Running head: CHAOS, VOCALISATIONS, AROUSAL - 1 -
Needing to shout to be heard? Affective dysregulation, caregiver under-responsivity,
and disconnection between vocal signalling and autonomic arousal in infants from
Wass, S.V.(*)(1), Goupil, L.(1), Smith, C.S.(2), Greenwood, E.M.G.(1)
1 – Department of Psychology, University of East London
2 – Institute of Psychiatry, Psychology & Neuroscience, King’s College London, UK
Corresponding author: Dr Sam Wass ORCID ID 0000-0002-7421-3493. Address: University
of East London, London E15 4LZ. Telephone: +44(0)7725369189. Email:
This research was funded by ESRC grant number ES/N017560/1, by ERC grant number
ONACSA 853251, by Project Grant RPG-2018-281 from the Leverhulme Trust and by an
ERC Marie Curie Fellowship JDIL 845859. Thanks to Kaili Clackson and Farhan Mirza for
help with data collection; to Caitlin Gibbs, Emma Bruce-Gardyne, Florian Andrey-Csolm,
Joan Eitzenberger, Leanne Barnes, Louise Stubbs, Deborah Scnatlebury and Anne
Hepworth for help with data coding. Thanks to members of the UEL BabyDev Lab for
comments and discussions on earlier drafts of this manuscript, and to all participating
children and caregivers.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 2 -
Data sharing: Due to the personally identifiable data of the data contained in this manuscript
(microphone recordings of infants), access to the data is only available through direct request
to the first author.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 3 -
Higher levels of household chaos have been related to increased child affect dysregulation
during later development. To understand why this relationship emerges, we used miniature
wearable microphones and autonomic monitors to obtain day-long recordings in home
settings from a cohort of N=74 12-month-old infants and their caregivers from the South-East
of the UK. Our findings suggest a disconnect between what infants communicate and their
physiological arousal levels, that are likely to reflect what they experience. Specifically, in
households which families self-reported as being more chaotic, infants were more likely to
produce negative affect vocalisations such as cries at lower levels of arousal. This
disconnection between signalling and autonomic arousal was also present in a lab still face
procedure, where infants from more chaotic households showed reduced change in facial
affect and slower physiological recovery despite equivalent change in arousal during the still
face episode. Finally, we found that this disconnect between what infants communicate and
their physiological arousal levels may influence the likelihood of a caregiver responding.
Implications for understanding the mechanisms underlying the relationship between
household chaos, emotion dysregulation and caregiver under-responsivity are discussed.
Keywords: emotion regulation, self-regulation, arousal regulation, caregiver sensitivity,
Running head: CHAOS, VOCALISATIONS, AROUSAL - 4 -
A lack of family and weekly routines, high noise levels, disorganisation and crowding are all
part of household chaos (Evans & Wachs, 2010; Matheny Jr. et al., 1995). Higher levels of
household chaos have been related to worse child outcomes across a range of areas including
cognitive and academic development (Shamama-tus-Sabah et al., 2011), language (Martin et
al., 2011) and mental and physical health outcomes (Coldwell et al., 2006; Marsh et al.,
2020; Mills-Koonce et al., 2016). One finding in particular that has attracted attention is the
well-replicated relationship found between household chaos and affect dysregulation,
evaluated using a variety of measures including response to challenge (Vernon-Feagans et al.,
2016), delayed gratification (Martin et al., 2011), caregiver-reported emotion regulation
(Miller et al., 2017) and caregiver- and clinician-rated behaviour problems (Bobbitt &
Gershoff, 2016; Coldwell et al., 2006; Deater‐Deckard et al., 2009) (see (Marsh et al., 2020)
for a review).
How and why chaos impacts on affect dysregulation remains unclear, but research suggests
that this is mediated by an impact on caregiver responsivity. Higher levels of household
chaos are known to relate to less responsive and more intrusive caregiving styles (Andeweg
et al., 2020; Deater‐Deckard et al., 2012; Dumas et al., 2005; Geeraerts et al., 2021). Vernon-
Feagans and colleagues examined the relationship between household chaos, early executive
function and behavioural regulation, and caregiver responsivity in a longitudinal study of
1145 infants. Household chaos and caregiver responsivity were rated by trained researchers at
each of the 5 home visits, and executive function tests were administered at older ages.
Results suggested that the relationship between household chaos and executive function was
Running head: CHAOS, VOCALISATIONS, AROUSAL - 5 -
mediated by caregiver responsivity (Vernon-Feagans et al., 2016) (see also (Geeraerts et al.,
2020; Song et al., 2018)).
The key to appreciating how caregiver under-responsivity leads to child affect dysregulation
is likely to lie in understanding the dynamic bi-directional influences between children and
the caregivers. Infants are thought to learn self-regulation by experiencing repeated cycles of
co-regulation with their caregiver; over time, the infant internalises the expectation of the
caregiver’s soothing response and through this learns self-regulation (Bronfenbrenner, 1977;
Ham & Tronick, 2009a; Kopp, 1982; Olson & Lunkenheimer, 2009). In home settings, for
example, caregivers respond to their infants’ distress by upregulating their own arousal state
to match their child’s, leading to short-term increases in arousal synchrony; greater caregiver
responsivity to infant stress associates with faster infant quieting (Wass et al., 2019, 2021). In
lab settings, studies have similarly shown that the recovery period of the infant following a
‘still face’ procedure is dependent on the response of the caregiver during the ‘reunion’ phase
(Enlow et al., 2014; Feldman, Gordon, et al., 2010; Provenzi et al., 2015a).
These studies have, however, concentrated on understanding how caregiver under-
responsivity is driven by caregiver adapting (or not) to their child. Less research has
investigated bi-directional links, examining simultaneously the role of the child as a sender in
these exchanges, as well as how the caregiver responds to the child. From birth, infants
typically express signs of heightened arousal and distress through cries and facial
expressions. It is thought that initially, infants mostly express these communicative displays
automatically - that is, directly as a function of autonomic arousal and physiological
responses to their environment (Craig, 1992; Zeskind, 2013). This predicts that, early on,
physiological states of arousal and facial and vocal displays of stress/distress should typically
Running head: CHAOS, VOCALISATIONS, AROUSAL - 6 -
align; e.g., in the absence of dysregulation, infants should mostly cry when they are aroused,
and not otherwise (Zeskind, 2013). Over time, infants develop the ability to use displays
flexibly and intentionally (Feldman, 2007a; Matthews, 2020), according to specific display
rules that are culture-specific. The onset of such intentional communication can for instance
be seen in gaze-coordinated vocalizations – where children vocalize while looking towards
their caregivers, or alternate their gaze between their caregiver and a situation or object
(Bates, 2014; Donnellan et al., 2020; Schieffelin, 2016).
Caregiver responsivity is thought to be crucial for this transition towards intentional
communication (Albert et al., 2018; Locke, 2006; Matthews, 2020). Infants’ dynamic
behaviours, such as crying and smiling, influence their caregivers’ behaviours (Feldman,
2007a), and negative vocalisations including cries elicit greater (Tronick, 2007; Wass et al.,
2019, 2021) and faster (Yoo et al., 2018) responses from caregivers, while vocalisations that
are more speech-like prompt more contingent responses (Albert et al., 2018). This signifies
that caregivers are sensitive to variations in the acoustic and phonetic characteristics of their
infants’ vocalisations, and interpret them as serving various functions (e.g., cries are typically
perceived as requiring a more urgent response). Yet it remains unclear whether, and if so
how, infants’ vocal and facial displays affect caregiver responsivity differently as a function
of their arousal level, and how these relationships may vary as a function of environmental
Few previous research studies have investigated simultaneously the relationship between
infants’ displays, their autonomic arousal, and caregiver responsivity. One recent study
looked at vocalisation and arousal co-fluctuations across the day in naturalistic home
environments; it found that negative affect vocalisations (including cries, which accounted
Running head: CHAOS, VOCALISATIONS, AROUSAL - 7 -
for 98% of negative vocalisations), occurred most frequently when the infant was highly
aroused and with lower arousal stability (Wass et al., 2021) (see also (McFarland, 2001;
McFarland et al., 2020; Wilder, 1974)). These types of vocalisations also elicited the highest
frequency of caregiver contingent responses when compared to more positive vocalisations.
High intensity infant vocalisations also coincided with bigger infant arousal changes,
followed by a period of arousal stability across the dyad. This confirms that cries – as well as
being particularly salient for caregivers – are an important tool for establishing co-regulation
within the dyad (Wass et al., 2019, 2021). It thus seems important to understand what
happens when a noisy, chaotic environment impacts how caregiver responsivity relates to
infants’ expressions of distress, potentially in turn affecting their self-regulatory skills.
To address this, in study 1 we used miniature wearable microphones and autonomic monitors
to obtain day-long recordings in home settings from a cohort of N=74 12-month-old infants
across the South-East of the UK. We examined how autonomic arousal, in infant and
caregiver, changed relative to naturally occurring infant vocalisations during the day.
Trained, blinded coders manually coded all infant vocalisations across two dimensions: first,
vocal affect, ranging from negative (fussy and difficult) to positive (happy and engaged); and
second, vocal intensity, reflecting the intensity with which the affect was expressed. For
technical reasons our microphones recorded a 5-second sample every minute and therefore
our analyses examined only large-scale arousal changes during the 10 minutes before and
after each vocalisation. Importantly, although our recordings do not capture all vocalisations
that occurred, the presence of undetected vocalisations can only have weakened any patterns
of event-related change that we did observe. Participating caregivers also attended a lab visit,
where the still face procedure (Ham & Tronick, 2009b), a widely used test of emotion
regulation in infancy, was recorded (study 2).
Running head: CHAOS, VOCALISATIONS, AROUSAL - 8 -
We hypothesised that caregiver sensitivity to their infant’s vocal (study 1) and facial (study 2)
displays of distress would be lower in high chaos households, and that as a consequence there
would be a more flexible association between arousal and displays in these infants. In study
1, we predicted that the level of caregiver responsivity would be related to infants’ vocal
displays; that lower levels of responsiveness would be related to less specific vocal displays;
and that vocal displays would be related to the infant’s autonomic arousal levels to varying
degrees, dependent on the level of caregiver responsivity. In study 2, we hypothesized that
there would be a similar disconnection between physiological arousal and facial displays in
infants from chaotic households during the still face episode, and that this should be
paralleled by reduced caregiver sensitivity.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 9 -
Experimental participant details
The project was approved by the Research Ethics Committee at the University of East
London. Participants were recruited from the London, Essex, Hertfordshire and Cambridge
regions of the UK. In total, 91 infant-caregiver dyads were recruited to participate in the
study, of whom usable autonomic data were recorded from 78. A further 4 participants failed
to return the chaos questionnaire. Further details, including exclusion criteria, and detailed
demographic details on the sample, are given in SM section 1.1 and Table S1. Of note, we
excluded families in which the primary day-time care was performed by the male caregiver,
because the numbers were insufficient to provide an adequately gender-matched sample. All
participating caregivers were, therefore, female. Participants received £30 in Love2Shop gift
vouchers as a token of gratitude for participation, split over two visits.
The Confusion Hubbub And Order Scale (henceforth ‘chaos’) questionnaire (Matheny Jr. et
al., 1995) asks caregivers to self-rate on a series of statements such as “it’s a real zoo in our
home” and “the atmosphere in our home is always calm”. Figure S1 (SM section 1.2) shows a
histogram of the results obtained from the questionnaire. For the time series analyses reported
in parts 2b, 2c, 3a-c and 4, where it is not possible to look at group differences based on a
continuous group variable, data have been split using a median split, with quartile splits also
reported in the Supplementary Materials. The median was 33; Table S1 provides a break-
down of the demographic data split by high/low chaos score, and associations between chaos
score and other demographic variables are reported in the results.
Experimental method details – home data
Running head: CHAOS, VOCALISATIONS, AROUSAL - 10 -
Participating caregivers were invited to select a day during which they would be spending the
entire day with their child but which was otherwise, as far as possible, typical for them and
their child. The researcher visited the participants’ homes in the morning (c. 7.30 - 10am) to
fit the equipment, and returned later (c. 4 - 7pm) to pick it up. The mean (std) recording time
per day was 7.3 (1.4) hours.
The equipment consisted of two wearable layers, for both infant and caregiver (see Figure 1).
For the infant, a specially designed baby-grow was worn next to the skin, which contained a
built-in Electrocardiogram (ECG) recording device (recording at 250Hz), accelerometer
(30Hz), Global Positioning System (GPS) (1Hz), and microphone (11.6kHz). For technical
reasons (limited storage capacity), the microphone recorded a 5-second snapshot of the
auditory environment every 60 seconds.
A T-shirt, worn on top of the device, contained a pocket to hold the microphone and a
miniature video camera (a commercially available Narrative Clip 2 camera). For the
caregiver, a specially designed chest strap was also worn next to the skin, containing the
same equipment. A cardigan, worn as a top layer, contained the microphone and video
camera. The clothes were comfortable when worn and, other than a request to keep the
equipment dry, participants were encouraged to behave exactly as they would do on a normal
Running head: CHAOS, VOCALISATIONS, AROUSAL - 11 -
Figure 1 – Equipment. a) picture showing the equipment used in the study. b) picture
showing a caregiver and child wearing the equipment.
Figure 2 raw data sample showing, from top to bottom: infant arousal composite score (see
SM sections 1.3-1.6); infant arousal after removal of the autocorrelation (AR (see SM section
1.7); infant vocal affect and infant vocal intensity (see Methods section); caregiver arousal;
caregiver arousal after removal of the autocorrelation.
Quantification and statistical analysis – home data
Running head: CHAOS, VOCALISATIONS, AROUSAL - 12 -
Autonomic data parsing and calculation of the autonomic composite measure. Further
details on the parsing of the heart rate (section 1.3; Figure S2, S3), heart rate variability
(section 1.4), and actigraphy (section 1.5) are given in the SM. As shown in the SM (Figure
S4) these three variables were highly interdependent, and so we collapsed them into a single
composite measure of autonomic arousal (see SM section 1.6 for further details). In section
1.7 we present a description of how the autocorrelation was removed from the arousal data.
Because we wished to examine large-scale arousal changes, all data were downsampled to
one-minute epochs for all analyses.
Home/Awake coding. Our preliminary analyses suggested that infants tended to be strapped-
in to either a buggy or car seat for much of the time that they were outdoors, which strongly
influenced their autonomic data. For this reason, all the analyses presented in the paper only
include data segments in which the dyad was at home and the infant was awake. A
description of how these segments were identified are given in the SM (section 1.8).
Following these exclusions, the mean (std) total amount of data available per dyad was 3.7
(1.7) hours, corresponding to 221.5 (102.4) 60-second epochs per dyad.
Vocal affect and intensity coding. Post hoc, trained, blinded coders identified samples in
which the infant or caregiver was vocalising, and coded them on a five-point scale for vocal
affect (negative: fussy and difficult, neutral or positive: happy and engaged), and on a five-
point scale for vocal intensity of the affect expressed, from 1 (least intense) to 5 (most
intense). Consistency of rating between coders was achieved through discussions and joint
coding sessions based on an ersatz dataset, before the actual dataset were coded. In order to
assess inter-rater reliability, 24% of the sample was double coded; Cohen’s kappa was 0.60,
which is considered acceptable (McHugh, 2012) . All coders were blind to study design,
Running head: CHAOS, VOCALISATIONS, AROUSAL - 13 -
participant details and hypothesised study outcome. In total, a mean (std) of 159 (36)
vocalisations per infant was analysed.
Permutation-based temporal clustering analyses. To estimate the significance of time-
series relationships, a permutation-based temporal clustering approach was used. This
procedure, which is adapted from neuroimaging (Maris, 2012; Maris & Oostenveld, 2007),
allows us to estimate the probability of temporally contiguous relationships being observed in
our results, a fact that standard approaches to correcting for multiple comparisons fail to
account for (Maris, 2012) (see also (Oakes et al., 2013)). See further details in SM section
Control analysis. Participant by participant, for each vocalisation that was observed, a
random ‘non-vocalisation’ moment was selected as a moment during the day when the dyad
was at home and the infant was awake but no vocalisation occurred. The same moving
window analysis described above was then repeated to examine change relative to this ‘non-
vocalisation event’. The same procedure was repeated 1000 times and the results averaged.
Real and observed data were compared using the permutation-based temporal clustering
analyses described above.
Experimental method details – lab data
The task was a standard version of the still face protocol (Weinberg & Tronick, 1996).
Caregiver and child were seated across a 80cm-wide table, and instructed to play naturally with
four toys positioned on the table. After four minutes, on an instruction from the experimenter,
Running head: CHAOS, VOCALISATIONS, AROUSAL - 14 -
the caregiver was instructed not to respond to the infant and to hold a neutral face for two
minutes. On a further instruction from the experimenter, the play resumed for a further two
minutes. If the infant became distressed during the still face period, as judged using the standard
guidelines (Weinberg & Tronick, 1996), the experiment was curtailed.
Quantification and statistical analysis – lab data
Facial and vocal affect coding. Facial and vocal affect was coded in 5-second bins using a
5-point scale, where -2 is extreme negative affect, 0 is neutral, and +2 is extreme positive
affect. To ascertain inter-rater reliability, 20% of the sample was double coded, and Cohen’s
Kappa was calculated. Inter-rater reliability was found to be 0.66, which is considered
Autonomic data parsing. The parsing of the ECG data was conducted using the same
procedures as used for the home data. Because of technical problems with equipment during
the recording of heart rate data from the still face protocol, data from this task are only
available from a N=17 subset of the sample. Further details of the parsing of the ECG data
are given in the SM (section 1.3; Figure S2, S3).
Running head: CHAOS, VOCALISATIONS, AROUSAL - 15 -
The results section is structured as follows. First (part 1) we include descriptive statistics,
showing how scores on the chaos scale correlated with household noise, demographics,
caregiver mental health, tonic arousal, and average vocalisation rate affect and intensity.
Second (part 2) we examine infant arousal around vocalisations in home settings (study 1),
subdivided by household chaos. In part 2a we examine the two-way split between arousal and
vocalisation affect/intensity; in part 2b we examine infant arousal changes around
vocalisations; in part 2c we examine the temporal clustering of infant vocalisations.
Third (part 3) we examine caregiver arousal around infant vocalisations in home settings
(study 1). In part 3a we examine caregiver arousal changes around vocalisations, subdivided
by household chaos; in part 3b we examine caregiver arousal changes around vocalisations,
subdivided by infant arousal at the time of the vocalisation; in part 3c we examine the
temporal clustering of adult vocalisations around infant vocalisations.
Fourth (part 4) we examine infant behaviour and physiological changes during the lab-based
still face protocol (study 2).
Part 1 - Descriptives
See Table S1 for demographic split.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 16 -
Before conducting our primary analyses (described from Part 2 onwards) we first calculated
some descriptive statistics to report on: i) face validity of the chaos scale – association with
microphone noise and people in the household; ii) association of chaos with demographics
and caregiver mental health; iii) association of chaos with tonic arousal; iv) association of
chaos with vocalisation rate, affect and intensity.
Face validity of chaos scale – association with microphone noise levels and people in the
household. First, we examined the face validity of the scale by examining the relationship
between household noise levels (taken from the average dB levels on the microphones worn
by infants) and caregiver household chaos ratings. Because not all variables were normally
distributed, more conservative non-parametric Spearman correlations are reported
throughout. A significant association was identified between chaos and microphone noise
levels while the infant was sleeping (rho=.28, p=.016), and a marginally non-significant
association was identified between chaos and ambient waking microphone noise levels (rho
=.22, p=.069). An association was also observed between chaos and the number of people
living in the household (rho =.41, p<.001).
Demographics and caregiver mental health. We examined the relationship of the chaos scale
to the demographic variables recorded: maternal education and household income. In our
sample, no significant relationships were identified (all ps>.55). In addition, we examined the
relationship of the chaos scale to caregiver mental health, as assessed using the Generalised
Anxiety Disorder 7-item screen (GAD-7) (Spitzer et al., 2006) and the Patient Health
Questionnaire 9-item (PHQ-9) screen for depression (Kroenke et al., 2001). An association
was observed between chaos and caregiver depression rho =.23, p=.026 but not anxiety
Running head: CHAOS, VOCALISATIONS, AROUSAL - 17 -
Association with tonic arousal. First we examined the association between tonic arousal,
defined from average heart rate, heart rate variability and movement levels in infant and
caregiver. Two categories were examined: segments while the dyad was at home and the
infant was awake; and those where the infant was asleep. No significant associations were
found between any of these variables and household chaos (all ps<.34).
Association with vocalisation rate, affect and intensity. No association was observed between
chaos and average infant vocal intensity (rho=-.07, p=.60); a marginally non-significant
positive association was observed between chaos and infant vocal affect (rho=.24, p=.084)
(higher chaos associated with more positive affect). No association was observed between
chaos and the vocalisation rate of infants or adults (rho=-.04/0.12, both ps>.3), but a
marginally non-significant positive association was noted between chaos and the proportion
of adult vocalisations that were infant-directed (rho=-.22, p=.098) (higher chaos associated
with a lower proportion of infant-directed vocalisations).
Part 2 – Infant arousal around vocalisations
a) Stacked bar charts
In order to examine how the relationship between vocalisation type and autonomic arousal
differed contingent on household chaos we first subdivided our all vocalisations by vocal
affect and physiological arousal, and plotted two separate stacked bar charts for the low and
high chaos groups, determined by a median split (Figure 3a). In the SM section 2.1 (Figure
S5) we also include the same plots subdivided using a quartile split.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 18 -
Statistical analyses were conducted by calculating ANOVAs separately for each vocal affect
bin and each group to examine whether, for that bin, there was a significant relationship with
arousal. Multiple comparisons were corrected for using the Benjamini-Hochberg procedure
(Benjamini & Hochberg, 1995). Figure 3a shows the results of this analysis for the low (left)
and high (right) chaos groups. In the low chaos group there is – as expected - a significant
relationship between extreme negative affect vocalisations and arousal F(4,134)=6.7, FDR-
corrected p=.006, such that extreme negative affect vocalisations are more likely at elevated
arousal. This relationship is not present in the high chaos group (p=.62). For moderate
negative affect vocalisations in the low chaos group the same relationship was present but did
not survive correction for multiple comparisons (p=.074). Again, this relationship is absent in
the high chaos group (p=.62). In the SM (Fig S5) we present the same analysis based on a
quartile (rather than median) split by chaos, showing that this relationship is observed across
all quartiles and is most marked in the bottom (lowest chaos) quartile group.
Figure 3b shows the same relationship, but examining the correspondence between vocal
intensity and arousal. In the low chaos group there is a similar relationship for extreme
intense vocalisations, which are more likely at elevated arousal, although this relationship did
not survive correction for multiple comparisons (pre-correction p=.046; post-correction
p=.082). This relationship is absent in the high chaos group (p=.92). Although differences
also appeared evidence for low intensity vocalisations, these were non-significant due to the
relatively lower frequency of this type of vocalisation in our data. Again, in the SM (Fig S5)
we present the same analysis based on a quartile split by chaos, showing that the relationship
is continuous across the sample.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 19 -
Finally, in Figure 3c we examine the relationship between vocal affect and intensity. This
shows that, for both groups, extreme negative and positive affect vocalisations are more
common at high vocal intensity. ANOVAs showed highly significant associations between
affect and intensity across all bins examined; no differences contingent on chaos were
Overall, these results show that, in the low chaos group, negative affect vocalisations such as
cries are more likely to occur at elevated arousal. This relationship is absent in the high chaos
Running head: CHAOS, VOCALISATIONS, AROUSAL - 20 -
Fig 3: Stacked bar charts showing: a) the relationship between vocal affect and arousal at
the time of vocalisation, split between the low (left) and high (right) chaos groups. * shows
the significance of the statistical analyses reported in the main text * - p<.05 after correction
for multiple comparisons; (*) – p<.05 before but not after correction; b) the relationship
between vocal intensity and arousal at the time of vocalisation, split between low (left) and
high (right) chaos groups; c) the relationship between vocal intensity and vocal affect, split
between low (left) and high (right) chaos households. For c), statistical analyses showed that
all bins in both groups were significant.
b) Arousal changes before and after vocalisations
Next, we examined arousal changes around vocalisations. First, we calculated the average
change in infant arousal around all negative (Figure 4a) and positive affect infant
vocalisations, subdivided using a median split into low/high chaos groups. In the SM we also
include the same analysis based on a quartile split by household chaos, to examine how
continuously these group differences were observed in our data (see SM Figure S6).
For each plot, the average change in arousal for the period from 10 minutes before to 10
minutes after each vocalisation is shown. In order to analyse the significance of group
differences, a permutation-based clustering analysis was conducted to correct for multiple
comparisons, as described in the Methods. For negative affect vocalisations, two significant
(p=.001) clusters were found: one between -1 to +1 minutes, and one from +4 to +5 minutes
following the vocalisation (Fig 4a), suggesting that the high chaos group showed smaller
arousal changes around vocalisations. For positive affect vocalisations, a significant (p=.025)
difference was observed from +1 to +2 minutes following the vocalisation (Fig 4b). For low
Running head: CHAOS, VOCALISATIONS, AROUSAL - 21 -
intensity vocalisations, a significant (p=.018) difference was observed from +2 to +3 minutes
following the vocalisation (Fig 4c); for high intensity vocalisations a significant (p=.025)
difference was observed from -3 to +3 minutes around vocalisations. In each case the
direction of these effects was the same: the high chaos group showed smaller arousal changes
around vocalisations. Results presented in the SM (Figure S6) show that, when the analysis is
repeated using a quartile split by household chaos, the relationship is continuous across the
Overall, these results show infants in the low chaos group show larger changes in autonomic
arousal across all types of vocalisations; differences appear most marked for negative affect,
and high intensity, vocalisations. This suggest that there is a disconnection between vocal
displays and physiological arousal in the high chaos group.
Figure 4: Line graphs showing average infant arousal around different types of
vocalisations, split by median split into children from low (blue) and high (red) chaos
households. a) Negative affect vocalisations; b) positive affect vocalisations; c) low intensity
Running head: CHAOS, VOCALISATIONS, AROUSAL - 22 -
vocalisations; d) high intensity vocalisations. Shaded areas show standard error. * - sections
identified as showing significant group differences by the permutation-based cluster analysis
c) Temporal clustering of vocalisations
Finally, as an additional analysis, we examined how likely infant vocalisations are to occur in
clusters, and how this differs contingent on household chaos, and between negative and
positive affect vocalisations. To do this we used the following procedure: for each
vocalisation, we calculated the average likelihood of another vocalisation during a given one-
minute time window (e.g. 1-2 minutes following an initial infant vocalisation – see Figure
5a). To estimate whether the observed likelihood differed from chance we performed a
control analysis in which we inserted random ‘non-vocalisation’ events into the data and
repeated the analysis relative to these ‘non-vocalisations’, and compared the ‘real’ and
‘control’ datasets using a nonparametric Mann-Whitney U test (see Figure 5a). We then
repeated this analysis across multiple time windows from 10 minutes before the vocalisation
to 10 minutes after.
For each time window, the significance of each Mann-Whitney U test compared the observed
vocalisation rate with change is shown in Figure 5b: a large dot indicates a significant
difference between the observed and the control data (i.e. that a vocalisation was significantly
more likely to occur during that time window compared with chance). Our results showed
that, in the low chaos group, a significantly greater than chance vocalisation likelihood was
observed from 9 minutes before each vocalisation to 7 minutes after (although not all
intermediate bins were significant – see Fig 5b). In the high chaos group, greater than chance
Running head: CHAOS, VOCALISATIONS, AROUSAL - 23 -
vocalisation rates were observed from 8 mins before to 10 mins after (although again not all
intermediate bins were significant).
In addition, we also directly compare the vocalisation rates between the high and low chaos
groups, using a Mann-Whitney U test and a permutation-based clustering analysis to correct
for multiple comparisons. This test identified a significantly higher vocalisation rate (p<.001)
in the low chaos group from -1 to +1 mins following the vocalisation, suggesting higher
caregiver responsivity for negative vocalisations in this group.
Figure 5c shows and identical plot of positive affect vocalisations. In the low chaos group,
greater than chance vocalisation likelihoods were observed between 8 minutes before each
vocalisation to 10 minutes after (although again not all intermediate time windows were
significant). In the high chaos group, greater than chance vocalisation rates were observed
between 9 mins before to 10 mins after. No direct group comparisons were significant,
suggesting no differences in caregiver responsivity for positive vocalisations.
Overall this suggests that, in the low chaos group, negative affect vocalisations are more
likely to be accompanied by other vocalisations during the period immediately around the
event. No significant group differences were observed for positive affect vocalisations.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 24 -
Fig 5: a) sample violin plot showing the analysis for one time interval that was then repeated
iteratively across multiple time intervals in b. The plot shows the likelihood of a subsequent
negative affect vocalisation in the time window 0-1 minutes following a negative affect
vocalisation, comparing real with control data. b) same analysis repeated across multiple
time windows, subdivided between low (red) and high (blue) chaos households. For each
time window, the size of the dot indicates the significance of the Mann-Whitney U test
comparing the observed and control data. Large dot – p>.05; small dot – p<.05. * - sections
identified as showing a significant group differences by the permutation-based cluster
Part 3 – Caregiver arousal around vocalisations
a) Caregiver arousal changes before and after infant vocalisations
In order to better understand why caregivers were less responsive in the high-chaos group we
also examined how caregiver arousal patterns change relative to infant vocalisations (Figure
6). These analyses were conducted identically to those described in Part 2 above (Figure 4),
but examining caregiver arousal changes relative to infant vocalisations, rather than infant
Running head: CHAOS, VOCALISATIONS, AROUSAL - 25 -
arousal changes. Figure S7 shows the same analysis, but subdivided using a quartile split by
For all infant vocalisations (Fig 6a), permutation-based clustering analyses revealed a
significant (p=.023) group difference from 0 to +2 minutes and from +5 to +6 minutes
following the vocalisation (Fig 6a), such that caregivers in the high chaos group showed
smaller arousal changes around infant vocalisations. For negative affect vocalisations, a
significant (p=.045) difference was observed from -1 to 0 and from +1 to +2 minutes relative
to infant vocalisations (Fig 6b). For positive affect vocalisations (Fig 6c) no group differences
were observed. Results presented in the SM (Figure S7) show the same analysis repeated
using a quartile split by household chaos. These analyses show that the relationship is linear
across the sample, and driven by greater caregiver arousal reactivity to infant vocalisations in
the bottom quartile (lowest household chaos) subgroup.
No significant group differences were observed when we examined caregiver arousal changes
to infant vocalisations subdivided by vocal intensity, subdivided by household chaos (Fig S8a
and S8b). However, the same analysis broken down using a quartile split by household chaos
(Fig S8c, S8c) show that, again, greater caregiver arousal reactivity to infant vocalisations is
observed in the bottom quartile (lowest household chaos) subgroup.
Overall this suggests that caregivers in the low chaos group show larger changes in
autonomic arousal around infant vocalisations, and that this is driven by greater responsivity
to negative affect vocalisations.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 26 -
Fig 6: Line graphs showing average caregiver arousal around different types of infant
vocalisations, split by median split into children from low and high chaos households. a) all
vocalisations; b) negative affect vocalisations; c) positive affect vocalisations. Shaded areas
show standard error. * - sections identified as showing significant group differences by the
permutation-based cluster analysis *<.05.
b) Caregiver arousal changes before and after infant vocalisations – subdivided by infant
We also repeated the same analyses as shown in Figure 6, but instead of subdividing
vocalisations by household chaos and manually coded vocal affect, we subdivided all
vocalisations according to the infant’s arousal at the time of the vocalisation. Separately for
each infant we performed a median split to differentiate between high and low arousal
vocalisations (relative to the average arousal level for that child). Results are shown in Figure
7a. The same permutation-based temporal clustering analysis was performed as described
above. This suggested that high arousal vocalisations were accompanied by significantly
(p<.001) greater changes in caregiver arousal during the time period from 0 to +3mins after
the vocalisation. We then further subdivided these vocalisations by household chaos, to create
four categories – high arousal vocalisations from the low chaos group, high arousal
vocalisations from the high chaos group, low arousal vocalisations from the low chaos group,
Running head: CHAOS, VOCALISATIONS, AROUSAL - 27 -
and low arousal vocalisations from the high chaos group. We found that only the low chaos
group appeared to respond differentially to high arousal vocalisations (Fig 7b).
Overall, this suggests that caregiver arousal changes are greater following high arousal infant
vocalisations compared with low arousal vocalisations. However, this finding is limited to the
low chaos group.
Fig 7: a) Line graph showing average caregiver arousal around infant vocalisations, split by
median split by the infant’s arousal at the time of the vocalisation. b) Same figure as 7a, but
further subdivided by household chaos, to create four categories. For a-b, shaded areas show
standard error. * - sections identified as showing significant group differences by the
permutation-based cluster analysis *<.05. c) sample violin plot showing the analysis for one
time interval that was then repeated iteratively across multiple time intervals in c. The plot
shows the likelihood of a subsequent caregiver vocalisation in the time window 0-1 minutes
following an infant vocalisation, comparing real with control data. d) same analysis repeated
Running head: CHAOS, VOCALISATIONS, AROUSAL - 28 -
across multiple time windows, subdivided between high and low arousal vocalisations. e)
same analysis repeated across multiple time windows, further subdivided by household
chaos. For c-e, the size of the dot indicates the significance of the Mann-Whitney U test
comparing the observed and control data. Large dot – p>.05; small dot – p<.05. * indicates
sections identified as showing a significant group differences by the permutation-based
cluster analysis, p<.05.
c) Caregivers’ vocal responsivity
We also repeated the same analyses as reported in Part 2c to examine the likelihood of
caregivers vocalising during the time period before and after infant vocalisations. Figure 7d
shows caregiver vocalisation likelihood subdivided between high and low arousal infant
vocalisations. As with the plots in Figure 5, the size of the dots in Fig 7d indicates the
significance of the statistical analyses comparing the observed vocalisation rate with chance;
the results suggested that, in all bins examined (apart from -9 mins for low arousal vocals) the
observed vocalisation rates were greater than chance. In addition, the observed vocalisation
rates were directly compared between the two conditions. Permutation-based temporal
clustering analyses suggested that caregivers are significantly more likely to vocalise during
the time period 0 to +2 mins following high arousal vocalisations. Figure 7e shows the same
plot, but further subdivided by household chaos. Overall, these results suggest that caregivers
are more likely to vocalise in the time period following high arousal infant vocalisations, with
the highest response rate observed from caregivers in the low chaos subgroup in response to
high arousal vocalisations.
Part 4 – Child affect regulation – lab battery
Running head: CHAOS, VOCALISATIONS, AROUSAL - 29 -
Figure 8a shows the experimental set-up that we used for the still face procedure. First, we
examined the change in infant facial affect during the still face procedure, subdivided into
high and low chaos groups (Fig 8b). The same permutation-based temporal clustering
analysis as described above was applied to test for group differences while correcting for
multiple comparisons. This identified a significant (p=.034) difference in facial affect in the
time period 1:20 to 1:40 during the still face procedure, such that the high chaos group
showed less negative affect. By contrast, when we examined physiological changes, we
found no significant difference between groups (although the N for this analysis is lower, as
described in the Methods). However, we did identify a trend-level correlation between chaos
and heart rate recovery following the still face (rho=.47, p=.056), such that higher household
chaos associated with slower recovery following the still face. Overall, these results suggest
that infants from high chaos households display less negative facial affect during the still
face, along with a trend towards slower physiological recovery afterwards.
Fig 8: a) screen grabs illustrating the experimental set-up; b) line graph showing change in
facial affect during the still face, subdivided by high and low household chaos; c) line graph
showing change in heart rate during all three phases of the experiment, subdivided by high
and low household chaos. Shaded areas show standard error. * - sections identified as
showing significant group differences by the permutation-based cluster analysis *<.05.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 30 -
Using day-long home recordings obtained using miniaturised wearable autonomic monitors
and microphones, we examined caregiver-infant physiological arousal and vocal affective
displays in high- and low-chaos households. We examined within-individual relationships
(how infant autonomic arousal relates to infant vocal affects) and cross-dyad relationships
(how caregiver arousal relates to infant vocalisations). We also examined differences in facial
displays and autonomic arousal in infants from the different households during the still face
paradigm. From our results the following conclusions can be drawn:
First, children in the low chaos group are more likely to produce negative affect vocalisations
such as cries at elevated arousal; this relationship is absent in the high chaos group (Fig 3).
Second, children from high chaos households showed smaller arousal changes around
vocalisations, in particular around negative affect and high intensity vocalisations such as
cries (Fig 4). The findings from the high-chaos group are in contrast with those from the low-
chaos group, and from previous research in general populations (i.e. not subdivided by
household chaos level) (Kreibig, 2010; Wass et al., 2021) which show that negative
vocalisations occur most often at high arousal states. This is consistent with the idea that
initially, infant vocalisation are not functionally flexible, but directly determined by levels of
arousal (Ghazanfar & Zhang, 2016; Wass et al., 2021; Zeskind, 2013). This suggests that
high chaos infants are not using high intensity vocalisations to communicate their high
arousal states, revealing a disconnect between infants’ levels of arousal – which is likely to
reflect what they are experiencing - and what they are actually expressing.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 31 -
Importantly, we observed no relationship between household chaos and infant vocalisation
rates, and no relationship between household chaos and overall vocalisation intensity or vocal
affect. Thus, it is not the case that infants in the high-chaos group are experiencing different
autonomic arousal, or vocalizing differently overall. Instead, it is how their vocalizations
relate to their arousal levels that is atypical, both as compared to the low-chaos group in the
present study, and as compared to what we expected based on past research (Wass et al.,
Our results also suggest that caregivers from high chaos households show smaller arousal
changes around infant vocalisations (Fig 6a), in particular around negative affect
vocalisations (Fig 6b). When we examined how caregiver responsivity varies as a function of
infant arousal at the time of the vocalisation (Fig 7) we found that all caregivers are less
responsive to low arousal vocalisations (Fig 7a, 7d). However, when we subdivided this by
household chaos we found that, whereas low chaos caregivers did differentiate between high
and low arousal vocalisations, high chaos caregivers did not (Fig 7b, 7e). In other words,
infant arousal at the time of the vocalisation influenced the likelihood of a caregiver response
in low chaos households, but the same relationship was not observed in high chaos
This is relevant to theories of selective reinforcement in communication (Albert et al., 2018;
Goldstein & Schwade, 2008; Locke, 2006; Oller & Griebel, 2020; Zhang & Ghazanfar,
2016). Our findings suggest that specific signals perhaps lose the potential functional
significance that we assume they had in earlier infancy for children in high chaos households,
potentially resulting in negative affect vocalisations such as cries becoming less specific.
Consistent with this is our finding that, in the low chaos group, negative affect vocalisations
such as cries are more likely to occur at elevated arousal, but this relationship is absent in the
Running head: CHAOS, VOCALISATIONS, AROUSAL - 32 -
high chaos group (Fig 3a). Possibly this indicates that, in contrast with previous studies in
general populations which found that negative affect vocalisations were more likely to occur
at high arousal states (Tronick, 2007; Wass et al., 2019, 2021), these infants have learned that
expressing distress when aroused does not have consequences, so they stop doing it. It
doesn’t matter how they communicate, the resulting feedback will be the same.
Interestingly, we found a similar disconnection between affective displays and physiological
arousal in a lab-based still face protocol: children from high chaos households showed less
negative facial affect (Fig 8b). This contrasts with previous selective reinforcement
interpretations of the still face paradigm in general populations, in which infants are said to
increase their vocal and facial displays in order to gain a previously conditioned response
from their caregiver (Ham & Tronick, 2009b). For instance. Goldstein and colleagues found
that 5-month-old infants responded to their caregiver’s neutral face with increased
behavioural signals (clusters of vocalisations) followed by quietness. The authors interpreted
this as a classic extinction burst (Goldstein et al., 2009).
Here, our findings mirror those in the more naturalistic data: infants from high chaos
household show similar levels of arousal during the still face; but, in contrast with children
from low chaos households, they do not match this with their facial display. There is again a
disconnection between their physiological response and their communicative behaviour. This
could mean that there is no/under regulation by the caregiver of their child's distress during
the post-still face; or, it could indicate that this behaviour had perhaps not been previously
reinforced; and so facial affect too has possibly lost potential functionality (Conradt &
Ablow, 2010; Enlow et al., 2014; Feldman, Singer, et al., 2010; Gunning et al., 2013; Haley
& Stansbury, 2003; Ham & Tronick, 2009b; Provenzi et al., 2015b). Our autonomic data
show the physiological correlates of this, indicating that infants from high chaos household
Running head: CHAOS, VOCALISATIONS, AROUSAL - 33 -
showed a trend towards slower recovery of their arousal level (return to baseline) during the
post-still face period. Of note, though, the sample for this last analysis was lower than the
other analyses in this paper.
Taken overall, our findings thus reveal a disconnect between the arousal patterns in infants in
high chaos households, and their vocalisation behaviours. This disconnect is not seen in
infants from low chaos households. Previous research has highlighted the relationship
between caregiver-child arousal levels (arousal coupling) as an important mechanism for co-
regulation across the dyad (Feldman, 2007b; Smith et al., 2021; Wass et al., 2021). In part,
infant arousal levels are communicated by vocalisations to elicit a caregiver response
(Ghazanfar & Zhang, 2016; Oller & Griebel, 2020): infants are able to utilise their
vocalisations to regulate their arousal levels (and overall arousal level across the child-
caregiver dyad). If communication is in effect the sending and receiving of signals, the
fundamental starting point must be that the overall outcome is beneficial to both partners
(Oller et al., 2013). Our data indicate that in highly chaotic households, senders and receivers
are not aligned anymore: infants are not expressing their levels of arousal, and caregiver
arousal levels are under-responsive to their infants. Thus, communication is not happening
within these dyads, and the interactions are not ‘beneficial’ in a co-regulation sense.
Although technical factors meant that we were confined to random sampling during the day
rather than continuous recordings, our analyses examine event-related changes relative to
vocalisations, and so the presence of undetected vocalisations can only have weakened the
patterns of event-related change that we have observed. Another limitation (and strength) of
our approach is that, although we only included data segments recorded while the dyad was at
home and the infant was awake, our home-based recordings nevertheless contained all types
of vocalisations across a variety of physical settings.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 34 -
In future it would be interesting to explore the relationship between autonomic arousal,
vocalisations, caregiver responsivity, and psychopathology. For example, one factor that may
mediate our present findings is caregiver depression, which was positively associated with
household chaos in our sample. Previous research has shown that more depressed mothers
tend to be under-responsive to their infants’ vocalisations and behaviours (Beebe et al., 2008;
Field et al., 1990). Household chaos has previously been found to act as a mediator between
caregiver depression and child outcomes (Hur et al., 2015). One further explanation for our
findings could be the higher levels of stress experienced by adults in high chaos households
(Bodrij et al., 2021), and associated lower levels of sensitivity (Andeweg et al., 2020). Yet,
we note that there was no association with anxiety in our sample.
Overall, our present results suggested that, in infants from high chaos households, infant
vocalisations tend to lose the functional significance that they originally held; and that this is
in some way related to caregiver responsiveness; meaning that autonomic regulation
following a stressor was slower. But other work from our group has shown that, in a general
population, moments of high infant arousal were more likely to be accompanied by infant
vocalisations, and by high intensity and negative affect vocalisations, including cries; and that
this served as a mechanism for co-regulation across the dyad (Wass et al., 2021).
Understanding how arousal level and vocalisations likelihood relate to caregiver responsivity
across typical and atypical development is an important goal for future research. Relatedly,
longitudinal studies are needed to examine the association between infant arousal,
vocalisations, and caregiver responsivity across development; how, for example, do
vocalisations become functionally flexible over time, what is the relationship between
functional flexibility and arousal, and how is this affected by caregiver responsiveness and
Running head: CHAOS, VOCALISATIONS, AROUSAL - 35 -
In summary, our data suggest that caregiver responsiveness largely shapes infant’s
communication of their distress, and that this influence may lead to adverse outcomes in high
chaos households. In these households, infant vocalisations (in particular negative
vocalisations and high intensity vocalisations) are not more likely to occur at times when
infant arousal is elevated, and they do not elicit responsiveness in the caregiver. Our findings
also suggest that vocalisations are an important driver of co-regulation: when sender-receiver
communicative behaviours are finely attuned, which seems to be favoured by a less chaotic
environment, vocalisations elicit caregiver responsiveness, aiding recovery in the infant.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 36 -
Albert, R. R., Schwade, J. A., & Goldstein, M. H. (2018). The social functions of babbling:
acoustic and contextual characteristics that facilitate maternal responsiveness.
Developmental Science, 21(5), e12641.
Andeweg, S. M., Bodrij, F. F., Prevoo, M. J. L., Rippe, R. C. A., & Alink, L. R. A. (2020).
Does sensory-processing sensitivity moderate the effect of household chaos on caregiver
sensitivity? An experimental design. Journal of Family Psychology.
Bates, E. (2014). The emergence of symbols: Cognition and communication in infancy.
Beebe, B., Jaffe, J., Buck, K., Chen, H., Cohen, P., Feldstein, S., & Andrews, H. (2008). Six‐
week postpartum maternal depressive symptoms and 4‐month mother–infant self‐and
interactive contingency. Infant Mental Health Journal: Official Publication of The
World Association for Infant Mental Health, 29(5), 442–471.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and
powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B
Bobbitt, K. C., & Gershoff, E. T. (2016). Chaotic experiences and low-income children’s
social-emotional development. Children and Youth Services Review, 70, 19–29.
Bodrij, F. F., Andeweg, S. M., Prevoo, M. J. L., Rippe, R. C. A., & Alink, L. R. A. (2021).
The causal effect of household chaos on stress and caregiving: An experimental study.
Comprehensive Psychoneuroendocrinology, 100090.
Bronfenbrenner, U. (1977). Toward an experimental ecology of human development.
American Psychologist, 32(7), 513.
Coldwell, J., Pike, A., & Dunn, J. (2006). Household chaos–links with parenting and child
behaviour. Journal of Child Psychology and Psychiatry, 47(11), 1116–1122.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 37 -
Conradt, E., & Ablow, J. (2010). Infant physiological response to the still-face paradigm:
Contributions of maternal sensitivity and infants’ early regulatory behavior. Infant
Behavior and Development, 33(3), 251–265.
Craig, K. D. (1992). The facial expression of pain Better than a thousand words? APS
Journal, 1(3), 153–162.
Deater‐Deckard, K., Mullineaux, P. Y., Beekman, C., Petrill, S. A., Schatschneider, C., &
Thompson, L. A. (2009). Conduct problems, IQ, and household chaos: A longitudinal
multi‐informant study. Journal of Child Psychology and Psychiatry, 50(10), 1301–1308.
Deater‐Deckard, K., Wang, Z., Chen, N., & Bell, M. A. (2012). Maternal executive function,
harsh parenting, and child conduct problems. Journal of Child Psychology and
Psychiatry, 53(10), 1084–1091.
Donnellan, E., Bannard, C., McGillion, M. L., Slocombe, K. E., & Matthews, D. (2020).
Infants’ intentionally communicative vocalizations elicit responses from caregivers and
are the best predictors of the transition to language: A longitudinal investigation of
infants’ vocalizations, gestures and word production. Developmental Science, 23(1),
Dumas, J. E., Nissley, J., Nordstrom, A., Smith, E. P., Prinz, R. J., & Levine, D. W. (2005).
Home chaos: Sociodemographic, parenting, interactional, and child correlates. Journal
of Clinical Child and Adolescent Psychology, 34(1), 93–104.
Enlow, M. B., King, L., Schreier, H. M. C., Howard, J. M., Rosenfield, D., Ritz, T., &
Wright, R. J. (2014). Maternal sensitivity and infant autonomic and endocrine stress
responses. Early Human Development, 90(7), 377–385.
Evans, G. W., & Wachs, T. D. (2010). Chaos and its influence on children’s development.
Washington, DC: American Psychological Association.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 38 -
Feldman, R. (2007a). Parent–infant synchrony and the construction of shared timing;
physiological precursors, developmental outcomes, and risk conditions. Journal of Child
Psychology and Psychiatry, 48(3‐4), 329–354.
Feldman, R. (2007b). Parent–infant synchrony and the construction of shared timing;
physiological precursors, developmental outcomes, and risk conditions. Journal of Child
Psychology and Psychiatry, 48(3‐4), 329–354.
Feldman, R., Gordon, I., Schneiderman, I., Weisman, O., & Zagoory-Sharon, O. (2010).
Natural variations in maternal and paternal care are associated with systematic changes
in oxytocin following parent–infant contact. Psychoneuroendocrinology, 35(8), 1133–
Feldman, R., Singer, M., & Zagoory, O. (2010). Touch attenuates infants’ physiological
reactivity to stress. Developmental Science, 13(2), 271–278.
Field, T. M., Healy, B. T., Goldstein, S., & Guthertz, M. (1990). Behavior-state matching and
synchrony in mother-infant interactions of nondepressed versus depressed dyads.
Developmental Psychology, 26(1), 7.
Geeraerts, S. B., Backer, P. M., & Stifter, C. A. (2020). It takes two: Infants’ moderate
negative reactivity and maternal sensitivity predict self-regulation in the preschool years.
Developmental Psychology, 56(5), 869.
Geeraerts, S. B., Endendijk, J., Deater-Deckard, K., Huijding, J., Deutz, M. H. F., van den
Boomen, C., & Deković, M. (2021). The role of parental self-regulation and household
chaos in parent-toddler interactions: A time-series study. Journal of Family Psychology,
Ghazanfar, A. A., & Zhang, Y. S. (2016). The autonomic nervous system is the engine for
vocal development through social feedback. Current Opinion in Neurobiology, 40, 155–
Running head: CHAOS, VOCALISATIONS, AROUSAL - 39 -
Goldstein, M. H., & Schwade, J. A. (2008). Social feedback to infants’ babbling facilitates
rapid phonological learning. Psychological Science, 19(5), 515–523.
Goldstein, M. H., Schwade, J. A., & Bornstein, M. H. (2009). The value of vocalizing: Five‐
month‐old infants associate their own noncry vocalizations with responses from
caregivers. Child Development, 80(3), 636–644.
Gunning, M., Halligan, S. L., & Murray, L. (2013). Contributions of maternal and infant
factors to infant responding to the Still Face paradigm: A longitudinal study. Infant
Behavior and Development, 36(3), 319–328.
Haley, D. W., & Stansbury, K. (2003). Infant stress and parent responsiveness: Regulation of
physiology and behavior during still‐face and reunion. Child Development, 74(5), 1534–
Ham, J., & Tronick, E. (2009a). Relational psychophysiology: Lessons from mother–infant
physiology research on dyadically expanded states of consciousness. Psychotherapy
Research, 19(6), 619–632.
Ham, J., & Tronick, E. (2009b). Relational psychophysiology: Lessons from mother–infant
physiology research on dyadically expanded states of consciousness. Psychotherapy
Research, 19(6), 619–632.
Hur, E., Buettner, C. K., & Jeon, L. (2015). Parental depressive symptoms and children’s
school-readiness: The indirect effect of household chaos. Journal of Child and Family
Studies, 24(11), 3462–3473.
Kopp, C. B. (1982). Antecedents of self-regulation: A developmental perspective.
Developmental Psychology, 18(2), 199.
Kreibig, S. D. (2010). Autonomic nervous system activity in emotion: A review. Biological
Psychology, 84(3), 394–421.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 40 -
Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2001). The PHQ‐9: validity of a brief
depression severity measure. Journal of General Internal Medicine, 16(9), 606–613.
Locke, J. L. (2006). Parental selection of vocal behavior. Human Nature, 17(2), 155–168.
Maris, E. (2012). Statistical testing in electrophysiological studies. Psychophysiology, 49(4),
Maris, E., & Oostenveld, R. (2007). Nonparametric statistical testing of EEG-and MEG-data.
Journal of Neuroscience Methods, 164(1), 177–190.
Marsh, S., Dobson, R., & Maddison, R. (2020). The relationship between household chaos
and child, parent, and family outcomes: a systematic scoping review. BMC Public
Health, 20, 1–27.
Martin, A., Razza, R. A., & Brooks-Gunn, J. (2011). Specifying the links between household
chaos and preschool children’s development. Early Child Development and Care, iFirst
Matheny Jr., A., Wachs, T. D., Ludwig, J., & Phillips, K. (1995). Bringing order out of chaos:
Psychometric characteristics of the Confusion, Hub- bub, and Order Scale. Journal of
Applied Developmental Psychology, 16, 429–444.
Matthews, D. (2020). Learning how to communicate in infancy. Current Perspectives on
Child Language Acquisition: How Children Use Their Environment to Learn.
Amsterdam, The Netherlands: John Benjamins Publishing Company, 11–38.
McFarland, D. H. (2001). Respiratory markers of conversational interaction.
McFarland, D. H., Fortin, A. J., & Polka, L. (2020). Physiological measures of mother–infant
interactional synchrony. Developmental Psychobiology, 62(1), 50–61.
McHugh, M. L. (2012). Interrater reliability: the kappa statistic. Biochemia Medica:
Biochemia Medica, 22(3), 276–282.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 41 -
Miller, A. L., Song, J., Sturza, J., Lumeng, J. C., Rosenblum, K., Kaciroti, N., & Vazquez, D.
M. (2017). Child cortisol moderates the association between family routines and
emotion regulation in low‐income children. Developmental Psychobiology, 59(1), 99–
Mills-Koonce, W. R., Willoughby, M. T., Garrett-Peters, P., Wagner, N., Vernon-Feagans,
L., & Investigators, F. L. P. K. (2016). The interplay among socioeconomic status,
household chaos, and parenting in the prediction of child conduct problems and callous–
unemotional behaviors. Development and Psychopathology, 28(3), 757–771.
Oakes, L. M., Baumgartner, H. A., Barrett, F. S., Messenger, I. M., & Luck, S. J. (2013).
Developmental changes in visual short-term memory in infancy: evidence from eye-
tracking. Frontiers in Psychology, 4. https://doi.org/10.3389/fpsyg.2013.00697
Oller, D. K., Buder, E. H., Ramsdell, H. L., Warlaumont, A. S., Chorna, L., & Bakeman, R.
(2013). Functional flexibility of infant vocalization and the emergence of language.
Proceedings of the National Academy of Sciences, 110(16), 6318–6323.
Oller, D. K., & Griebel, U. (2020). Functionally flexible signaling and the origin of language.
Frontiers in Psychology, 11.
Olson, S. L., & Lunkenheimer, E. S. (2009). Expanding concepts of self-regulation to social
relationships: Transactional processes in the development of early behavioral
Provenzi, L., Borgatti, R., Menozzi, G., & Montirosso, R. (2015a). A dynamic system
analysis of dyadic flexibility and stability across the Face-to-Face Still-Face procedure:
application of the State Space Grid. Infant Behavior and Development, 38, 1–10.
Provenzi, L., Borgatti, R., Menozzi, G., & Montirosso, R. (2015b). A dynamic system
analysis of dyadic flexibility and stability across the Face-to-Face Still-Face procedure:
application of the State Space Grid. Infant Behavior and Development, 38, 1–10.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 42 -
Schieffelin, B. B. (2016). Looking and talking: the functions of gaze direction in the
conversations of a young child and her mother. In Acquiring conversational competence
(pp. 50–65). Taylor and Francis.
Shamama-tus-Sabah, S., Gilani, N., & Wachs, T. D. (2011). Relation of home chaos to
cognitive performance and behavioral adjustment of Pakistani primary school children.
International Journal of Behavioral Development, 35(6), 507–516.
Smith, C. G., Jones, E. J. H., Charman, T., K., Clackson., Mirza, F. U., & Wass., S. v. (2021).
Anxious parents show higher physiological synchrony with their infants. Psychological
Song, J.-H., Miller, A. L., Leung, C. Y. Y., Lumeng, J. C., & Rosenblum, K. L. (2018).
Positive parenting moderates the association between temperament and self-regulation
in low-income toddlers. Journal of Child and Family Studies, 27(7), 2354–2364.
Spitzer, R. L., Kroenke, K., Williams, J. B. W., & Löwe, B. (2006). A brief measure for
assessing generalized anxiety disorder: the GAD-7. Archives of Internal Medicine,
Tronick, E. (2007). The neurobehavioral and social-emotional development of infants and
children. WW Norton & Company.
Vernon-Feagans, L., Willoughby, M., & Garrett-Peters, P. (2016). Predictors of behavioral
regulation in kindergarten: Household chaos, parenting, and early executive functions.
Developmental Psychology, 52(3), 430.
Wass, S. v, Phillips, E., Smith, C., & Goupil, L. (2021). Vocalisations and the Dynamics of
Interpersonal Arousal Coupling in Caregiver-Infant dyads.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 43 -
Wass, S. v, Smith, C. G., Clackson, K., Gibb, C., Eitzenberger, J., & Mirza, F. U. (2019).
Parents mimic and influence their infant’s autonomic state through dynamic affective
state matching. Current Biology, 29(14), 2415-2422. e4.
Weinberg, M. K., & Tronick, E. Z. (1996). Infant affective reactions to the resumption of
maternal interaction after the still‐face. Child Development, 67(3), 905–914.
Wilder, C. N. (1974). Respiratory patterns in infant cry. Developmental Psychology, 6, 293–
Yoo, H., Bowman, D. A., & Oller, D. K. (2018). The origin of protoconversation: An
examination of caregiver responses to cry and speech-like vocalizations. Frontiers in
Psychology, 9, 1510.
Zeskind, P. S. (2013). Infant crying and the synchrony of arousal. Evolution of Emotional
Communication: From Sounds in Nonhuman Mammals to Speech and Music in Man,
Zhang, Y. S., & Ghazanfar, A. A. (2016). Perinatally influenced autonomic system
fluctuations drive infant vocal sequences. Current Biology, 26(10), 1249–1260.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 44 -
Supplementary Materials for:
Needing to shout to be heard? Affective dysregulation, parental under-responsivity, and
a disconnection between vocal signalling and autonomic arousal in infants from chaotic
Table of Contents
1 Supplementary Methods ................................................................................................. 45
1.1 Experimental participant details .......................................................................... 45
1.3 Heart rate data ....................................................................................................... 46
1.4 Heart-Rate Variability (HRV) .................................................................................... 49
1.5 Actigraphy .................................................................................................................... 49
1.6 Arousal composite .................................................................................................. 50
1.7 Removal of autocorrelation from arousal data ......................................................... 51
1.8 Home/Awake coding .............................................................................................. 52
1.8.1 Home/not home ....................................................................................................... 52
1.8.2 Sleeping/waking ...................................................................................................... 52
1.9 Vocal intensity coding ............................................................................................ 52
1.10 Permutation-based clustering analyses ................................................................ 52
2 Supplementary Results .................................................................................................... 54
2.1 Part 2a – repeated with quartile split by household chaos ...................................... 54
2.2 Part 2b – repeated with quartile split by household chaos ...................................... 54
2.3 Part 3 – repeated with quartile split by household chaos ........................................ 55
2.4 Part 3 - Caregiver arousal changes around low- and high-intensity infant
vocalisations ........................................................................................................................ 55
Running head: CHAOS, VOCALISATIONS, AROUSAL - 45 -
1 Supplementary Methods
2 Experimental participant details
This sample size was selected prior to the commencement of the study based on power
calculations presented, and approved by peer review, in the funding application that supported
this work (ESRC ES/N017560/1). Exclusion criteria included: complex medical conditions,
skin allergies, heart conditions, parents below 18 years of age, and parents receiving care from
a mental health organisation or professional. Full demographic details of the participants are
Infant age (days) – mean
Gender (% male)
Infant Ethnicity (%)
Asian, Indian & Pakistani
Mixed - White/Afro-Carib
Mixed - White/Asian
Household Income (%)
No formal qualifications
Running head: CHAOS, VOCALISATIONS, AROUSAL - 46 -
Table S1: Demographic details for: a) the whole sample; b) and c) – data subdivided into
low/high CHAOS groups using a median split, as described in the main text.
1.2 CHAOS questionnaire results
Figure S1: Histogram of CHAOS questionnaire results
2.3 Heart rate data
ECG was recorded at 250Hz. To ensure good quality recordings, the ECG device was attached
using standard Ag-Cl electrodes, placed in a modified lead II position. Due to technical
problems with the ECG recording leads (N=9) and to problems with attaching the ECG
recording electrodes securely (N=2), the ECG data were unavailable for 11 of the 93
participants originally tested.
To ensure the accuracy of these recording devices, they were cross-validated by recording heart
rate and heart rate variability using both the new devices at home and established recording
devices (a Biopac MP150 amp recording at 2000Hz) in lab settings. High reliability was
observed both for heart rate (rho=.57, p<.001) and heart rate variability (rho=.70, p=.01).
Analysis of the Inter-Beat Intervals (IBIs) was performed using custom-built Matlab scripts.
These scripts were designed through an extensive piloting process to be optimal for the ECG
device used for this study. First, data were parsed using a simple amplitude threshold (see e.g.
(Aurobinda, Mohanty, & Mohanty, 2016) for a similar approach), with R peaks identified as
moments where the raw ECG signal exceeded the threshold value. Initially, the threshold value
was set high; the same process was then repeated at incrementally decreasing thresholds.
At each threshold value, the R peaks identified were automatically subjected to the following
checks. These threshold values were set following extensive piloting and visual inspection of
our infant ECG data using the visualisation shown in Figure S2. i) minimum temporal
threshold: does the R peak occur at a time interval of greater than 300 msecs since the previous
R peak (corresponding to a heart rate of 200 BPM); ii) maximum temporal threshold: does the
Running head: CHAOS, VOCALISATIONS, AROUSAL - 47 -
R peak occur at a time interval of less than 850 msecs since the previous R peak (corresponding
to a heart rate of 70 BPM); iii) maximum rate of change: when we calculate the R to R interval
between this peak and the subsequent peak, and compare it with the R to R interval between
this peak and the previous peak, is this difference less than 300msecs? In setting these threshold
values, careful attention was paid to visual inspection to determine the maximum and minimum
‘genuine’ heart rates observed in our infant data; in setting the maximum rate of change
criterion, careful attention was paid to identify the maximum rate of vagally mediated heart
rate changes in infants.
Figure S2 shows a sample screenshot from the Matlab processing algorithm that was used. Two
separate types of artefact are shown. The first, highlighted by the call-out figures at a and d,
are instances where the ECG signal for a particular beat was lower than the threshold, and a
genuine beat was missed. It can be seen that in both instances, the R peaks either side of this
missing beat have been automatically identified, and excluded. These artifacts were identified
based on the maximum temporal threshold criterion in example a and d, and additionally based
on the maximum rate of change criterion in example d. The second, highlighted by the call-out
figures at b and c, are instances where the ECG signal exceeded the amplitude threshold, and
an incorrect R peak was identified. In both instances, the incorrect beat has been identified
based on the minimum temporal threshold criterion, and the R peaks either side of this incorrect
beat have been identified and excluded. Please note also that the sample below has been
selected in order to demonstrate how the program identified the most common artefacts in the
data. Overall, the occurrence of both types of artefact in our data is relatively rare, as is shown
in Figure S3, below.
These three criteria were applied separately to data after it had been parsed at each threshold
value. Following this, at each threshold value, the proportion of candidate R peaks that were
rejected was compared with the proportion of candidate R peaks that passed all three criteria.
The threshold value with the lowest proportion of rejected candidate R peaks was chosen as
the threshold used for that participant.
In addition, and as a further check, a trained coder who was naïve to study hypotheses double
coded a randomly selected subsample of 1000 beats for 20% of the participants, coding them
as genuine or artefactual. Cohen’s kappa was calculated to measure inter-rater reliability
between the manual coding and the automatic coding, based on the best-fitting threshold level.
This was found to be 0.97, which is high (McHugh, 2012).
Running head: CHAOS, VOCALISATIONS, AROUSAL - 48 -
Figure S2: Sample screenshot from ECG parsing algorithm. 60 seconds’ data is shown. From
top to bottom: i) raw ECG signal. Coloured dots show the results of the three checks described
in the main text, below (see legend); ii) smoothed second derivative of ECG signal. This
measure was not used as our pilot analyses found it to be less effective than applying the
Running head: CHAOS, VOCALISATIONS, AROUSAL - 49 -
processing to the raw signal; iii) raw (unprocessed) actigraph data. This information was only
used for visual inspection, and was not used in parsing; iv) RR intervals (in BPM), with rejected
data segments excluded.
Figure S2 below shows a histogram of the proportion of candidate R peaks rejected for each
participant, based on the best-fitting threshold value. The median (st. err.) is 1.07 (0.36) % data
rejected. This relatively low figure was achieved through very close attention during the
piloting phase to the selection and placement of the ECG electrodes, to the design of the device,
and the gain settings on the recording device.
Figure S3: Histogram showing the proportion of rejected R peaks (as identified using the three
criteria described above).
1.4 Heart-Rate Variability (HRV)
HRV was calculated using the PhysioNet Cardiovascular Signal Toolbox (Vest et al., 2018).
In these scripts, which performed a completely separate analysis of the ECG data, a 60-second
window with an increment of 60 seconds was implemented, and the default settings were used
with the exception that the min/max inter-beat interval was set at 300/750 ms for the infant data
and 300/1300 ms for the adult data. The Root Mean Square of Successive Differences
(RMSSD) measure was taken to index Heart Rate Variability, but other frequency domain
measures were additionally inspected and showed highly similar results, as expected (Vest et
Actigraphy was recorded at 30Hz. To parse the actigraphy data we first manually inspected the
data, then corrected artifacts specific to the recording device used, then applied a Butterworth
Running head: CHAOS, VOCALISATIONS, AROUSAL - 50 -
low-pass filter with a cut-off of 0.1 Hz to remove high-frequency noise, and then averaged
from three dimensions into one. Actigraphy data were available for all participants tested.
2.6 Arousal composite
Previous research has shown significant patterns of tonic and phasic covariation between
different autonomic measures collected from infants (Wass, Clackson, & de Barbaro, 2016;
Wass, de Barbaro, & Clackson, 2015). Here, we include plots showing that the present dataset
replicated and extended these results. The plots only show the sections of the data when
participants were at home, comparing sections in which the infants were awake and asleep.
Figure S2a shows cross-correlation plots examining the relationship between heart rate and
movement. In both waking and sleeping sections the zero-lag correlation is 0.5. Figure S2c
shows how these zero-lagged correlations vary on a per-participant basis. S2b shows an
illustrative sample from a single participant. Sleeping sections show very low movement levels
and lower heart rate. Of note, heart rate and movement do still inter-relate during the sleeping
sections of the data (Figure S2c), albeit that the variability in heart rate and movement is lower.
Figure S2 d)-f) show similar relationships between heart rate and heart rate variability,
illustrating the strong and consistent negative relationships that were observed between these
variables, as predicted.
Figure S4: Illustrating the relationship between the individual physiological measures
included in the composite measure. a) Cross-correlation of the relationship between HR and
Movement. b) Scatterplot from a sample participant. Each datapoint represents an individual
60-second epoch of data. c) Histograms showing the average zero-lagged correlation between
60-second epochs, calculated on a per-participant basis and then averaged. d)-f) Equivalent
plots for Heart rate and Heart rate variability.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 51 -
Extensive previous research has identified fractionation, and differentiation, within our
autonomic response systems (Janig & Habler, 2000; Kreibig, 2010; Lacey, 1967; Levenson,
2014; Quas et al., 2014) – suggesting, for example that the sympathetic and parasympathetic
subdivisions operate, to an extent, in a non-additive manner (Samuels & Szabadi, 2008).
Although indubitable, these findings should be seen as rendering incorrect our treatment here
of autonomic arousal as a one-dimensional construct. Like many other arguments concerned
with general versus specific factors, the question is rather one of the relative proportions of
variance that can be accounted for by a single common factor in comparison with the variance
accounted for by the sum of specific factors (Graham & Jackson, 1970) (see also (Calderon,
Kilinc, Maritan, Banavar, & Pfaff, 2016)).
As a result of these considerations, the three autonomic measures were collapsed into a single
composite measure for Analysis 1. To do this, the actigraphy data was first subjected to a log
transform (Thomas & Burr, 2008), to correct the raw results, which showed a strong positive
skew (Wass et al., 2016; Wass et al., 2015) (see also SM section 1.6, below). Second, all three
variables were converted to z-scores. Third, the HRV data were inversed because of the overall
negative relationships noted between HRV and the other two measures (see Figure S4). Fourth,
the three z-scores were averaged.
On the occasions where heart rate data were excluded due to artifact, data from actigraphy
alone was used for the composite variable. Note that these occasions were relatively rare
(accounting for a median ~=1% of all data - see Figure S3), and that the zero-lag cross-
correlation between movement and heart rate across all available data was high (~=.5 – see
1.7 Removal of autocorrelation from arousal data
Autonomic arousal data are known to show autocorrelation (Wass et al., 2016). In order to
preclude the possibility that differences in the autocorrelation may have influenced results, the
autocorrelation was removed from the data prior to performing all calculations, using the
following procedure. First, best-fit bivariate polynomials were calculated for the two time
series independently, in order to remove linear and quadratic trends, and the residuals obtained
were subjected to the Dickey-Fuller test to check that they showed stationarity, which they did.
The residuals were used in subsequent analyses. Next, in order to remove the autocorrelation
component from each time series independently, univariate autoregressive models were fitted
to each time series, and the residuals were calculated (see e.g. Feldman, Greenbaum, &
Yirmiya, 1999; Feldman, Magori-Cohen, Galili, Singer, & Louzoun, 2011; Jaffe et al., 2001;
Suveg et al., 2016 for similar approaches). The residual values (shown in Figure 1) were
converted into z-scored values. These z-scored values were then used for all analyses. The only
exception to this is the analyses specifically examining changes in autocorrelation relative to
vocalisations, for which the raw uncorrected data were used.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 52 -
2.8 Home/Awake coding
1.8.1 Home/not home
Coding of when participants were at home was performed using the GPS monitors built into
the recording devices. The position of the participant’s home was calculated based on the
postcode data that they supplied, and any GPS samples within a 50m area of that location were
treated as Home (corresponding to the accuracy of the GPS devices that we were using).
To identify samples in which infants were sleeping, parents were asked to fill in a logbook
identifying the times of infants’ naps during the day. This information was manually verified
by visually examining the actigraphy and ECG data collected, on a participant by participant
basis. Actigraphy, in particular, shows marked differences between sleeping and waking
samples (see Figure 1 in main text), which allowed us to verify the parental reports with a high
degree of accuracy. N=4 of the participants recorded did not sleep during the day that we were
2.9 Vocal intensity coding
The same analysis as described for the vocal affect coding in the manuscript was conducted to
code intensity on a scale from 1 (least intense) to 9 (most intense), and the same criteria were
used to collapse these into low intensity/neutral/high intensity. In order to assess inter-rater
reliability, 24% of the sample was double coded; Cohen’s kappa was 0.60, which is considered
acceptable (McHugh, 2012). All coders were blind to intended analyses.
3 Permutation-based clustering analyses
To estimate the significance of the time-series relationships in the results, a permutation-based
temporal clustering approach was used. This method examines temporally contiguous patterns
of change in instances where the centre-point of the expected response window is unknown, or
unimportant (Maris & Oostenveld, 2007). In each case, the test statistic (always specified in
the text) was calculated independently for each time window. Series of significant effects
across contiguous time windows were identified using an alpha level of .05. 1000 random
datasets were then generated with the same dimensions as the original input data. To ensure
that the same level of autocorrelation was present in the simulated data as in the original
datasets, multivariate autoregressive models were fitted to each sample included in the original
dataset using the Matlab function ARfit.m (Neumaier & Schneider, 2001), and the matching
AR parameters were used to generate each of the random datasets using the Matlab function
ARsim.m (Neumaier & Schneider, 2001). Then, the same sequence of analyses was repeated,
and the longest series of significant effects across contiguous time windows was identified.
The results obtained from the random datasets were used to generate a histogram, and the
likelihood of observed results have been obtained by chance was calculated by comparing the
observed values with the randomly generated values using a standard bootstrapping procedure.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 53 -
Thus, a p value of <.01 indicates that an equivalent pattern of temporally contiguous group
differences was observed in 10 or fewer of the 1000 simulated datasets created.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 54 -
4 Supplementary Results
2.1 Part 2a – repeated with quartile split by household chaos
Figure S5: Identical to Figure 3, but based on a quartile split by household chaos.
2.2 Part 2b – repeated with quartile split by household chaos
Figure S6: Identical to Figure 4, but based on a quartile split by household chaos.
Running head: CHAOS, VOCALISATIONS, AROUSAL - 55 -
2.3 Part 3 – repeated with quartile split by household chaos
Figure S7: Identical to Figure 6, but based on a quartile split by household chaos.
2.4 Part 3 - Caregiver arousal changes around low- and high-intensity infant
Figure S8: Identical to Figure 6 but examining caregiver arousal changes to infant
vocalisations subdivided by vocal intensity. a) and b) show results broken down using a
median split by household chaos. c) and d) show the same analysis, but subdivided using a
quartile split by household chaos.