We selected the following species for this study based on their phylogenetic divergence within the evolutionary tree, cranial size (allowing electrode placement in comparable positions), and ease of handling by caretakers:

Table 1 summarizes basic information for the experimental cases, specifying the number of animals used, the number of recording sessions per subject, and the starting hours and recording duration in each species. A total of thirteen sessions were analyzed across six animals. Experimental protocols for all three species were approved by the respective animal ethics and welfare committees of the institutions where the animals resided: Oceanogràfic of Valencia and Fundación Parques Reunidos (Zoo of Madrid and Faunia).

The main challenge for non-invasive EEG acquisition was the design of a lightweight, autonomous device with sufficient battery capacity to record high-quality EEG signals for 8–12 h. In collaboration with SmartEEG®, we developed and adapted a next-generation EEG device through an electronic engineering process that included the design of new optimized chips for EEG recording, representing a technological advance for both human and animal applications (Figure 1).

Spectral analysis was performed using a Short-Time Fourier Transform (STFT) with the following parameters to ensure high frequency resolution:

Instead of a standard electrode set, we manually configured two electrodes adapted to the cranial morphology of each species, ensuring easier placement over brain regions. Cup electrodes were modified by shortening the cable and micro-soldering them to the base connections of the UMindMirror PCB. Different attachment mechanisms were used depending on species (Figure 2):

The device was always fitted on conscious animals, following animal welfare guidelines. Procedures were conducted within their enclosures with the assistance of regular caretakers to minimize stress. The protocol included:

To ensure reproducibility of superficial EEG recordings across the three species, electrode placement over the central region of the cerebral hemisphere (sensorimotor cortex) was standardized using external cranial landmarks (Figure 2). In all specimens, the active electrode was positioned in a referential montage relative to the sagittal midline and the inter-aural axis (IAA).

Tortoise (Aldabrachelys gigantea): Accounting for the specialized cranial shield, the electrode was positioned 5 mm lateral to the sagittal midline and 5 mm anterior to the IAA. The midpoint of the frontal plate served as the primary landmark to center recordings over the dorsal cerebral hemispheres, over the anterior part of parietal bones (45).

Penguin (Pygoscelis papua): The electrode was placed 12 mm lateral to the sagittal midline and 10 mm anterior to the IAA. This location aligns with the posterior margin of the orbital rim, overlying the frontal bone to capture EEG activity while avoiding the supraorbital glands (44).

Aardvark (Orycteropus afer): Given its dolichocephalic morphology, the electrode was placed over the parietal bone, 15 mm lateral to the sagittal midline and 10 mm anterior to the IAA. This targets the post-orbital region, minimizing artifacts from the temporal musculature (43).

After placement, the EEG device was connected via Bluetooth to the freely available Android Sleep Recorder 1.2.0 Pro app, installed on a tablet within a 15 m line-of-sight range. Signal quality, electrode placement, and battery status were verified before starting the recording session.

EEG data were transmitted via Bluetooth and stored on the tablet within the directory structure of the Sleep Recorder app. The steps required for proper recording were:

Data were saved in European Data Format (EDF), an open, non-proprietary standard for medical time-series data, ensuring interoperability and compatibility with different analysis platforms and facilitating data sharing between institutions.

EEG signals were analyzed using the open-source, cross-platform software EDFbrowser (https://www.teuniz.net/edfbrowser/). To characterize the electroencephalographic (EEG) profile across different behavioral states, we calculated the relative power of the standard frequency bands (delta, theta, alpha, beta, and gamma) as a percentage of the total spectral power. For each species (tortoise, penguin, and aardvark), we analyzed five 30-s epochs obtained from representative episodes of each state (Wakefulness, NREM, and REM sleep). Statistical comparisons of these relative power values were performed using a nested repeated-measures ANOVA (StatView 5.0, SAS Institute Inc., Cary, USA). This design accounted for the hierarchical structure of the data by nesting multiple recordings within individual subjects. By partitioning the total variance into within-subject and between-subject components, this approach avoided pseudoreplication and ensured that group comparisons were based on the appropriate level of replication. Post-hoc analyses were conducted using Tukey’s Honest Significant Difference (HSD) test to identify specific pairwise differences between states and species, with the threshold for statistical significance set at p ≤ 0.05.

As per standard nomenclature in sleep and arousal research, desynchronized activity refers to a brain state where the EEG displays low-amplitude, mixed-frequency patterns. In all three species, three distinct EEG patterns were consistently observed based on synchronization-desynchronization features present in the recordings:

To ensure the accuracy of these observations, we utilized continuous video monitoring during the study. A comparative analysis of this footage confirmed that the animals’ behavioral parameters remained consistent with baseline periods when the device was not present. Furthermore, since recordings were primarily conducted during habitual rest periods, the video data verified that sleep patterns were undisturbed and occurred naturally as on non-recordin days.

Spectral analysis confirmed the presence of delta (0.5–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), and beta (12–30 Hz) bands across all species, indicating that the frequency ranges of brain oscillations are conserved among amniotes. However, signal amplitude varied between species, largely reflecting anatomical differences such as skull thickness, tissue composition, and electrode placement.

The non-invasive system used in this study maintained stable recordings for 10–12 h, covering complete sleep–wake cycles. Bluetooth transmission was effective within a round perimeter up to 15 m radius, allowing naturalistic observation of the animals in their enclosures.

The use of wet cup electrodes (Ag/AgCl) provided low impedance and minimized motion artifacts, even in active animals. Minor variations in amplitude across species were attributable to anatomical differences rather than methodological limitations.

To assess the significance of EEG frequency band power densities of the different behavioral vigilance states across species, we designed a nested three-way ANOVA (factors: species, state, frequency band) evaluating whether there was significant variability in relative EEG power with respect to (1) species, (2) state and/or (3) frequency band. The overall test revealed a significant three-way interaction between species, state and frequency band (F_16,96 = 31.41, p < 0.0001), indicating that the distribution of spectral power across different frequency bands during sleep–wake states varies fundamentally between the tortoise, penguin and aardvark. Significant two-way interactions were also found for state x frequency band (F_8,96 = 523.64, p < 0.0001) and species x frequency band (F_8,48 = 210.24, p < 0.0001). Given the significant three-way interaction, subsequent analyses were decomposed by species to examine the specific state-dependent spectral shifts within each group.

Overall, this methodology enabled the identification of conserved sleep–wake states (wakefulness, NREM, and REM) in reptiles, birds, and mammals. Spectral frequency analysis demonstrated that the distribution of EEG frequency bands (delta, theta, alpha, beta) is consistent across these groups, supporting the evolutionary conservation of sleep mechanisms. However, species-specific EEG patterns were also overtly detected. The system proved effective, reproducible, and compatible with animal welfare standards, confirming its potential for future applications in comparative neuroscience, veterinary medicine, and zoological research.

Traditional EEG studies in non-human animals have relied primarily on invasive methods requiring anesthesia, surgical electrode implantation, and/or physical restraint, procedures that inevitably compromise animal welfare and alter normal behaviors (9, 21). In contrast, the present methodology is based on a lightweight, autonomous device coupled with wet Ag/AgCl cup electrodes, providing stable, long-duration recordings without the need for invasive preparation or prolonged immobilization.

This approach represents a significant advancement in both comparative neuroscience and animal welfare, as it aligns with the five-domain model of welfare (1, 2), enabling neurophysiological monitoring without compromising behavioral integrity. The ability to obtain 10–12 h of continuous EEG data under naturalistic conditions opens new opportunities for longitudinal studies in zoological institutions, veterinary practice, and wildlife conservation programs.

Our results provide evidence of conserved sleep–wake dynamics across amniotes. In all three species, wakefulness was marked by low-amplitude, high-frequency activity; NREM sleep by synchronized, high-amplitude delta waves; and REM sleep by desynchronized, high-frequency activity without locomotion. These findings are consistent with previous descriptions of sleep in reptiles, birds, and mammals (3, 22, 23).

The identification of REM-like states in Aldabrachelys gigantea further supports the growing body of evidence that reptiles exhibit homologous sleep states to those described in birds and mammals, reinforcing the hypothesis of an ancient evolutionary origin of REM sleep (24, 25). In the homeothermic species examined (gentoo penguin and aardvark), the observed NREM sleep patterns, characterized by high-amplitude delta waves, and REM sleep, characterized by low-voltage desynchronized activity, are consistent with established descriptions in avian and mammalian literature (3, 8). The fragmented nature of REM sleep in penguins is in line with avian-specific adaptations, likely related to ecological pressures such as predation risk (26). The aardvark recordings, meanwhile, exhibited patterns analogous to those of other large terrestrial mammals, validating the method’s applicability in diverse mammalian taxa.

The electrophysiological profile of the aardvark provides a compelling illustration of the mammalian sleep state. Our findings align with comparative data indicating that mammals exhibit a high-frequency, low-amplitude EEG during REM sleep, characterized by a notable concentration of power in the beta and gamma bands (27, 28, 30, 31). This high-frequency power is significantly more pronounced in the aardvark than in non-mammalian amniotes such as the tortoise or the penguin, where active (REM) sleep states often lack the high-spectral intensity found in eutherian mammals (25). This reinforces the view that robust cortical activation during REM is a hallmark of the mammalian lineage. Also, a primary finding in the aardvark was the significantly higher alpha power during wakefulness compared to the other species. In mammals, alpha oscillations are recognized as essential for top-down inhibitory control and orchestrating long-range cortical communication (29). This alpha dominance, which was significantly suppressed during both NREM and REM sleep in our subjects, suggests a high degree of cortical integration during arousal, consistent with models of synaptic homeostasis where the transition to NREM serves to downscale the high-energy synaptic activity built up during alpha-dominant wakefulness (32). The specific suppression of alpha during sleep reinforces the theory that mammalian sleep requires a fundamental reorganization of the thalamocortical loops that sustain waking consciousness (24).

The comparison between the tortoise and the penguin offers a unique window into the evolution of sleep from the common stem-amniote ancestor. While both species showed state-dependent modulations, the tortoise exhibited a marked spectral homogeneity in REM, with delta levels significantly higher than those of the other species. This suggests that reptilian NREM sleep might represent a less differentiated and highly stable state compared to the more complex spectral distributions seen in birds and mammals (33). In contrast, the penguin displayed a shift towards faster oscillatory dynamics, with significant theta dominance during wakefulness and REM sleep. This shift likely supports the metabolic and cognitive demands of endothermy—where high-frequency neural activity facilitates the complex cognitive maps and efficient foraging necessary to sustain high metabolic rates (34). Furthermore, the specific dominance of theta oscillations is increasingly recognized as a requirement for complex navigation, particularly for encoding spatial trajectories and managing high-level information processing in avian species (36).

The successful detection of sleep patterns in such different animals using non-invasive surface electrodes supports the sensitivity and reliability of our methodology, indicating that anatomical differences—such as the presence of feathers, skull thickness, and subcutaneous fat—do not impede the acquisition of interpretable EEG signals when appropriate recording technology is employed. These findings validate the feasibility of non-invasive neurophysiological monitoring across phylogenetically distant taxa and establish a methodological framework applicable to both research and clinical contexts.

The capacity to record sleep and wake states under non-invasive, stress-free conditions is of particular relevance for zoological institutions, aquariums, sanctuaries, and conservation programs. Sleep quality is a key indicator of animal welfare, with inadequate sleep negatively impacting cognitive, immune, and metabolic functions (4). By enabling routine EEG monitoring, this methodology provides an objective tool for assessing welfare in captive and managed animals.

Furthermore, this non-invasive methodology also demonstrates immediate and transformative potential for clinical veterinary practice. From a diagnostic perspective, it enables routine neurological screenings without the need for sedation, thereby facilitating the early detection of epilepsy, sleep disorders, age-related cognitive decline, encephalitis, and intracranial neoplasms (37, 35). For therapeutic monitoring—particularly in epileptic animals (both human and non-human)—it provides an objective means to assess treatment efficacy through serial recordings documenting seizure frequency and severity. In zoological and wildlife medicine, where anesthesia poses considerable risk, this technique allows for the evaluation of neurological function without compromising animal welfare. Furthermore, the relatively low cost of the device compared to conventional veterinary EEG systems, together with the simplicity of the recording procedures, democratizes access to advanced neurophysiological assessments. Such accessibility has the potential to markedly enhance standards of veterinary neurological care by enabling earlier and more accurate diagnosis, as well as more effective follow-up of therapeutic interventions (38). Its potential applications extend also to agricultural and domestic contexts where sleep quality may influence productivity and health (20).

While the methodology proved effective, certain limitations should be acknowledged. Variability in signal amplitude was observed across species, likely reflecting anatomical differences such as skull thickness and integumentary structures (20, 39). Although these differences did not preclude identification of sleep–wake states, further optimization of electrode placement may improve signal consistency (40). Additionally, aquatic species such as penguins could not be monitored in water due to the technical limitations of the current device. Future developments should focus on waterproofing and expanding the range of wireless transmission, enabling full ecological validity in aquatic environments (41). Finally, while this study demonstrated the feasibility of non-invasive EEG across three species, broader application to other taxonomic groups is necessary to establish its generalizability and refine species-specific protocols (42).

Overall, this study provides the first demonstration of a non-invasive and wireless EEG methodology capable of recording naturalistic sleep–wake cycles in unanesthetized, unrestrained reptiles, birds, and mammals. The technique represents an important step forward for comparative neurobiology and animal welfare, offering a reproducible, welfare-compatible approach that bridges technological innovation with biological and veterinary applications.