJ and dlPFC are considered an important component of neural mechanisms supporting social interactions. The TPJ is involved in mentalizing, tracking socio-emotional value of the incoming input, and related to social connectedness. The dlPFC is involved in top-down cognitive control during cooperation and to track task-relevant information. Some argue that increased INS in these regions reflect increased attunement and allocation of attention to the interaction (Gvirts & Perlmutter, 2020). Prior hyperscanning research with adults showed that INS increased in left TPJ within follower-leader pairs (Jiang et al., 2015). Given the limitations of fNIRS and our limited probe set-up (e.g., Koike et al., 2015), our interpretations about the specificity of the spatial distribution in our results should be interpreted with caution. Overall, our results tentatively suggest that parents in the current study were both leading the task and maintaining the interaction while tracking mental states of their children. Conversely, children mainly took a follower role.

Our focus was to build on the existing literature and extend the use of INS to study dynamic changes in parent-child interactions during a cognitive task. We explored how parental feedback relates to changes INS and how the role of feedback might vary as a function of children’s responses. In terms of parental feedback, the results showed that positive and neutral feedback were most prevalent during a challenging cognitive task. In contrast, negative feedback was relatively rare. These lower levels of negative feedback may stem from the younger age range of our participants, as parental criticism increases across elementary and middle school years as parents gradually place heavier emphasis on academic performance (Gunderson et al., 2018).

Our main question was regarding how parental verbal feedback and child affective responses related to changing interpersonal neural synchrony patterns during parent-child interactions. Our results showed that parental praise and child positive affect both influenced the synchrony between the parent and the child. It is important to highlight that our parental measure was verbal whereas the child measure was primarily a nonverbal measure, in keeping with our goal of analyzing different components in the parent and the child. While the synchrony between the dyad showed attenuation throughout the interaction, praise, and child positive affect both predicted a lower decline of synchrony. In other words, when parents provided praise and children displayed positive affect, the decline in synchrony was less pronounced and a higher level of synchrony was maintained. Overall, our results are consistent with the hypothesis that decreasing neural synchrony might indicate emerging cooperative problem solving (e.g. Mayseless et al., 2019), and suggest that maintaining a certain level of synchrony could create an optimal interaction. The attenuation of the decline in synchrony suggests that praise and positive affect contribute to maintaining a higher level of alignment throughout the interaction.

Importantly for our purposes, our two main factors of interest interacted. When praise and child positive affect were both present, interpersonal neural synchrony was higher than all other conditions – especially when compared to conditions where feedback was present, but child positive affect was not. These findings corroborate behavioral studies suggesting that feedback is not always associated with better interactional outcomes (e.g., Henderlong & Lepper, 2002; Mueller & Dweck, 1998). Here, we, for the first time, identified a neural basis for cases where praise might backfire, documenting the role of positive praise might vary depending on child affect. We also documented changing associations between parental feedback and child affect in terms of their relationship with interpersonal neural synchrony. Importantly, these associations were observed even after controlling task-specific information provided by parents, as well as baseline variation in children’s independent spatial skill. The specificity of these associations highlights the importance of considering not only task-relevant feedback from parents, but also the broader support they afford to children during learning interactions. Overall, observing the ebb and flow of interpersonal neural synchrony across successive time windows via fNIRS hyperscanning opens the way for future studies examining dynamic and bidirectional relations between parent feedback and child affect.

Our study had certain limitations and raises questions for future studies. While we leveraged prior literature, patterns in our INS and behavioral data to determine our window length and frequency selection, future hyperscanning studies should strive to establish more standardized guidelines for these values (Marzoratti & Evans, 2022; Nguyen et al., 2021a). For example, while we used a frequency band like prior studies on parent-child interactions leveraging similar stimuli (Nguyen et al., 2020; Zhao et al., 2022), future studies should discuss the role of different frequency bands in further detail as frequency bands might vary depending on the context (Hakim et al., 2023). We captured naturalistically varying parental feedback. This limited our ability to make causal claims and examine the effect of different feedback categories. For example, negative feedback was rare. We also did not have sufficient variability in our data to consider different types of positive praise. According to the attribution/social-cognitive theory of motivation (Hong et al., 1999), the efficacy of praise may hinge upon the specific way it is phrased. For instance, praise focusing on the children’s generic ability (e.g., “You are so good at this!”) or emphasizing their effort (e.g., “You can do it!”) has been found to influence children’s motivation (Cimpian, 2010; Zentall & Morris, 2010). Similarly, Brummelman and collagues (2016) observed that the utilization of praise directed toward persons with low self-esteem (e.g., “You’re smart”) and inflated praise (e.g., “That’s incredibly beautiful!”) could diminish children’s motivation and feelings of self-worth when confronted with challenges or setbacks such as instances of struggle or failure. In the current study, person praise was rare possibly due to the nature of the task presented (only 2 of the 40 parents produced a person praise), which limited our ability to examine the role of praise type. Future work should more closely examine specific relations between different types of positive praise and neural synchrony. One potential direction for future research would be to experimentally manipulate parental feedback. We targeted the brain regions most relevant to dyadic interactions but did not have whole brain coverage based on prior literature. Thus, we were limited in our ability to make conclusions about the involvement of specific brain areas outside the ones that we focused on in interpersonal neural synchrony. We included both mothers and fathers as our focus was on primary caregivers, but prior work showed that the neural basis of mother-child interactions might differ from that of father-child interactions (Nguyen et al., 2021c) – see supplementary materials for follow-up analyses on mothers only. Finally, our sample consisted of families from relatively high socioeconomic backgrounds, which might limit the generalizability of our findings.

Loving relationships with caregivers are fundamental for children’s development both broadly and more specifically during learning interactions. Praise is considered one crucial component of multi-faceted loving parent-child relationships in various theoretical models, including behavioral, attachment and cognitive perspectives. However, empirical work suggests that the role of praise might vary depending on a host of factors. The current study examined the role of parental praise in a challenging cognitive task and asked whether the role of praise varied as a function of children’s affective states. Leveraging fNIRS-based hyperscanning, we, for the first time, showed that the neural basis of parent-child interactions dynamically changes as function of interactive relationships between parental feedback and child affect. We showed that neural synchrony is highest and maintained at a higher level during the interaction when parental praise coincides with child’s positive affect. This profile of synchrony is correlated with more optimal child performance. Our work adds to the prior literature on parental praise by pinpointing factors that moderate the role of praise in the context of dynamic parent-child interactions. In doing so, our research contributes to the broader field by utilizing fNIRS hyperscanning to deepen our understanding of the mechanisms shaping the dynamics of parent-child relations as they evolve in real time.

Research Highlights.

1. The level of interpersonal neural synchrony between parents and children dynamically varies during a cognitively challenging (tangram) task.

2. The left temporo-parietal regions of the child and the right dorsolateral prefrontal and right temporo-parietal regions of the parent demonstrate the strongest parent-child neural synchrony.

3. The relationship between parental praise (positive feedback) and parent-child neural synchrony is contingent upon child positive affect during the task.

4. Change in parent-child neural synchrony relates to children’s performance on an independent visuospatial processing measure.

Data Availability Statement:

The data that support the findings of this study are available from the corresponding author upon reasonable request.