Developmental features of sleep electrophysiology in family dogs

Abstract

Age-related differences in dog sleep and the age at which dogs reach adulthood as indexed by sleep electrophysiology are unknown. We assessed, in (1) a Juvenile sample (n = 60) of 2–14-month-old dogs (weight range: 4–68 kg), associations between age, sleep macrostructure, and non-rapid eye movement (NREM) EEG power spectrum, whether weight moderates associations, and (2) an extended sample (n = 91) of 2–30-months-old dogs, when sleep parameters stabilise. In Juvenile dogs, age was positively associated with time in drowsiness between 2 and 8 months, and negatively with time in rapid eye movement (REM) sleep between 2 and 6 months. Age was negatively associated with delta and positively with theta and alpha power activity, between 8 and 14 months. Older dogs exhibited greater sigma and beta power activity. Larger, > 8-month-old dogs had less delta and more alpha and beta activity. In extended sample, descriptive data suggest age-related power spectrum differences do not stabilise by 14 months. Drowsiness, REM, and delta power findings are consistent with prior results. Sleep electrophysiology is a promising index of dog neurodevelopment; some parameters stabilise in adolescence and some later than one year. Determination of the effect of weight and timing of power spectrum stabilisation needs further inquiry. The dog central nervous system is not fully mature by 12 months of age.

Full text

1
Vol.:(0123456789)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports
Developmental features of sleep
electrophysiology in family dogs
Vivien Reicher1,2,5*, Nóra Bunford1,3,5, Anna Kis1,4, Cecília Carreiro1, Barbara Csibra1,
Lorraine Kratz1 & Márta Gácsi1,2
Age-related dierences in dog sleep and the age at which dogs reach adulthood as indexed by
sleep electrophysiology are unknown. We assessed, in (1) a Juvenile sample (n = 60) of 2–14-month-
old dogs (weight range: 4–68 kg), associations between age, sleep macrostructure, and non-rapid
eye movement (NREM) EEG power spectrum, whether weight moderates associations, and (2) an
extended sample (n = 91) of 2–30-months-old dogs, when sleep parameters stabilise. In Juvenile dogs,
age was positively associated with time in drowsiness between 2 and 8 months, and negatively with
time in rapid eye movement (REM) sleep between 2 and 6 months. Age was negatively associated
with delta and positively with theta and alpha power activity, between 8 and 14 months. Older dogs
exhibited greater sigma and beta power activity. Larger, > 8-month-old dogs had less delta and more
alpha and beta activity. In extended sample, descriptive data suggest age-related power spectrum
dierences do not stabilise by 14 months. Drowsiness, REM, and delta power ndings are consistent
with prior results. Sleep electrophysiology is a promising index of dog neurodevelopment; some
parameters stabilise in adolescence and some later than one year. Determination of the eect of
weight and timing of power spectrum stabilisation needs further inquiry. The dog central nervous
system is not fully mature by 12 months of age.
Sleep is a relatively immobile, reversible state, characterised by absent or limited responsiveness to the exter-
nal environment (without loss of consciousness) and is regulated by circadian and homeostatic processes1. It
is a complex combination of behavioural and physiological processes that aects several aspects of cognitive
development2 as well as dimensions of functioning3. Findings across studies indicate that sleep plays a role in
animal (neuro)development (for review, see2,4).
Sleep is characterised by the ultradian rhythms of two distinct forms of electroencephalographic (EEG)
activity: desynchronised, rapid waveforms with a small amplitude constitute rapid eye movement (REM) sleep,
and synchronised, slow waveforms with high amplitude constitute non-REM (NREM) sleep. In humans, the
rst sleep stage following wakefulness is NREM 1, during which alpha waves (8–12Hz) characteristic of quiet
wakefulness cease and low-voltage theta activity (4–8Hz) appears. e second stage, NREM 2, is characterised
by theta and delta (1–4Hz) activity and/or sleep spindles (sigma activity bursts, 12–16Hz). In the third stage,
also called slow wave sleep (SWS), high-voltage delta frequency waveforms are dominant. NREM is followed by
REM, characterised by mixed theta and beta activity, rapid saccadic eye movements and muscle atonia5.
is description of basic sleep architecture is applicable to humans and to other mammals, though there are
some inter-species dierences. e transition from wakefulness to sleep, for example, is not as clear in carnivore
and insectivore mammals6,7, as in humans5. ese species exhibit drowsiness, a sleep stage that bears character-
istics of both human NREM 1 sleep and quiet wakefulness. Dierent levels of NREM sleep are also not as dier-
entiated in other species as in humans, but constitute one universal stage, slow wave sleep8 (though more recent
ndings indicate this might be a methodological issue9). Yet another dierence is related to NREM/REM cycles.
In most cases in humans, REM is followed by NREM sleep, whereas other mammals are more likely to wake
up aer REM sleep, potentially reecting an evolutionary adaptation to limit time spent in a vulnerable state10.
Developmental changes as reected by neural indices of sleep are not specic to humans (for review see2) but
are commonly observed across many mammalian species11. For example, in several species, there is longer total
sleep duration during early, relative to later developmental stages (human12; rat, cat, guinea pig13). Furthermore,
in the rst month of life, a decrease in REM sleep and an increase in NREM sleep has been observed in the cat,
OPEN
1Department of Ethology, Institute of Biology, Eötvös Loránd University, Pázmány Péter sétány 1/C,
Budapest 1117, Hungary. 2MTA-ELTE Comparative Ethology Research Group, Budapest, Hungary. 3Developmental
and Translational Neuroscience Research Group, Institute of Cognitive Neuroscience and Psychology, Research
Centre for Natural Sciences, Budapest, Hungary. 4Institute of Cognitive Neuroscience and Psychology, Research
Centre for Natural Sciences, Budapest, Hungary. 5
These authors contributed equally: Vivien Reicher and Nóra
Bunford. *email: vivien.reicher@gmail.com
Content courtesy of Springer Nature, terms of use apply. Rights reserved

2
Vol:.(1234567890)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
rat and guinea pig13,14. It has been proposed that REM sleep may serve as an endogenous source of activation
that supports neurodevelopment, e.g., in the form of muscle twitches during sleep15 and this may explain why
there is an age-related decrease in this sleep parameter. As aging proceeds, changes in sleep architecture are less
pronounced during pubertal development: for example, in rats, there is a decrease in both REM and NREM sleep
accompanied by an increase in wakefulness14. To our knowledge, developmental changes in sleep parameters
have not been documented in non-human primates, except for cross-sectional comparisons of young adult and
aged animals, e.g., in rhesus monkeys16 or in mouse lemurs17.
Developmental changes are not only reected in age-related alterations in sleep macrostructure but also in
EEG spectrum. Given their assumed role in cortical maturation, the majority of available, pertinent develop-
mental studies are focused on delta power (1–4Hz) and/or slow wave activity (SWA; 1–4.5Hz) changes during
NREM sleep (e.g.14,18). Both rat and human data show that the activity of slow waves increases during the rst
years of life, reaches a maximum by puberty, and declines throughout adolescence (rat14; human19,20). ese
changes likely reect synaptic pruning18,21. A similar pattern has been found in dogs, using invasive methods22.
Taken together, neurodevelopmental changes mirrored by dierences in distinct aspects of sleep are observ-
able across some mammalian species. Yet, not just with regard to development but more generally, it is a notable
limitation to the literature that available ndings on animal sleep have been mostly obtained with methods that
limit their comparability across humans and nonhuman animals. Specically, in available animal studies, invasive
manipulations and techniques, such as drugs, lesions and/or sensory deprivations have been used (see review4).
Given dierences between chemically-induced and natural sleep23 and that intracranial electrodes can only be
used with a restricted subgroup of animals (i.e., laboratory-bred and -kept animals), the generalizability of these
results is limited, underscoring need for animal models with which sleep can be studied naturalistically, using
non-invasive methods.
Relatedly, the family dog has been recognised as unique among domesticated species, because both its evolu-
tionary history and shared social environment with humans have contributed to changes in dog socio-cognitive
skills24,25. Regarding sleep research specically, dogs have been repeatedly shown to be an ideal model species
for non-invasive sleep EEG studies7,26–30. e general architecture of human sleep is better approximated by dog
sleep than by that of commonly used laboratory animals. For example, the primary diurnal sleep phase in humans
and dogs is in the dark, whereas that of mice and rats is in the light; the daily sleep duration in humans is around
8h, in dogs it is 8–14h, while in mice, rats, and cats it is longer, around 12–15 h31. Although early studies on dog
sleep were focused on neurological conditions (e.g., epilepsy) and involved invasive methods with laboratory
animals32,33, more recent investigations capitalised upon the advantages of this species, by using a non-invasive
polysomnography method7. is novel method has been successfully implemented in several studies assessing
the basic architecture of dog sleep (i.e., in the absence of pre-sleep active experiences and/or handling)7,34,35 and
the correspondence between awake functioning, sleep architecture and pre-sleep activities including learning27
and emotional exposure28 as well as timing and location of sleep26. Accordingly, the dog is a promising model for
comparative sleep research and ndings indicating age-related changes and dierences in dog behaviour indicate
it may also be suitable to model developmental dierences reected in neural indices of sleep.
Available data suggest developmental dierences in dogs’ socio-cognitive abilities, this concerns for example
age-related dierences in how well dogs follow human pointing gestures, with lower performance in 10–12-
month olds relative to younger (2–10) and older (12–14) dogs36. Reports on similar abilities in puppies and
adult dogs can also be found, for example, social referencing seems to develop early (puppies37,38; adults39,40).
Executive functions, such as working memory, were observed in puppies of 8–10weeks old41, but less developed
than in dogs above 1year old42.
Morphological and structural brain changes in dogs have also been described but only with regard to very
young animals (8–36weeks)43 or in old dogs (> 7years)44. Beyond these studies, the majority of the literature
examining age-related changes in dog behaviour (e.g.45), cognition (e.g.46), brain anatomy (e.g.,47) or neural
processes (e.g.,7), consists of studies with ‘adult’, i.e., 12-month-old or older animals or declare that dogs older
than 12months of age are adults (e.g.48). Yet others focused on changes throughout the lifespan but not explic-
itly on early development (despite having the data to do so), with such focus potentially obscuring age-related
dierences in very young animals (e.g.49). Despite notable absence of empirical data on when brain maturation
in dogs may mark the onset of/transition into adulthood as well as evidence that dogs from any size category
reach their nal weight by 12 months50, it is a generally accepted (unspoken) assumption in the literature that
dogs over one year of age are adults.
In addition to age, dierences in morphological features across dogs and dierences in developmental rates
between breeds may inuence EEG data. Regarding the former, dogs present signicant intraspecic variability
with regard to head musculature and skull shape and thickness. To circumvent measurement error that might
arise as a result of these dierences, dog EEG data are analysed using not absolute, but relative EEG power7, as in
human studies51. Certain breeds may mature at a faster rate than others. For example, dogs with greater muscle
and/or skull thickness are suggested to develop at a slower rate and reach sexual maturity later, given hormonal52
and sexual53,54 changes. As the muscle and skull thickness of a dog is typically associated with their adult weight,
dierences in developmental status might be indexed as a dog’s adult weight. For example, when comparing
beagles and great Danes, basal plasma growth hormone (GH) increases in both breeds until the age of 7weeks,
but high GH release persisted only in great Danes until the age of 24 weeks52. Further, in fox terriers, a breed
with an average adult weight of 8kg, the rst spermatozoa in the ejaculate occurs when the animals are between
8 and 9months old53 whereas in Collies, a breed with an average adult weight of 23kg, the rst spermatozoa are
observed when the animals are between 11 and 12months old54.
Taken together, evidence in both humans and non-human animals indicates that developmental dierences
and, as such, developmental status, may be reected in parameters of sleep. However, the majority of available
ndings have been obtained using invasive methods and performed mainly on what are assumed to be adult
Content courtesy of Springer Nature, terms of use apply. Rights reserved

3
Vol.:(0123456789)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
animals. us, developmental dierences in specic aspects of animal sleep or when such dierences stabilise
(and thus indicate maturity) are unknown.
For the present study, we assumed that age-related dierences in the sleep architecture of dogs are generally
similar to corresponding dierences in other mammals. To test this, we analysed age-related dierences in sleep
macrostructure (drowsiness, NREM and REM) and spectral power (delta 1–4Hz, theta 4–8Hz, alpha 8–12Hz,
sigma 12–16Hz, beta 16–30Hz) variables. Weight was also included in our analyses.
Methods
Subjects. Altogether, data from 91, 2–30-month-old family dogs were analysed. First, to determine whether
sleep parameters stabilise by the time dogs reach the age of 1year, data were collected from 60, 2–14-month-old
dogs (Juvenile sample) (Mean age = 7.8, SD = 4.0). is consisted of 50 purebreds from 25 dierent breeds and
10 mongrels; 32 females (Mean age = 8, SD = 4.0; of the 32 females, 15 were neutered and for 1, there was no
such information available) and 28 males (Mean age = 7.7, SD = 4.2; out of the 28 male dogs 3 were neutered).
Second, to assess whether sleep parameters change beyond the age of 14months, we extended our sample (and
thus, the age range thereof) by adding data from our previous dog sleep EEG studies7,26–29, so that the total sam-
ple consisted of 91 family dogs (Extended sample encompassing the Juvenile sample + n = 31, > 14month-old
dogs) with an age range of 2–30months. is sample consisted of 75 purebreds from 30 dierent breeds and 16
mongrels; 48 females (Mean age = 12.9, SD = 8.3; of the 48 females, 26 were neutered, for 3, there was no such
information available, and the remaining 19 were not in heat at the time of the sleep recording) and 43 males
(Mean age = 12.8, SD = 8.1; of the 43 males, 10 were neutered and for 1, there was no such information available).
In case of purebred dogs, actual weight was not used to index weight, as actual weight can be more of an
indicator of the animals’ body condition rather than of size-driven maturational speed. Rather, we assigned to
each dog the standard adult, breed-specic weight based on the Federation Cynologique Internationale/American
Kennel Club database (http:// www. fci. be/ en/ nomen clatu re/; https:// www. akc. org/). In case of mixed breeds, we
used the dog’s owner-reported 12-month-old weight (assessed when the dog reached 12months of age). For
details on sample demographics, see Supplementary TableS1.
Ethics statement. is research was approved by the Hungarian “Animal Experiments Scientic and Ethi-
cal Committee” (PE/EA/853–2/2016) and was conducted in compliance with ARRIVE Guidelines and with
Hungarian regulations on animal experimentation and Guidelines for use of animals in research, as outlined
by the Association for the Study Animal Behaviour (ASAB). All owners participated voluntarily and signed an
informed consent form.
Procedure. All sleep recordings had a minimum duration of 0.5h and a maximum duration of 3h and were
conducted aer a relatively active day (e.g., mentally and physically loaded due to advanced training, excursion),
during the aernoon, with a start time between 12 and 6pm. Dogs were measured at various unfamiliar loca-
tions (i.e., camp, canine sleep laboratory, unfamiliar room at breeders’, owners’ and friends’ homes), all appropri-
ate for basic sleep recording: constantly dark, quiet environment, mattress and reading lamp for the owner and
water for the dog, available adlibitum.
Before measurements, the experimenter explained the process to the owner while the dog could explore the
room (5–10min). Prior to sleep, in case of the Juvenile sample only, owners were asked to have their dog execute
a few familiar, simple tasks (e.g., lay down, give paw, sit) and praise them with treats/verbally, so as to ease any
excitement the dog may have experienced in the novel environment.
Aer such familiarization, the owner and dog settled on the mattress and the owner gently held the dog’s
head while the experimenter applied the surface electrodes. During electrode placement, dogs were rewarded
using food and/or social (e.g., petting, praise) reward.
In cases where a dog did not appear calm and/or comfortable with electrode placement, the experimenter
added brief breaks (~ 5min) before continuing with electrode placement. If more than 60min passed since arrival
time and electrode placement was unsuccessful, the owner was asked to come back with the dog on another
occasion. e rst attempt to measure sleep EEG was unsuccessful in the case of 17 dogs, two of whom were not
invited back for a second attempt due to aggressive behaviour, such as biting. Of those invited back for a second
attempt (n = 15), the majority were measured successfully (n = 8). If the second attempt was also not successful
for measure, the dog was excluded from the study.
Aer electrode placement and checking of appropriate quality of polysomnography (PSG) signals, owners
were asked to mute their cell phones and, for the duration of the measurement, engage in a quiet activity such
as reading, watching a movie on a laptop with earphones or sleeping. e experimenter le the room and moni-
tored the measurement in an adjacent room. If the displacement or malfunction of an electrode occurred, the
experimenter entered the sleep lab to x it.
EEG/PSG method. Dogs were measured between 2012 and 2020. During this period, our lab improved and
updated its electrode placement and recording methods. Specically, in our prior studies, only the frontal (Fz)
channel was recorded7,26–28. Since then, the electrode placement has been updated and instead of one, four EEG
channels were recorded29. Further, in the Juvenile sample, four EEG channels and an eye movement channel had
been recorded (and not ECG, EMG and respiration as in case of the > 14-month-old dogs).
e two electrodes placed on the right and le zygomatic arch next to the eyes (F8, F7) and the scalp elec-
trodes over the anteroposterior midline of the skull (Fz, Cz) were referred to the G2, a reference electrode which
was in the posterior midline of the skull (occiput; external occipital protuberance). e ground electrode (G1)
Content courtesy of Springer Nature, terms of use apply. Rights reserved

4
Vol:.(1234567890)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
was attached to the le musculus temporalis. See Fig.1 for photo of a dog with all electrodes applied and Sup-
plementary Fig.S1 for a schematic drawing of electrode positions on a dog’s head.
As at least the frontal electrode (Fz) was active in all dogs (i.e., in some cases of the > 14-month-old dogs),
data from this electrode were used for spectral analyses. For electrode placement, Signa Spray Electrode Solution
was used to separate the dog’s fur where Gold-coated Ag/AgCl electrodes were attached onto the skin using EC2
Grass Electrode Cream (Grass Technologies, USA) and impedance values were kept below 20kΩ. As further
visualisation of these electrode placement procedures, a representative video is available on our youtube channel
(see QR code on Fig.1).
Recordings were obtained with one of the following three technical arrangements.
(1) In case of 47 dogs (51.6% of the total sample) the signal was collected, amplied and digitised at a sampling
rate of 1000Hz/channel, using the 40-channels NuAmps amplier (© 2018 Compumedics Neuroscan)
and DC-recording, and saved in .cnt format with the Scan 4.3 Acquire soware (© 2018 Compumedics
Neuroscan), converted to .edf format using MatLab EEG Toolbox.
(2) In case of 37 dogs (40.7% of the total sample) the signal was collected, pre-ltered, amplied and digitised
with a sampling rate of 1024Hz/channel using a SAM 25 R style MicroMed Headbox (MicroMed Inc.,
Houston, TX, USA). e hardware passband was set at 0.5–256Hz, sampling rate of 512Hz, anti-aliasing
lter with cut-o frequency at 1kHz, and 12-bit resolution covering a voltage range of ± 2mV as well as
second-order soware lters (high pass > 0.016Hz, low pass < 70Hz) using System Plus Evolution soware
(MicroMed Inc, Houston, TX, USA), which exported data in .edf format.
(3) In case of 7 dogs (7.7% of the total sample) the signal was collected, pre-ltered, amplied and digitised with
a sampling rate of 249Hz/channel using a 30-channel Flat Style SLEEP La Mont Headbox with implemented
second-order lters (high pass > 0.5Hz, low pass < 70Hz), and HBX32-SLP 32 channel pre-amplier (La
Mont Medical Inc., USA).
To correct for dierences in EEG lter characteristics across recording devices, a calibration process was
implemented on devices (1) and (2). Specically, a waveform generator at the Fz electrode input of both devices
was used to apply 40 and 355μV amplitude sinusoid signals at various amplitudes (0.05Hz, every 0.1Hz between
0.1 and 2Hz, every 1Hz between 2 and 20Hz, every 10Hz between 10 and 100Hz). e amplitude reduction
rate for each recording system was determined by calculating the proportion of digital (measured) and analog
(generated) amplitudes of sinusoid signals. Next, amplitude reduction rates were calculated for each device and
EEG spectrum amplitudes were corrected by dividing such calculated values by the obtained amplitude reduc-
tion rate for the recording system. e calibration process could not be implemented on device (3), as it stopped
working before the end of the project. us, data recorded with that device (n = 7 dogs) were only included in
sleep macrostructure analyses, but not in EEG spectrum analyses.
Data analysis. Both sleep macrostructure and spectral data were analysed. Sleep macrostructure vari-
ables were examined in 91 dogs (Mean age = 12.9, SD = 8.2), and spectral variables in 79 dogs (Mean age = 12.1,
SD = 7.9) (this discrepancy is due to artifacts on Fz (n = 6) and the inability to calibrate device (3) (n = 7).
Sleep recordings were visually scored in accordance with standard criteria55, adapted for dogs7 and previ-
ously shown to reliably identify stages of wake, drowsiness, NREM and REM in dogs7,56. Data was analysed and
exported using Fercio’s EEG Plus 2009–2020 soware (developed by Ferenc Gombos). Sleep data consisted of
0.5–3-h-long recordings, with variability in recording length being due to dierences in subject compliance.
Juvenile dogs were more likely to fall asleep early (see “Results”), yet, if and when they woke up, they oen
became active and in such cases, recording had to be discontinued regardless of the length of elapsed time.
Conversely, if and when > 14-month-old dogs woke up, they continued lying next to their owner, relaxed, and/
or fell back asleep, and thus recording could be continued for the duration of 3h. FiguresS2 and Figs.S3 (see
“Supplementary”) show that age was associated with time spent awake (with older dogs spending more time
Figure1. Photo of a dog with electrode placement before the sleepmeasurement.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

5
Vol.:(0123456789)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
awake) but not with time spent asleep (minutes of drowsiness + NREM + REM sleep during the recording) (see
“Results”). Given these inherent age-related dierences between dogs (resulting in procedural, i.e., recording
duration dierences), to control for any potential biases, relative values were used in all sleep macrostructure
variables (time spent in drowsiness, NREM and REM). Further, as age was associated with time spent awake,
wake-related variables (e.g. sleep latency 1: elapsed minutes from the onset of recording until the rst epoch
scored as drowsiness; sleep latency 2: elapsed minutes from the onset of recording until the rst epoch scored
as NREM sleep) were excluded from analyses.
Relative power spectra were calculated only for NREM sleep, which provided most artifact-free traces and
only for the Fz channel that was uniformly recorded for all dogs. Subjects with artifacts throughout the whole
recording on this channel (n = 3 dogs) were excluded. For the remaining subjects (n = 57) artifact rejection of the
EEG trace was carried out manually in 4s epochs. Average power spectral values (1–30Hz) were calculated by a
mixed-radix Fast Fourier Transformation (FFT) algorithm, applied to the 50% overlapping, Hanning-tapered 4s
windows of the EEG signals of the Fz-G2 derivation. e relative power spectra were calculated as the propor-
tion of total power (1–30Hz), and frequency ranges of delta (1–4Hz), theta (4–8Hz), alpha (8–12Hz), sigma
(12–16Hz) and beta (16–30Hz) were used.
Analytic plan. Data for 2–14-month-old dogs (n = 60) were statistically analysed and the data of older dogs
(n = 31; 2–30months old) were included in Figures to illustrate trends in the data across a broader age range.
Age-related dierences in sleep parameters, including sleep latency, wake and sleep duration were analysed
with Kendall rank correlation (because age, sleep latency and wake were non-normally distributed).
We used general additive models (GAMs)57 to assess possibly non-linear eects of age on sleep parameters. As
an extension of the generalized linear (regression) model, in addition to linear relationships between predictors
and dependent variables, non-linear terms can also be included in GAMs. As such, GAMs are more exible than
simple linear models, which makes them especially useful in analysis of developmental data58,59. A predened
non-linear function does not have to be specied, as the nonlinear function (i.e., smooth) is determined auto-
matically, in any number of dimensions. e estimated degrees of freedom (edfs) provide an estimation of the
complexity of the smooth, with greater edf values indicating greater complexity (e.g., edf = 1 indicates a linear
association). In our model, the appropriate degree of smoothness was determined based on Restricted Maxi-
mum Likelihood (REML) to prevent overtting. As model validation and visualization are further informative
given that model values are always approximate, we also inspected (visually) our data for appropriateness of the
automatically applied smooth and applied a smoothing parameter when warranted (e.g., in cases of overtting).
GAM analyses were performed using the mgcv R package (version 3.6.360) and results were plotted using the
ggplot R package (version 3.6.3). GAMs were conducted to assess the eect of age, weight, and the interaction
between age and weight as the independent variables (IV) (entered into GAMs as covariates) on macrostructural
and spectral sleep parameters as dependent variables (DV), i.e., drowsiness, NREM and REM sleep as well as
delta and theta power activity. Prior to analyses, assumptions of GAM (i.e., normality distribution of residuals)
were checked. Except for NREM sleep, which was normally distributed, the DVs were positively skewed. As a
result, distribution of model residuals was also non-normal. us, GAMs were tted using the Gamma family
(as recommended for variables with no possibility for negative values) in case of drowsiness and delta and theta
power activity and Tweedie family (as recommended for variables with the possibility for zero values, but not
negative values) in case of REM. Consequently, distributions of residuals were normal.
In case of covariates, collinearity and concurvity were checked. In GAMs, when there is no collinearity
between covariates, an additional pitfall is if two covariates exhibit concurvity (one may be a smooth curve
of another, i.e., there may be a non-linear interaction between them). Concurvity would not only increase the
variance of coecients but also enlarge their standard deviation. Age and weight showed neither collinearity
(p = 0.93, Adj. R2 = − 0.017) nor concurvity (indicators of concurvity range from zero to one, and a value > 0.5
suggests concurvity; in our data the highest value was 0.38).
Results of statistical analyses are presented rst (Juvenile sample [2–14months]) and gures to illustrate
trends in the data across a broader age range are presented second (Extended sample (2–30months)).
Results
In the Extended sample older age was associated with longer sleep latency (τb = 0.272, p < 0.001) and more wake
duration (τb = 0.381, p < 0.001) but not with sleep duration (τb = 0.057, p = 0.422), (see Supplementary Figs.S2
and S3). is, combined with the resulting procedural dierences (early termination of the recordings when
juvenile dogs woke up and became active, see “Methods”), led us to exclude sleep latency and wake-related sleep
parameters from all further analyses. Further, examples of representative hypnograms for dogs with dierent
ages are included in our Supplementary document (Fig.S4).
Juvenile sample (2–14 months). Drowsiness, NREM and REM sleep (n = 60). ere was a positive as-
sociation between age and relative time spent in drowsiness (F = 9.517, p < 0.001; Fig.2a). Specically, visual in-
spection indicated there was a strong positive association between these variables between the ages of 2months
to roughly 8months and that aer the age of 8months, time spent in drowsiness sleep seemed to stabilise,
though its variance increased. Weight and the interaction of age and weight had no eect (all ps > 0.05).
Age was not associated with relative time spent in NREM (F = 1.598, p = 0.192; Fig.2b) and weight and the
interaction between age and weight also had no eect (all ps > 0.05).
Age was negatively associated with relative time spent in REM (F = 4.363, p = 0.014; Fig.2c). Visual inspec-
tion indicated that there was a negative association between these variables from the age of 2months to roughly
6months and that aer the age of 6months, time spent in REM seemed to stabilise. At the ages of 2–3months,
Content courtesy of Springer Nature, terms of use apply. Rights reserved

6
Vol:.(1234567890)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
REM variance was large (Mean = 27.6%, SD = 16.0%). Individual dierences in sleep location or weight were
highly unlikely to cause this eect, as two of the ve dogs that spent the most time in REM slept at the laboratory,
and three of the ve dogs had a standard adult weight of 35kg (the other two had a standard adult weight of 9.5
and 23kg) (for details see Supplementary TableS1, dogs are marked with *). Weight and the interaction of age
and weight had no eect (all ps > 0.05). See Table1 for a summary of macrostructural GAM results.
Spectral analysis (in NREM sleep) (n = 57). Age was negatively associated with delta power activity (F = 3.337,
p = 0.046; Fig.3a). Specically, visual inspection indicated that before the age of 8months no dierences in delta
activity were observable but between the ages of 8 and 14months, at greater ages, there were lower delta power
values. Moreover, there was an interaction between age and weight (F = 2.080, p = 0.049, Fig. 4), with visual
inspection suggesting that aer the age of 8months, larger dogs had less delta activity (weight range: 4–68kg).
Weight had no main eect (F = 0.103, p = 0.804).
Of note, in the current sample, there were n = 4 extra large dogs (with weights of 42.5, 55, 60.8, and 68kg), and
these were not proportionally distributed in terms of their age. To check whether the interaction eect of age and
weight on delta is driven by this bias, analyses were repeated with these four dogs excluded. As previously, age
was negatively associated with delta power activity (F = 7.284, p < 0.001), but the interaction eect between age
and weight was only present at a trend level (F = 1.489, p = 0.053). Weight had no main eect (F = 1.489, p = 0.324).
Age was positively associated with theta power activity (F = 3.361, p = 0.022; Fig.3b). Visual inspection showed
that before the age of 8months no dierences in theta activity were observable but between the ages of 8 and
14months, at greater ages, there were greater theta power values. Weight and the interaction between age and
weight had no eect (all ps > 0.05).
Figure2. e association between age and relative duration of (a) drowsiness, (b) NREM and (c) REM sleep.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

7
Vol.:(0123456789)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
Age was positively associated with alpha power activity (F = 3.289, p = 0.045; Fig.3c). Visual inspection indi-
cated that before the age of 8months no dierences in alpha activity were observable but between the ages of 8
and 14months, at greater ages, there were greater alpha power values. Moreover, there was an interaction between
age and weight (F = 2.154, p = 0.039), with visual inspection suggesting that aer the age of 8months, larger
dogs had more alpha activity. Weight had no main eect (F = 1.489, p = 0.324). Aer exclusion of the four extra
large dogs, analyses were repeated and the interaction eect between age and weight was no longer signicant
(F = 1.777, p = 0.109). As previously, age was positively associated with alpha power activity (F = 5.478, p = 0.002)
and weight had no main eect (F = 1.418, p = 0.240).
Age was positively associated with sigma power activity (F = 9.746, p = 0.003; Fig .5a). Visual inspection
showed a linear association, at greater ages, there were greater sigma power values. Weight and the interaction
between age and weight had no eect (all ps > 0.05).
Age was positively associated with beta power activity (F = 38.333, p < 0.001; Fig.5b). Visual inspection indi-
cated that at greater ages, there were greater beta power values. Moreover, there was an interaction between age
and weight (F = 2.495, p = 0.016; see Supplementary Fig.S5), with visual inspection suggesting that aer the age
of 8months, larger dogs had more beta activity. Weight had no main eect (F = 0.123, p = 0.727). Aer exclusion
of the four extra large dogs, analyses were repeated and the interaction eect between age and weight remained
signicant (F = 2.164, p = 0.037). As previously, age was positively associated with beta power activity (F = 52.186,
p < 0.001) and weight had no main eect (F = 2.606, p = 0.114).
See Table2 for a summary of spectral GAM results.
To see the possible inuence of body condition of mix breeds dogs on our relevant variables (delta, alpha
and beta power activity), we run additional analysis with mix breed dogs excluded. For results see the “Sup-
plementary” document.
Extended sample (2–30 months). To illustrate trends in the data, descriptive data on sleep parameters
for the Extended sample are depicted in gures (Figs.6 and 7).
Based on visual inspection, with age, dogs seem to spend more time in drowsiness and less time in NREM,
and these associations seem to be closer to linear (see Fig.6a,b) than in the Juvenile sample. Moreover, except
from more REM sleep in early life, REM appears to remain stable with age (see Fig.6c), though it is noteworthy
that 2–3-month-old dogs showed a lot of variance in this regard.
More linear associations also appear with respect to spectral values; in older dogs, there seem to be lower
delta power but greater theta, alpha, sigma and beta power activity (see Fig.7a–e).
Table 1. Results of generalized additive models. A vector of smoothing parameter (sp) was added in the GAM
of NREM sleep, because in the original model the degree of smoothness was overtted, assuming a more
complex association between age and NREM due to great variance in the data.
Drowsiness. Formula: drows. ~ s(age) + s(weight) + ti(age, weight)
Parametric coecients Estimate SE t p
Intercept 3.079 0.065 47.1 < 0.001
Smooth terms edf Ref. df F p
s (age) 2.340 2.870 9.517 < 0.001
s (weight) 1 1 0.245 0.625
ti (weight, age) 6.499 8.582 1.895 0.083
R2 (adj) = 0.461, deviance explained = 52.8%
NREM. Formula: NREM ~ s(age, sp = 10) + s(weight) + ti(age,
weight)
Parametric coecients Estimate SE t p
Intercept 59.679 1.751 34.09 < 0.001
Smooth terms edf Ref.df F p
s (age) 1.204 1.371 1.598 0.192
s (weight) 1 1 1.066 0.307
ti (weight, age) 5.685 7.688 1.334 0.260
R2 (adj) = 0.19, deviance explained = 29.8%
REM. Formula: REM ~ s(age) + s(weight) + ti(age, weight)
Parametric coecients Estimate SE t p
Intercept 2.728 0.094 28.92 < 0.001
Smooth terms edf Ref.df F p
s (age) 2.1 2.591 4.363 0.014
s (weight) 1 1 2.543 0.117
ti (age, weight) 1.82 2.275 1.013 0.311
R2 (adj) = 0.282, deviance explained = 22.9%
Content courtesy of Springer Nature, terms of use apply. Rights reserved

8
Vol:.(1234567890)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
Discussion
is study aimed to address some gaps in knowledge via investigating family dogs’ sleep from a developmental
perspective based on a large cross-sectional sample of non-invasive polysomnography data. We report associa-
tions between age and indices of sleep electrophysiology—macrostructure and power spectrum, as well as an
interaction between these associations and dogs’ size (weight).
Figure3. e association between age and (a) delta, (b) theta and (c) alpha power activity.
Figure4. e association between age and delta power activity, given weight. Darker and larger dots indicate
larger dogs.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

9
Vol.:(0123456789)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
In a subset of the Juvenile sample (between the ages of 2 and 8months) we found that older dogs spent more
time in drowsiness. To our knowledge, in other non-human mammals drowsiness sleep stage was not analysed in
the context of neuro-development. In human infants, NREM 1 sleep, the stage most similar to dog drowsiness7,
is observable from the age of < 2 months61,62, based on the scoring suggestions of Anders etal.63. Relative to
younger age groups (4–12months and 13–24months), infants between 25 and 28months exhibit greater NREM
1 sleep62. However, later on in development (> 2years, up to adolescence) age is not associated with this sleep
parameter19,64. Taken together, these results (dog drowsiness, human NREM 1) suggest parallels between age-
related dierences in dog and human sleep with regard to time spent in drowsiness, in terms of older age being
linked to more time in drowsiness until a certain point (human: 2years; dog: 8months), but there being no
age-related dierences in this sleep parameter thereaer (suggesting beginning of stabilisation). Earlier ndings
show that in humans, this stabilisation occurs during childhood, whereas the current data suggest that in dogs,
it may occur at around the age of 8months.
We did not nd an association between age and the relative duration of NREM sleep. is is inconsistent with
observations in the rat showing an age-related decrease in NREM beginning in adolescence13. Observations in
humans also described a developmental decrease in slow wave sleep19. However, others did not nd such age-
related changes in NREM stage 2 nor in slow wave sleep (stage 3)62,64. is inconsistency may be a result of our
ndings with NREM including both stage 2 and slow wave sleep (stage 3), as in dogs, NREM sleep stages are not
as dierentiated as in humans, but, rather, constitute one stage. Of further note, there were a few dogs with very
large NREM proportion values, who may have contributed to the age-NREM association being non-signicant.
We did not remove these animals from the analyses as there was no indication of their NREM data being inac-
curate or resulting from a measurement error. As such, additional dogs with such values in a larger sample may
tip the scale towards the negative age-NREM relation.
Regarding REM sleep, we observed similar age-related changes in dogs as in other species. In rats and cats13 as
well as in humans12,62 there is an age-related decrease in the relative time spent in REM. In humans, this decrease
stabilises around the age of 5 years12,64. However, others observed age-related changes in REM sleep, specically,
relative to younger and older age groups, young adults between 19 and 29years exhibited greater REM sleep19. In
dogs, we observed a similarly negative relation between age and REM sleep that stabilised aer 6months, which
roughly corresponds to the mid-late juvenile period (i.e., between infancy and before puberty/1–12years) in
humans as indicated by research on epigenetic translation between dog and human age65.
In the current Juvenile sample, older dogs exhibited lower delta and higher theta, alpha, sigma and beta power
activity, mostly showing the above associations evident between 8 and 14months of age, partially in line with
earlier ndings. First, an age-related decrease in delta power has also been observed in invasive studies in dogs22
as well as in a previous non-invasive study on a sample of adult dogs (2+years)7. Second, in rats and humans, slow
wave activity follows an inverted U-shaped trajectory14,20,64. Specically, in humans, age is positively associated
with delta power during the rst years of life but there is an age-related decrease in delta power from childhood
to adolescence20,64. Higher frequencies increase in power over childhood62,66 and age-related dierences in fre-
quency ranges of delta, theta, alpha, sigma and beta are still present in older age groups64. As such, at least insofar
Figure5. e association between age and (a) sigma and (b) beta power activity.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

10
Vol:.(1234567890)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
as age-related dierences in delta power go, the rst 8months in dogs may correspond to the rst few years in
humans. Further, brain electrophysiology in dogs between 8 and 14months is comparable to that observed in
humans in early childhood and beyond. Both sets of ndings suggest human–dog age-related correspondences
that are consistent with earlier estimates of such correspondence65.
Of further note, relative to earlier human studies with a limited number of electrodes19,20,67, the authors of
recent studies successfully mapped changes in sleep EEG with high spatial resolution62,64,66. As in dog sleep EEG
studies, four EEG channels are used (e.g.29,68), and in the current sample we analysed only the frontal EEG chan-
nel, changes in cortical topography of dog sleep could not be explored.
It has been argued that the decrease in delta power is a correlate of synaptic pruning in humans18,21. ere is
evidence of synaptic pruning in animals, such as in mice (as indexed by the number of spines within the brain
and alterations in molecular signals involved in pruning69) and adolescent/pubertal rats (indexed by the num-
ber of the number of synaptophysin‐immunoreactive boutons70 and adolescent primates (e.g., Macaca mulatta;
dened by overlapping vesicular glutamate transporter 1-positive (VGlut1+) and postsynaptic density 95-posi-
tive (PSD95+) puncta71), though we could not identify any studies examining whether decreased delta power is
related to pruning in these species. In the current study, we have shown an age-related decrease in delta power
Table 2. Data of generalized additive models.
Delta. Formula: delta ~ s(age) + s(weight, sp = 10) + ti(age, weight)
Parametric coecients Estimate SE t p
Intercept 4.49 0.006 79.9 < 0.001
Smooth terms edf Ref. df F p
s (age) 2.142 2.625 3.337 0.045
s (weight) 1.086 1.134 0.103 0.804
ti (weight, age) 7.093 9.182 2.08 0.049
R2 (adj) = 0.36, deviance explained = 46.5%
eta. Formula: theta ~ s(age) + s(weight) + ti(age, weight)
Parametric coecients Estimate SE t p
Intercept 1.926 0.044 44.12 < 0.001
Smooth terms edf Ref. df F p
s (age) 2.672 3.275 3.361 0.022
s (weight) 1 1 0.271 0.605
ti (weight, age) 5.815 7.801 1.311 0.226
R2 (adj) = 0.26, deviance explained = 41.7%
Alpha. Formula: alpha ~ s(age) + s(weight) + ti(age, weight)
Parametric coecients Estimate SE t p
Intercept 0.587 0.048 12.13 < 0.001
Smooth terms edf Ref.df F p
s (age) 2.301 2.817 3.289 0.045
s (weight) 1 1 0.250 0.619
ti (weight, age) 7.665 9.753 2.154 0.039
R2 (adj) = 0.39, deviance explained = 48.8%
Sigma. Formula: sigma ~ s(age) + s(weight) + ti(age, weight)
Parametric coecients Estimate SE t p
Intercept − 0.278 0.064 − 4.345 < 0.001
Smooth terms edf Ref. df F p
s (age) 1 1 9.746 0.003
s (weight) 1 1 0.114 0.737
ti (weight, age) 7.267 9.39 1.942 0.065
R2 (adj) = 0.39, deviance explained = 48.7%
Beta. Formula: beta ~ s(age) + s(weight) + ti(age, weight)
Parametric coecients Estimate SE t p
Intercept − 0.0379 0.079 − 0.477 0.635
Smooth terms edf Ref.df F p
s (age) 1 1 38.333 < 0.001
s (weight) 1 1 0.123 0.727
ti (weight, age) 8.114 10.27 2.495 0.016
R2 (adj) = 0.47, deviance explained = 62.4%
Content courtesy of Springer Nature, terms of use apply. Rights reserved

11
Vol.:(0123456789)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
activity in dogs and whether such decrease reects synaptic pruning in this species is a testable hypothesis for
future research.
Although there was an interaction eect between age and weight on delta, alpha and beta power activity (aer
the age of 8months, larger dogs had less delta and more alpha and beta activity), this may have been driven by
four large dogs that were disproportionately distributed in terms of age, as when analyses were repeated without
these animals, the interaction eect became trend-level in case of delta and alpha power activity, and remained
signicant only in case of beta power activity. If this tendency reects a true eect (between 8 and 14months;
weight ranged between 4 and 68kg), it may be that at some point, these larger and thus slower-maturing animals
(e.g.,52), begin to neurally “catch up” with their same-aged but smaller-sized counterparts, in terms of reaching
or transitioning into adolescence. In addition, in the current study, we used the standard weight of pure-bred
dogs, but an owner-reported weight for mongrels. It may be worth to consider the body condition of mongrels
in future research, to gain a more accurate measure resembling breed standard weight.
We extended the analysis sample with data from previous dog sleep EEG studies7,26–29 to examine trends in
when adult-like sleep parameters stabilise, if not before the age of 14months, with the ages of the Extended
sample ranging from 2 to 30months. In the Extended sample, visual inspection showed similar associations
between age and drowsiness and power spectra data as in the analysis sample, but these relations were closer
to linear. In the Extended sample, visual examination indicated no association between age and REM but a
negative association between age and NREM, unlike in the Juvenile sample where age was statistically associ-
ated with REM but not with NREM. is pattern across observations and results underscores the importance
of careful consideration of age-related research question and sample denition; that is, the same phenomenon
may manifest quite dierently depending on the age-range within which it is examined, atter or non-signicant
Figure6. e association between age and relative duration of (a) drowsiness, (b) NREM and (c) REM sleep in
the Extended sample.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

12
Vol:.(1234567890)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
associations become steeper or signicant and vice versa. In the Extended sample, visual examination further
suggested that age-related dierences in power spectrum do not stabilise by 14months of age or, for that matter,
even by 30months of age. By analogy, the quantity of dog fast frontal sleep spindles also increases throughout
the lifespan35, indicating that in dogs, certain sleep-related physiological processes mature later and their matu-
ration continues on for longer.
Future directions and limitations
In this study, data from single measurements were used, thus we did not examine whether a rst-night eect29
inuences the observed relations between variables. Further, short aernoon naps were measured, in locations
unfamiliar to the dogs indicating our ndings may not generalise to sleep at nighttime and/or at home. Determin-
ing whether the developmental patterns observed here can be replicated under dierent circumstances in these
regards is thus a potential next step in this line of research. Nevertheless, although we had previously shown that
dierences in sleep parameters are associated with dierences in both the timing and the location of sleep26, all
dogs in the current study were measured in the aernoon and in unfamiliar locations, thus within-individual
relations between age and sleep parameters may not be aected by these factors.
Age-related dierences were measured cross-sectionally. us, the ndings obtained in the current study are
indicative of associations between age-related dierences and the measured sleep parameters but not of tem-
poral, within-animal changes over time. As such, these results indicate it is also warranted to undertake larger,
longitudinal studies to assess the questions examined herein.
Regarding the interaction eect between weight and age on delta power activity, it is unclear whether the
observed nding reects a true eect or is a spurious one. e eect of weight—and of the variables weight was
a proxy for—on the relation between age and delta power activity remains an area in need of further inquiry,
ideally in studies that are explicitly designed to assess such eect. As the relation between heart rate and weight
in dogs has been recently shown to be more complex than previously assumed34, the association between age,
delta power activity—or sleep electrophysiology more generally—and weight may not be straightforward.
Figure7. e association between age and (a) delta, (b) theta, (c) alpha, (d) sigma, (e) beta power activity in
the Extended sample.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

13
Vol.:(0123456789)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
In the current research, weight was used as a proxy for dierences in morphological features across dogs
and dierences in developmental rates between breeds. Although adult weight in dogs tends to correspond to
muscle and skull thickness, there may be nuanced aspects of these characteristics that are not well approximated
by weight but do impact sleep EEG. Nevertheless, the current sample was representative both with regard to
head musculature and skull shape and thickness and with regard to weight, which, although prevented direct
comparisons of subgroups, also excludes the possibility of any strong biases in our data that could have been
due to these features.
As data trends indicated age-related dierences in power spectrum may not stabilise by 14months of age
(as in humans and rats, where such dierences do not stabilise by adolescence2—or, for that matter, not even by
30months—there is a need to quantitatively analyse age-related dierences in these variables across a broader age
range. Nevertheless, based on the current ndings, stabilization of neurodevelopment in the dog brain does not
occur by the commonly adopted 12-month mark. is is similar to humans, where neuromaturation continues
well into “young adulthood”, with gray/white matter changes continuing beyond the age of 1872 and white matter
developing into the late 20s73, despite the commonly accepted 18-year cuto of adulthood72,73.
A potential parallel with regard to dierent relations between age and indices of sleep electrophysiology
depending on the specic sleep parameter examined, in humans, there is a negative association between declara-
tive memory and REM in ~ 5-year-old children, whereas there is an increasingly stronger negative association
between working memory and NREM slow bands from childhood to adolescence74. Establishing age-related
cutos or transition periods with regard to the association between age, indices of sleep electrophysiology, and
dierent cognitive domains in dogs might also be worth examining in the future.
Conclusions
Dogs are among the most studied species, both on their own right as pet animals and as a model of human social
cognition and the neural correlates and underpinnings thereof. Yet, developmental dierences in dog sleep elec-
trophysiology were previously unknown. Further, in the absence of empirical evidence, it has been (implicitly)
assumed in the dog behaviour and brain function literature that dogs reach adulthood at the age of 12months.
In this rst, large-scale non-invasive study on the association between age and sleep parameters in dogs, we
have shown that measured indices of sleep are a good index of neurodevelopment in dogs, with some parameters,
such as drowsiness and REM sleep, stabilizing in around 6–8months, and some, such as power spectrum, not
stabilizing even by 30months. As such, the stabilization of neurodevelopment in the dog brain is a complex
matter and denitive conclusions about the age at which dogs reach adulthood cannot yet be drawn. Further,
the relevant eects of dierences across breed will need to be addressed. Our ndings do suggest that, given
robust link between developmental status of the nervous system and time spent in REM sleep—which, in dogs,
stabilises around 6months—samples including dogs younger than 6months of age should not be conceptualised
as adult samples or as homogenous with regard to maturity of the nervous system. Alternatively, if the aim is
to make comparisons across age groups, those should be demarcated in light of this maturational pattern. Our
results also suggest both dierences and similarities in terms of age-related variation in sleep macrostructure
and power spectrum across dogs and humans as well as across dogs and other model species, further underscor-
ing the validity of the dog for modelling human sleep but also highlighting potentially important inter-species
developmental dierences.
Received: 4 February 2021; Accepted: 10 November 2021
References
1. Siegel, J. M. Do all animals sleep?. Trends Neurosci. 31, 208–213 (2008).
2. Kurth, S., Olini, N., Huber, R. & LeBourgeois, M. Sleep and early cortical development. Curr. Sleep Med. Reports 1, 64–73 (2015).
3. 1. Carskadon, M. A. & Dement, W. C. Monitoring and staging human sleep. in Principles and practice of sleep medicine (eds. Kryger,
M. H. et al.) 16–26 (Elsevier Saunders, 2011).
4. Frank, M. G. Sleep and developmental plasticity: Not just for kids. Prog. Brain Res. 193, 221–232 (2011).
5. Colrain, I. M. Sleep and the brain. Neuropsychol. Rev. 21, 1–4 (2011).
6. Ruckebusch, Y. e relevance of drowsiness in the circadian cycle of farm animals. Anim. Behav. 20, 637–643 (1972).
7. Kis, A. et al. Development of a non-invasive polysomnography technique for dogs (Canis familiaris). Physiol. Behav. 130, 149–156
(2014).
8. Z epelin, H., Siegel, J. & Tobler, I. Mammalian sleep. in Principles and practice of sleep medicine (eds. Kryger, M., Roth, T. & Dement,
W.) 91–100 (Elsevier Saunders, 2005).
9. Fernandez, L. M. J. & Lüthi, A. Sleep spindles: Mechanisms and functions. Physiol. Rev. 100, 805–868 (2020).
10. Voss, U. Functions of sleep architecture and the concept of protective elds. Rev. Neurosci. 15, 33–46 (2004).
11. Hagenauer, M. H. & Lee, T. M. Adolescent sleep patterns in humans and laboratory animals. Horm. Behav. 64, 270–279 (2013).
12. Rowarg, H. P., Muzio, J. N. & Dement, W. C. Ontogenetic development of the human sleep-dream cycle. Science (80-). 152,
604–619 (1966).
13. Jouvet-Mounier, D., Astic, L. & Lacote, D. Ontogenesis of the states of sleep in rat, cat, and guinea pig during the rst postnatal
month. Dev. Psychobiol. 2, 216–239 (1969).
14. Olini, N., Kurth, S. & Huber, R. e eects of caeine on sleep and maturational markers in the rat. PLoS ONE 8, e72539 (2013).
15. Tiriac, A., Uitermarkt, B. D., Fanning, A. S., Sokolo, G. & Blumberg, M. S. Rapid whisker movements in sleeping newborn rats.
Curr. Biol. 22, 2075–2080 (2012).
16. Zhdanova, I. V. et al. Aging of intrinsic circadian rhythms and sleep in a diurnal nonhuman primate, Macaca mulatta. J. Biol.
Rhythms 26, 149–159 (2011).
17. Aujard, F., Cayetanot, F., Bentivoglio, M. & Perret, M. Age-related eects on the biological clock and its behavioral output in a
primate. Chronobiol. Int. 23, 451–460 (2006).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

14
Vol:.(1234567890)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
18. Feinberg, I. & Campbell, I. G. Sleep EEG changes during adolescence: An index of a fundamental brain reorganization. Brain Cogn.
72, 56–65 (2010).
19. Gaudreau, H., Carrier, J. & Montplaisir, J. Age-related modications of NREM sleep EEG: From childhood to middle age. J. Sleep
Res. 10, 165–172 (2001).
20. Campbell, I. G. & Feinberg, I. Longitudinal trajectories of non-rapid eye movement delta and theta EEG as indicators of adolescent
brain maturation. Proc. Natl. Acad. Sci. USA 106, 5177–5180 (2009).
21. Huttenlocher, P. R. Synaptic density in human frontal cortex—Developmental changes and eects of aging. Brain Res. 163, 195–205
(1979).
22. Fox, M. W. Postnatal development of the EEG in the Dog—II: Development of electrocortical activity. J. Small Anim. Pract. 8,
77–107 (1967).
23. Koelsch, S., Heinke, W., Sammler, D. & Oltho, D. Auditory processing during deep propofol sedation and recovery from uncon-
sciousness. Clin. Neurophysiol. 117, 1746–1759 (2006).
24. Hare, B., Brown, M., Williamson, C. & Tomasello, M. e domestication of social cognition in dogs. Science (80-). 298, 1634–1636
(2002).
25. Topál, J. et al. e dog as a model for understanding human social behavior. Adv. Study B ehav. 39, 71–116 (2009).
26. Bunford, N. et al. Dierences in pre-sleep activity and sleep location are associated with variability in daytime/nighttime sleep
electrophysiology in the domestic dog. Sci. Rep. 8, 7109 (2018).
27. Kis, A. et al. e interrelated eect of sleep and learning in dogs (Canis familiaris); an EEG and behavioural study. Sci. Rep. 7,
41873 (2017).
28. Kis, A. et al. Sleep macrostructure is modulated by positive and negative social experience in adult pet dogs. Proc. R. Soc. B Biol.
Sci. 284, https:// doi. org/ 10. 1098/ rspb. 2017. 1883 (2017).
29. Reicher, V. et al. Repeated aernoon sleep recordings indicate rst-night-eect-like adaptation process in family dogs. J. Sleep Res.
https:// doi. org/ 10. 1111/ jsr. 12998 (2020).
30. Bódizs, R., Kis, A., Gácsi, M. & Topál, J. Sleep in the dog: Comparative, behavioural and translational relevance. Curr. Opin. Behav.
Sci. 33, 25–33 (2020).
31. Toth, L. A. & Bhargava, P. Animal models of sleep disorders. Comp. Med. 63, 91–104 (2013).
32. Shimazono, Y. et al. 6. e correlation of the rhyihmic waves of the hippocampus with the behaviors of dogs. Neurol. Med. Chir.
(Tokyo). 2, 82–88 (1960).
33. Wauquier, A., Verheyen, J. L., Van Den Broeck, W. A. E. & Janssen, P. A. J. Visual and computer-based analysis of 24 h sleep-waking
patterns in the dog. Electroencephalogr. Clin. Neurophysiol. 46, 33–48 (1979).
34. B álint, A. et al. Potential physiological parameters to indicate inner states in dogs: e analysis of ECG, and respiratory signal
during dierent sleep phases. Front. Behav. Neurosci. 13, 207 (2019).
35. Iotchev, I. B. et al. Age-related dierences and sexual dimorphism in canine sleep spindles. Sci. Rep. 2, 9 (2019).
36. Gácsi, M., Kara, E., Belényi, B., Topál, J. & Miklósi, Á. e eect of development and individual dierences in pointing comprehen-
sion of dogs. Anim. Cogn. 12, 471–479 (2009).
37. Gácsi, M. et al. Explaining dog wolf dierences in utilizing human pointing gestures: Selection for synergistic shis in the develop-
ment of some social skills. PLoS ONE 4, e6584 (2009).
38. Fugazza, C., Moesta, A., Pogány, Á. & Miklósi, Á. Presence and lasting eect of social referencing in dog puppies. Anim. Behav.
141, 67–75 (2018).
39. Salamon, A., Száraz, J., Miklósi, A. & Gácsi, M. Movement and vocal intonation together evoke social referencing in companion
dogs when confronted with a suspicious stranger. Anim. Cogn. 23, 913–924 (2020).
40. Merola, I., Prato-Previde, E. & Marshall-Pescini, S. Dogs’ social referencing towards owners and strangers. PLoS ONE 7, e47653
(2012).
41. Bray, E. E. et al. Cognitive characteristics of 8- to 10-week-old assistance dog puppies. Anim. Behav. 166, 193–206 (2020).
42. Bray, E. E., MacLean, E. L. & Hare, B. A. Context specicity of inhibitory control in dogs. Anim. Cogn. 17, 15–31 (2014).
43. Gross, B., Garcia-Tapia, D., Riedesel, E. & Norman Matthew Ellinwood, J. K. J. Normal canine brain maturation at magnetic
resonance imaging. Vet. Radiol. Ultrasound 51, 361–373 (2010).
44. Hecht, S. & Hodshon, A. Aging changes of the brain. in Diagnostic MRI in Dogs and Cats (ed. Mai, W.) 318–325 (CRC Press, 2018).
45. Kubinyi, E. & Iotchev, I. B. A preliminary study toward a rapid assessment of age-related behavioral dierences in family dogs.
Animals 10, 1–10 (2020).
46. Tapp, P. D. et al. Size and reversal learning in the beagle dog as a measure of executive function and inhibitory control in aging.
Learn. Mem. 10, 64–73 (2003).
47. Gunde, E. et al. Longitudinal volumetric assessment of ventricular enlargement in pet dogs trained for functional magnetic reso-
nance imaging (fMRI) studies. Vet. Sci. 7, 127 (2020).
48. Passalacqua, C. et al. Human-directed gazing behaviour in puppies and adult dogs, Canis lupus familiaris. Anim. Behav. 82,
1043–1050 (2011).
49. Wallis, L. J. et al. Lifespan development of attentiveness in domestic dogs: Drawing parallels with humans. Front. Psychol. 5, 71
(2014).
50. Salt, C. et al. Growth standard charts for monitoring bodyweight in dogs of dierent sizes. https:// doi. org/ 10. 1371/ journ al. pone.
01820 64. (2017).
51. Law, S. K. ickness and resistivity variations over the upper surface of the human skull. Brain Topogr. 6, 99–109 (1993).
52. Favier, R. P., Mol, J. A., Kooistra, H. S. & Rijnberk, A. Large body size in the dog is associated with transient GH excess at a young
age. J. Endocrinol. 170, 479–484 (2001).
53. Mialot, J. P., Guerin, C. & Begon, D. Growth, testicular development and sperm output in the dog from birth to post pubertal
period. Andrologia 17, 450–460 (1985).
54. Ford, L. Testicular maturation in dogs. Am. J. Vet. Res. 30, 331 (1969).
55. Berry, B. R. et al. e AASM Manual for the Scoring of Sleep and Associated Events. Rules, Terminology and Technical Specications
(American Academy of Sleep Medicine, 2012).
56. Gergely, A. et al. Reliability of Family Dogs’ Sleep Structure Scoring Based on Manual and Automated Sleep Stage Identication.
Animals 10, 927 (2020).
57. Wood, S. N. Generalized Additive Models: An Introduction with R (Chapman and Hall, 2006).
58. Meulman, N., Wieling, M., Sprenger, S. A., Stowe, L. A. & Schmid, M. S. Age eects in L2 grammar processing as revealed by ERPs
and how (not) to study them. PLoS ONE 10, 1–27 (2015).
59. Tremblay, A. & Newman, A. J. Modeling nonlinear relationships in ERP data using mixed-eects regression with R examples.
Psychophysiology 52, 124–139 (2015).
60. RCoreTeam. R: A language and environment for statistical computing. (2017).
61. Jenni, O. G., Borbély, A. A. & Achermann, P. Development of the nocturnal sleep electroencephalogram in human infants. Am. J.
Physiol. Regul. Integr. Comp. Physiol. 286, 528–538 (2004).
62. Novelli, L. et al. Mapping changes in cortical activity during sleep in the rst 4 years of life. J. Sleep Res. 25, 381–389 (2016).
63. Anders, T., Emde, R. & Parmelee, A. A Manual of Standardized Terminology, Techniques and Criteria for Scoring of States of Sleep
and Wakefulness in Newborn Infants (UCLA Brain Information Service/Brain Research Institute, 1971).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

15
Vol.:(0123456789)
Scientic Reports | (2021) 11:22760 | https://doi.org/10.1038/s41598-021-02117-1
www.nature.com/scientificreports/
64. Kurth, S. et al. Mapping of cortical activity in the rst two decades of life: A high-density sleep electroencephalogram study. J.
Neurosci. 30, 13211–13219 (2010).
65. Wang, T. et al. Quantitative translation of dog-to-human aging by conserved remodeling of the DNA methylome. Cell Syst. 11,
176-185.e6 (2020).
66. Chu, C. J., Leahy, J., Pathmanathan, J., Kramer, M. A. & Cash, S. S. e maturation of cortical sleep rhythms and networks over
early development. Clin. Neurophysiol. 125, 1360–1370 (2014).
67. Jenni, O. G. & Carskadon, M. A. Spectral analysis of the sleep electroencephalogram during adolescence. Sleep 27, 774–783 (2004).
68. Bolló, H. et al. REM versus Non-REM sleep disturbance specically aects inter-specic emotion processing in family dogs (Canis
familiaris). Sci. Rep. 10, 1–8 (2020).
69. de Cossío, L. F., Guzmán, A., van der Veldt, S. & Luheshi, G. N. Prenatal infection leads to ASD-like behavior and altered synaptic
pruning in the mouse ospring. Brain. Behav. Immun. 63, 88–98 (2017).
70. Drzewiecki, C. M., Willing, J. & Juraska, J. M. Synaptic number changes in the medial prefrontal cortex across adolescence in male
and female rats: A role for pubertal onset. Synapse 70, 361–368 (2016).
71. Chung, D. W., Wills, Z. P., Fish, K. N. & Lewis, D. A. Developmental pruning of excitatory synaptic inputs to parvalbumin interneu-
rons in monkey prefrontal cortex. Proc. Natl. Acad. Sci. USA 114, E629–E637 (2017).
72. Bennett, C. M. & Baird, A. A. Anatomical changes in the emerging adult brain: A voxel-based morphometry study. Hum. Brain
Mapp. 27, 766–777 (2006).
73. Lebel, C. & Beaulieu, C. Longitudinal development of human brain wiring continues from childhood into adulthood. J. Neurosci.
31, 10937–10947 (2011).
74. Gorgoni, M., D’Atri, A., Scarpelli, S., Reda, F. & De Gennaro, L. Sleep electroencephalography and brain maturation: Developmental
trajectories and the relation with cognitive functioning. Sleep Med. 66, 33–50 (2020).
Acknowledgements
is project has received funding from National Research Development and Innovation Oce (OTKA FK
128242; K 132372), the MTA-ELTE Comparative Ethology Research Group (01031), MTA Lendület (“Momen-
tum”) Grant (#LP2018-3/2018), BIAL Foundation (Grant no. 169/16), the János Bolyai Research Scholarship of
the Hungarian Academy of Sciences, ÚNKP-20-3&ÚNKP-21-5 New National Excellence Program of the Minis-
try for Innovation and Technology from the source of the National Research, Development and Innovation Fund.
Author contributions
Conceptualization: V.R., N.B. and M.G.; methodology: V.R., N.B., A.K and M.G.; validation: V.R., N.B. and M.G.;
formal analysis: V.R.; investigation: V.R., N.B., A.K., L.K., B.Cs., C.C.; data curation: V.R, L.K., A.K.; writing—
original dra: V.R., N.B., C.C. and M.G.; writing—review and editing: all authors, visualization: V.R., supervision:
N.B., and M.G., funding acquisition: M.G.
Competing interests
e authors declare no competing interests.
Additional information
Supplementary Information e online version contains supplementary material available at https:// doi. org/
10. 1038/ s41598- 021- 02117-1.
Correspondence and requests for materials should be addressed to V.R.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access is article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© e Author(s) 2021, corrected publication 2022
Content courtesy of Springer Nature, terms of use apply. Rights reserved

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com