ABSTRACT
In infants, as in adults, social context is known to influence attentional allocation during communication. The sharing of attention between individuals potentiates learning, but little is understood about the interpersonal neural mechanisms that support this process. Recently, it has been demonstrated that during spoken communication, spontaneous neural coupling (temporal synchronization) arises between speaker and listener, and their coupling strength predicts communicative success. Here, we assess whether gaze, a salient cue that elicits joint attention, moderates endogenous levels of neural coupling in adult-infant speaker-listener dyads. Electroencephalography (EEG) was concurrently measured in 19 adult experimenter-infant dyads at left and right central electrode locations. The adult sang nursery rhymes to the infant whilst either looking directly at the infant, or with her gaze averted by 20°. Gaze-related changes in adult-infant neural network connectivity were measured using Partial Directed Coherence (PDC), a statistical measure of causality and directional influence. Our results showed that bi-directional connectivity between adults and infants was significantly higher during periods of Direct than Indirect gaze in Theta, Alpha and Beta EEG bands. Further analyses suggested that these effects were not attributable to differences in task engagement, EEG power, or basic neural processing of speech between gaze conditions. Further, in Alpha and Beta bands, but not other bands, infants influenced adults more strongly than vice versa. This is the first demonstration that mutual direct gaze increases adult-infant neural coupling during social communication. Future research should explore the role of neural coupling in learning and other aspects of social behavior.
SIGNIFICANCE STATEMENT During infancy, the social context exerts powerful influences on learning. This context arises from dynamic interactions between social partners, yet all known neural infancy studies have only considered what occurs within one partner, the infant. Consequently, the contingency (temporal dependency) of infant’s neural activity on the adult’s and vice versa has never been measured. Yet, recent adult studies suggest that strong interpersonal neural contingency (coupling) predicts successful communication. Here, we report the first ever study to examine adult-infant neural coupling and characterize its causal architecture. We observed strong bidirectional adult-infant coupling which was significantly modulated by social gaze. These results are important because they challenge the current thinking about how social effects on early learning are understood and investigated.
1 INTRODUCTION
Social interactions between adults and infants involve both verbal and non-verbal modes of communication (Csibra & Gergely, 2009). These are known to play a vital role in supporting early learning across multiple domains of language, cognition and socio-emotional development (Rogoff, 1990; Leslie, 1994; Csibra & Gergely, 1998). The neural mechanisms that support these social interactions are now beginning to be understood. Speech perception, for example, is known to involve the synchronization (or phase-locking) of neural activity to temporal structures in speech (Giraud & Poeppel, 2012). Infants, like adults, show neural oscillatory phase-locking to speech (Leong et al., in revision; see also Telkemeyer et al, 2009).
Infants are also known to rely heavily on non-verbal cues (such as eye contact, gaze direction, pointing and gestures) to infer meaning, intention and causality (Csibra & Gergely, 2000). Neonates are already sensitive to adults’ non-verbal social communicative cues (Meltzoff & Moore, 1977; Farroni et al, 2002) but the ability to engage in non-verbal communication develops rapidly over the first year of life (Mundy et al, 2000, 2003). A highly salient ostensive signal in human communication is direct gaze towards the addressee, which usually results in mutual eye contact (Csibra & Gergely, 2009). From birth, infants prefer to look at pictures of faces with direct gaze over averted gaze (Farroni et al., 2002). By 4 months, ERP studies show enhanced neural processing of pictures of faces showing direct relative to indirect gaze (Farroni et al., 2002). In adults, direct gaze leads to activation in similar neural areas as those evoked by other communicative signals (e.g. direct gaze vs eyebrow raise), which are both interpreted as ostensive signals (Kampe et al., 2003).
Direct gaze potentiates joint attention (a state of shared focus between individuals), which in turn enhances infants’ neural processing of jointly-attended stimuli. For example, infants are more likely to follow an adult’s gaze towards another object when it is preceded by a moment of direct, mutual gaze (Senju & Csibra, 2008). When infants attend jointly to an object with an adult, they show a larger Nc ERP (Striano et al., 2006; Parise et al., 2010), along with changes in neural activity in Theta- and Alpha- EEG bands (Hoehl et al., 2014; St John et al., 2016). However, joint attention is a dyadic activity, and previous investigations into the neural correlates of joint attention have, almost exclusively, only explored the neural correlates of joint attention within the child alone (although see Lachat et al, 2012). Consequently, the contingency (temporal dependency) of one partner’s neural activity with respect to the other during joint attention has not previously been investigated.
Adult-infant temporal contingencies have long been observed in behavioural and physiological domains. For example, patterns of temporally synchronous activity between parent and child during shared attention have been noted both for gaze (Kaye & Fogel, 1980) and affect (Cohn & Tronick, 1988; Feldman et al., 2006). These episodes of interaction also lead to physiological synchronization of heart rhythms (Feldman et al., 2011). Interpersonal neural dynamics, however, are only just beginning to be studied using neuroimaging methods that allow the simultaneous capture of brain activity between two individuals, such as dual electroencephalography (dual-EEG; e.g. Dumas et al, 2010, 2012; Lachat et al, 2012), dual functional magnetic resonance imaging (dual-fMRI; e.g. Stephens et al, 2010; Anders et al, 2011) and dual functional near infrared spectroscopy (dual-fNIRS; e.g. Jiang et al., 2012), see Hasson et al (2012), Hari et al (2013) and Koike et al (2015) for recent reviews.
In adults, neural synchronisation between individuals has been shown in frontal areas during face-to-face communication (Jiang et al, 2012). Further, the strength of speaker-listener neural synchronisation has been associated with listeners’ success in comprehending the speaker (Stephens et al, 2010; Dikker et al, 2014; Silbert et al, 2014). Neural synchronization between adults has also been observed in the absence of speech. For example, Saito et al (2010) found that adult pairs who performed a gaze cueing task in real time together, maintaining eye contact and gaze following, had stronger mutual neural synchronization in their right inferior frontal gyri (IFG) than control data from non-matched pairs.
Here, we aimed to measure the temporal dependency (synchronization) between adult and infant neural signals during conditions of direct or indirect adult speaker gaze. The measure we used was Partial directed coherence (PDC), which not only captures the strength of instantaneous interactions between pairs of neural signals, but also provides insights into their causal relationship and direction of influence (Baccala & Sameshima, 2001; Baccala et al, 2007; Faes & Nollo, 2010, 2011). PDC is based on the concept of Granger causality (Granger, 1969) which reflects the extent to which one signal statistically predicts another signal in time. Further, PDC reveals only direct couplings between channels (e.g. i -> j), and does not measure the effect of intermediary channels (e.g. i -> k -> j). Thus, PDC is a suitable measure of changes in information flow (both directionality and strength) during different network states. Here, we apply the PDC metric to measure the causal architecture and information flows in a dyadic social network comprising an infant and an adult. We manipulate social context so that the adult is either looking directly at the infant, or indirectly, at a 20° oblique angle.
In terms of affect and physiological changes, research has shown that the influence of infants and parents on one another is bi-directional (Feldman et al., 2006, 2011). In older infants, however, such as the age range studied here, research suggests that mothers are more likely to be responsive to their children than vice versa (Cohn & Tronick, 1986.)
Accordingly, we predicted that: i) significant neural coupling would exist between adults and infants during social interaction, ii) that direct gaze would be associated with higher interpersonal neural connectivity than indirect gaze and iii) that infants would influence adults more than vice versa.
2 METHODS
2.1 Participants
The participants were twenty-nine infants (15M/14F) and one female adult experimenter. The infants’ mothers were all native English speakers. The median age was 8.3 months (st. err. 0.44 months). All infants had no neurological problems and had normal hearing and vision, as assessed by maternal report.
2.2 Materials
Seven familiar nursery rhymes such as ‘Old MacDonald’ were used as sung stimuli, as listed in Table S1 (Supplementary materials). Prior to starting the experiment, mothers confirmed that the nursery rhymes were familiar to their infant. As the stimuli were produced live the experimenter was recorded during each experimental session to ensure that the durations were consistent between the Direct and Indirect conditions (see Table S1). Bonferroni-corrected t-tests did not identify any significant differences between conditions.
2.3 Protocol
Infants sat upright in a high chair facing the female experimenter, who was the same for all testing sessions. Each nursery rhyme was presented in two gaze conditions. In the Direct condition the experimenter looked directly at the infant while singing; in the Indirect condition she averted her gaze by fixating at a constant visual target exactly 20° to the left or right side of the infant (see Figure 1). The experiment was divided into two blocks, with a short break in-between. In each block, the experimenter sang each nursery rhyme to the infant twice (once Direct and once Indirect, order counterbalanced). Thus, infants heard each nursery rhyme 4 times in total. The total stimulus presentation time (7 nursery rhymes x 4 repetitions) was 377.65 seconds on average (range = 350.73s to 391.90s, SD = 12.42s).
2.4 EEG acquisition
EEG was recorded simultaneously from the infant and the female adult experimenter from two electrodes in the central region (C3 and C4), referenced to the vertex (Cz) according to the International 10–20 placement system. EEG was recorded from central sites to reduce potential confounding influences of muscle artefacts and blinking while still capturing a robust neural response. The vertex reference location was used because it produces comparable results to other reference sites (Tomarken, Davidson, Wheeler, & Kinney, 1992), and is the least invasive for young infants. Prior to electrode attachment, electrode sites were marked and wiped with alcohol. Electrodes were then affixed to the scalp using Signa conductive electrode gel (Parker Laboratories Inc, NJ). EEG signals were obtained using a Biopac MP150 Acquisition System with filters at 0.1 Hz highpass and 100 Hz lowpass. Wireless dual-channel BioNomadix amplifiers were used to reduce distraction for the infant during testing. EEG was recorded at 1000 Hz using AcqKnowledge software (Biopac Systems Inc). All further analysis was performed using Matlab software (Mathworks Inc). Both participants’ data was recorded concurrently in a single acquisition session on the same computer, ensuring accurate time synchronization of the two data streams.
2.5 EEG artifact rejection and pre-processing
To ensure that the EEG data used for analysis reflected only awake, attentive and movement-free behaviour we performed a two-stage artifact rejection procedure. First, each dyad was video-taped and the videos were reviewed frame-by-frame (30 fps) to identify the onset and offset times of movement artifacts, including blinks, head and limb motion, and chewing. Only periods when infants were still and looking directly at the experimenter were accepted. Next, manual artifact rejection was performed on this still, attentive data to further exclude segments where the amplitude of infants’ or adults’ EEG exceeded +100 μV.
Following this two-stage process, 19/29 infants (10M/9F), gave sufficient data for inclusion in the final analyses. The median (st.err.) age of retained infants was 8.52 (0.57) months. On average, the retained infants contributed 45.52 seconds (range = 8.00s to 107.00s, SD = 28.18s) of still and attentive data in the Direct gaze condition, and 43.92 seconds (range = 11.00 to 122.62s, SD = 30.07s) in the Indirect gaze condition. Adult data was only analysed for those segments in which the infant data were retained. A paired t-test confirmed that there was no significant difference in the amount of still and attentive data collected between Direct and Indirect gaze conditions (t(18) = 0.44, p = .66). Therefore, infants were not more inattentive during the Indirect gaze condition.
The cleaned EEG data was resampled to a lower frequency of 200 Hz, which permitted the use of a low order multivariate autoregressive (MVAR) model in the subsequent connectivity analysis, whilst retaining sufficient spectral detail. The data was then low-pass filtered under 45 Hz to suppress electrical line noise. Finally, the data was segmented into 1.0s-long epochs (200 data samples per epoch) for connectivity analysis.
2.6 EEG analyses : Power spectrum and GPDC network connectivity
A detailed description of all methods is given in the Supplementary Materials (Section 2). Briefly, we first assessed the EEG power spectrum of infant and adult signals for each experimental condition and hemisphere (left or right). Second, to assess network connectivity in each gaze condition, we measured Generalised Partial Directed Coherence (GPDC), which is a directional casual measure of direct information flows between channels in a network (Baccala & Sameshima, 2001; Baccala et al, 2007; Faes & Nollo, 2010, 2011). The GPDC measure is based on the principles of Granger Causality (Granger, 1969), and measures the degree of influence that channel j (‘Sender’) directly has on channel i (‘Receiver’) with respect to the total influence of j on all channels in the network. Here, each individual electrode (Infant L, Infant R, Adult L, Adult R) was taken as one channel (see Figure 2). We computed directed coherence values for all 12 possible (non-self) pairwise connections, both within individual (e.g. Infant L -> Infant R) as well as across individuals (e.g. Infant L -> Adult L).
2.7 Control analyses
In the first control analysis, we generated a surrogate dataset comprising all 3876 possible combinations of non-matched adult-infant data (i.e. combinations of adult and infant data from different experimental sessions). We then performed an identical connectivity analysis on this surrogate dataset (see Supplementary Materials Section 2.2). The resulting values provided a baseline measure of the ‘random’ level of coherence that did not specifically arise from the task (e.g. due to being present in a similar physical testing environment). In the second control analysis, we examined whether basic sensory processing of the speech stimulus differed between Direct and Indirect mutual gaze conditions. If basic auditory processing differences were present, then any observed neural connectivity changes would not be solely gaze-related. Accordingly, wavelet coherence between the neural EEG signal and the speech amplitude envelope was measured in each gaze condition for adults and infants. A description of this method is given in the Supplementary Materials (Section 2.3).
2.8 Statistical analysis of GPDC network changes
To assess whether the Gaze experimental manipulation resulted in statistically-significant changes in GPDC strength within the neural network, we conducted 2 x 2 x 2 x 2 x 2 Mixed ANOVAs taking Gaze condition ([2], Direct or Indirect), ‘Sender’ ([2], Infant or Adult), ‘Sending’ Hemisphere ([2], L or R) and ‘Receiving’ Hemisphere ([2], L or R) as within-subjects factors. This allowed us to determine whether there was an overall global effect of Gaze (as indexed by a significant main effect of Gaze), and also whether patterns of connections in the network displayed a consistent causal architecture. For example, if adults were influencing infants more strongly than infants were influencing adults overall, this would be indexed by a significant main effect of ‘Sender’. Table S2 (Supplementary Materials) summarises the predicted pattern of effects. In addition, we also examined whether there were differences in connectivity as a function of infant age by using a median split at 9 months, and entering Age Group as a between-subjects factor in the ANOVA analysis. Finally, infants’ mean looking time across all conditions was entered as a co-variate, to control for individual differences in infant attentiveness.
3 RESULTS
3.1 Analysis of EEG power spectra across conditions
Figure 3 shows the mean power spectrum of the EEG signal for infants and adults, for the experimental conditions of Direct (solid line) and Indirect (dotted line) gaze, for left and right hemispheres respectively. Paired t-tests conducted for each EEG frequency band (Delta [1-3 Hz], Theta [3-6 Hz], Alpha [6-9 Hz], Beta [9-25 Hz], and Gamma [25-40 Hz]) indicated that there was no significant difference in EEG power between the Direct and Indirect conditions in infant or adult signals for either hemisphere in any EEG band (p>.05 for all comparisons, Benjamini-Hochberg FDR-corrected [Benjamini & Hochberg, 1995, 2000]). Therefore, the gaze manipulation did not generate any detectable power changes that might systematically bias the PDC metric (see Supplementary Materials 2.1 for an explanation of the effect of power on PDC).
3.2 Comparison of real and non-matched control surrogate data
The mean GPDC values for each EEG band, for each pairwise connection, are shown in Figure 4 (full values are given in Tables S3 and S4 of the Supplementary Materials). We assessed whether the mean GPDC values for each non-self pairwise connection, and for each condition and each EEG frequency, were significantly above their respective threshold values established from the surrogate dataset with non-matched adult-infant pairs. One-sample t-tests revealed that all real GPDC values were significantly above their respective surrogate values for all frequencies and non-self pairwise connections (Benjamini-Hochberg FDR-corrected p<.05 for all pairwise connections [Benjamini & Hochberg, 1995, 2000]). Therefore, significant bi-directional patterns of connectivity occurred both across and within infant and adult signals in all EEG bands.
For illustration, Figure 5 depicts a network representation of Alpha band (6-9 Hz) data, where nodes are channels (electrodes) and edges are pairwise GPDC connections. Network graphs for Direct and Indirect gaze conditions are shown in separate subplots, along with the difference between conditions (a positive value indicates higher connectivity in the Direct condition). Inspection of Figure 5 indicates that stronger Alpha connectivity was observed in the Direct gaze condition as compared to the Indirect gaze condition in the majority of across-individual connections. Further, infant→adult GPDC values (shown in red) were generally higher than adult→infant values (shown in blue) for the same pairwise connection (e.g. in the Direct condition, Ad L→Inf L = 0.42 whereas Inf L→Ad L = 0.75).
3.3 Network connectivity patterns and modulation by gaze
As outlined previously, Mixed Design ANOVAs were used to assess specific features of network connectivity (such as ‘Sender’-‘Receiver’ patterns and hemispheric lateralization patterns), and the modulation of these connectivity patterns by Gaze. To assess across-individual (interpersonal) neural connectivity patterns, five separate ANOVA analyses were conducted for Delta, Theta, Alpha, Beta and Gamma frequency bands. To assess within-individual connectivity patterns, a separate set of five ANOVAs were again conducted for Delta, Theta, Alpha, Beta and Gamma bands. The full results of these ANOVAs are reported in the Supplementary Materials (Table S5), and a summary of the major effects is presented here.
3.3.1 Gaze effects
Stronger low-frequency interpersonal connectivity for Direct gaze than Indirect gaze. Overall interpersonal connectivity was significantly stronger for Direct as compared to Indirect gaze in Theta, Alpha and Beta bands (Theta : F(1,16) = 5.40, p<.05, η2p =.25; Alpha: F(1,16) = 7.94, p<.05, η2p = .33; Beta: F(1,16) = 7.95, p<.05, η2p = .33) (see Figure 6). No significant interactions were observed between Gaze and Hemisphere or Age Group in any EEG band (see Table S5). This suggests that the overall increase in interpersonal neural connectivity for Direct relative to Indirect gaze was consistently observed across left and right hemispheres, and in both younger and older infants. In the Alpha and Beta bands, a marginally non-significant interaction between Gaze and ‘Sender’ was observed (Alpha: F(1,16) = 4.01, p=.06, η2p =.20; Beta (F(1,16) = 4.30, p=.05, η2p = .21). Tukey post hoc analyses of this interaction indicated that in both bands, infants were stronger ‘Senders’ in the Direct as compared to the Indirect gaze condition (p<.05 for both bands), but the difference between conditions for the adult was not significant. In the Theta band, however, the Gaze x Sender interaction was not significant (F(1,16) = .13, p=.73, η2p =.01) indicating that both adult and infant were sending more strongly in the Direct condition.
3.3.2 Direction of adult-infant influences
Infants influence adults more strongly at Alpha and Beta frequencies. In Alpha and Beta bands, infants were stronger ‘Senders’ than adults (Alpha : F(1,16) = 7.35, p<.05, η2p =.32; Beta: F(1,16) = 7.41, p<.05, η2p =.32). By contrast, in the Delta, Theta and Gamma bands, there was no significant main effect of ‘Sender’, indicating that infants and adults were influencing each other equally at these frequencies (Delta : F(1,16) = 0.02, p=.90, η2p =.00; Theta : F(1,16) = 0.56, p=.46, η2p =.03; Gamma : F(1,16) = 2.87, p=.11, η2p =.15).
In the Supplementary Materials Section 5.3 we also describe the results of the within-individual connectivity analyses. In brief, stronger within-individual connectivity was observed in infants than adults in four out of five frequency bands assessed (all except Beta). Further, for both infants and adults, stronger within-individual connectivity was observed for Direct gaze relative to the Indirect gaze in the Delta band, but not in other EEG bands. Finally, in the Supplementary Materials Section 6 we describe the results of the second control analysis. This was conducted to examine whether interpersonal connectivity gaze effects could be attributed to differences in basic speech processing (see Figure S1). In brief, no significant differences in speech-brain coherence were observed between the Direct and Indirect gaze conditions at any EEG frequency.
4 DISCUSSION
Recently, neural coupling (synchronization) between adults has been observed during social interaction. For example, previous studies using dual-fMRI have shown that when two adults engage in eye contact, neural activity in areas such as the right inferior frontal gyrus becomes synchronized (Saito et al, 2010; Tanabe et al, 2012). Here, we aimed to (1) determine whether neural coupling also exists between adults and infants during social interaction, (2) to characterize the causal architecture of this coupling, and (3) to assess whether neural coupling is modulated by the social gaze context (direct/indirect gaze).
Our results indicated, first, that significant neural coupling does indeed exist between infants and adults during social interaction. Tests showed that all pairwise coherence values were significantly above their respective surrogate values at all frequencies. That is, one partner’s neural activity was temporally dependent on the other partner’s neural activity across all timescales measured. Further, when investigating the causal architecture of the adult-infant neural network, we found that each channel (adult or infant; left or right hemisphere) had a significant and direct influence on every other channel, suggesting that causal patterns of influence were bi-directional between adult and infant, and also bi-hemispheric. Of note, significant bi-directional coupling was also observed in the Indirect condition. This was expected, since the infant was facing the adult directly in both conditions, and, for the adult, the infant was positioned at 20° eccentricity from the fixation point, and so still clearly visible.
Second, we found that in Theta, Alpha and Beta bands, stronger adult-infant connectivity was observed during Direct relative to Indirect gaze. This effect was not due to a reduction in attention from the infant listener during adult gaze aversion, as infants showed equivalent time durations of still and attentive looking at the adult across both conditions. This result also cannot be due to underlying power differences in the EEG spectra, as analyses indicated no change in power for infants or adults between the two gaze conditions. The lack of a power effect may initially appear surprising as previous infant studies have reported relative alpha desynchronization in conditions of joint attention (Hoehl et al, 2014; St John et al, 2016). However, it may be the case that alpha desynchronization is only observed in situations involving triadic attention between the infant, an adult and an object. Indeed, St John et al (2016) found that when the adult interacted face-to-face with the infant without looking or pointing to an object (similar to our current paradigm), no alpha desynchronization was observed. Our gaze effect also cannot be explained as a meta-phenomenon of changes in basic sensory processing of the speech signal, as both infants’ and adults’ accuracy of neural tracking of the speech signal remained at the same level across gaze conditions. Third, in Alpha and Beta bands only, we found that infants influenced adults more strongly than vice versa. This finding is consistent with behavioural studies which show that mothers are more likely to be responsive to their children than vice versa (Cohn & Tronick, 1986). Finally, we found no significant difference in the pattern of effects for younger (<9 months) and older infants, and we also found no overall hemispheric differences in connectivity patterns.
Previous dual-EEG studies on the temporal architecture of interpersonal neural coupling have generally focused on the coordination of motor activities such as finger-tapping tasks (Dumas et al, 2010; Konvalinka et al, 2014; Naeem et al, 2012). Our findings are novel insofar as they were observed in the absence of motor co-ordination within the dyad, but rather as an effect of modulation of social context (gaze). As such our results are most directly comparable to the dual-fNIRS findings from Jiang and colleagues, who observed greater neural synchronization in the left inferior frontal cortex during face-to-face dialog, relative to back-to-back dialog (Jiang et al., 2012). They are also comparable to results from Saito and colleagues who used dual-fMRI and identified, in adults, greater neural synchrony in the inferior frontal gyrus during eye contact. Here, we found that increased adult-infant neural coupling during direct gaze was observed only at low neural oscillatory frequencies such as Theta and Alpha, which are frequency bands that have also been implicated in previous EEG studies of joint attention with infants and adults (Lachat et al, 2012; Hoehl et al, 2014; St John et al, 2016).
One potential mechanism that might mediate interpersonal neural coupling is mutual phase-resetting in response to salient social signals. The phase of cortical oscillations (the neural feature with which GPDC values are computed) reflects the excitability of underlying neuronal populations to incoming sensory stimulation (Schroeder et al, 2009). Sensory information arriving during high receptivity periods is more likely to be encoded than information arriving during low receptivity periods. Consequently, neuronal oscillations have been proposed to be a mechanism for temporal and spatial sampling of the environment (Giraud & Poeppel, 2012; Kayser et al, 2012; Lisman, 2005) as well as for attentional selection (Schroeder & Lakatos, 2009). Specifically, salient events in the world are thought to reset the phase of on-going neuronal oscillations to match the temporal structure of these events and optimize their encoding (Schroeder & Lakatos, 2009). Consequently, interpersonal neural synchronization could increase within a dyad during the course of social interaction because each partner is continuously producing salient social signals (such as gaze, gestures, or speech) that act as synchronization triggers to reset the phase of their partner’s on-going oscillations, bringing each in close alignment with the other. Here, we replicate previous findings with adults that gaze (eye contact) is one social cue that moderates interpersonal neural synchronization.
There are however a number of limitations to this work. The first is that only two EEG channels were recorded from each individual (C3 and C4). While these locations were chosen due to their low contamination by speech articulatory artifacts (see Supplementary Materials Section 7), joint attention is thought to involve a distally distributed neural network of frontal temporal and parietal cortical and subcortical neural regions, which our recording techniques would not capture fully (Mundy, 2003). The advantage of this low-density approach was that it improved the interpretability of the PDC measure, as the number of pairwise connections increases exponentially with the number of recording channels. Other limitations include possible contamination of our data by the adult’s speech articulatory artifacts. As we discuss in detail in the Supplementary Materials (Section 7), these would have been consistent between Direct and Indirect conditions, and therefore a more likely cause of a Type II rather than a Type I statistical error.
In conclusion, the current study is (to our knowledge) the first demonstration that significant neural coupling can be measured between infants and adults, and that the strength of adult-infant neural coupling is higher during Direct than Indirect mutual gaze. Further research should aim to assess relationships with attention states, learning and behavior using a wider range of neural recording sites, in more diverse settings, to investigate causal mechanisms, as well as other questions, such as individual differences.
ACKNOWLEDGEMENTS
This research was funded by an ESRC Transforming Social Sciences collaboration grant (ES/N006461/1) to VL and SW, a Lucy Cavendish College Junior Research Fellowship to VL, and by a British Academy Post-Doctoral Fellowship and an ESRC FRL Fellowship (ES/N017560/1) to SW.