Fetal neurobehavior and consciousness: a systematic review of 4D ultrasound evidence and ethical challenges

Study design

This review was conducted in accordance with the PRISMA 2020 guidelines for systematic reviews. The primary objective was to assess the relationship between observable fetal behaviors – particularly as visualized through four-dimensional (4D) ultrasonography – and key neurodevelopmental processes. The secondary aim was to examine how these findings have been interpreted in terms of sensory integration and emerging neural complexity, without asserting direct evidence of fetal consciousness. Given the heterogeneity in study designs and outcome measures, a narrative synthesis approach was adopted in place of meta-analysis.

Data sources and search strategy

A systematic search was carried out across four electronic databases: PubMed, Scopus, Embase, and Web of Science. The search covered literature published between January 2000 and April 2025. Controlled vocabulary and free-text terms were used in combination, including “fetal behavior,” “fetal movement,” “4D ultrasound,” “prenatal neuroimaging,” “fetal MRI,” “thalamocortical connectivity,” “neurodevelopment,” “sensory integration,” and “consciousness.” Boolean operators and database-specific syntax were applied to optimize the search strategy. The final set of results was filtered to include only peer-reviewed articles written in English and involving human subjects.

Eligibility criteria

Studies were eligible for inclusion if they involved human fetuses, utilized 4D ultrasound, fetal MRI, EEG, or fetal magnetoencephalography (fMEG), and reported observable behavioral phenomena in relation to developmental milestones. In addition, included studies were required to engage, either implicitly or explicitly, with interpretive frameworks addressing the significance of fetal behavior in terms of neural function, sensory processing, or ethical reflection. Exclusion criteria encompassed studies based solely on animal models, technical imaging studies without behavioral analysis, non-peer-reviewed literature, and articles not published in English. Reviews and theoretical papers were included only when they contributed original frameworks or critical insights relevant to the subject matter.

Study selection

All retrieved citations were imported into a reference management system, where duplicates were removed. Two independent reviewers screened the titles and abstracts for relevance. Full-text versions of potentially eligible studies were then assessed using pre-defined inclusion and exclusion criteria. Disagreements between reviewers were resolved through discussion, and when necessary, a third senior reviewer was consulted. Of the 218 records initially identified, 74 met the full eligibility criteria. Among these, 23 studies were designated as the core analytical set based on methodological rigor, specificity of fetal behavior interpretation, and engagement with developmental or ethical implications.

Quality assessment

All 74 included studies were evaluated using the AMSTAR-2 framework, a validated tool for appraising the methodological quality of systematic and observational research. This appraisal considered 16 domains, including protocol transparency, appropriateness of inclusion criteria, reliability of data extraction, and the validity of interpretations. To distinguish higher-quality studies, we applied a pre-defined threshold: studies that met at least 10 of the 16 AMSTAR-2 criteria were classified as high-quality and designated as “core studies” (n=23). These studies consistently demonstrated methodological rigor in their design, reporting, and engagement with developmental or ethical implications. The remaining 51 studies were retained as “contextual studies,” offering broader thematic relevance despite some limitations in design or reporting clarity. Although we did not include a full scoring table in supplementary materials, all scoring decisions were reached independently by two reviewers and were internally cross-validated to ensure consistency.

Data extraction

A standardized protocol was used for data extraction. For each eligible study, extracted data included author and year of publication, gestational age at observation, imaging modality used, types of fetal behaviors observed (such as yawning, startle reflex, or facial movements), associated neurodevelopmental interpretations (e.g., reflexive vs. integrative responses), and any discussion of ethical, legal, or philosophical implications.

Data synthesis

Due to the heterogeneity of included studies – particularly in terms of methodologies, gestational windows, and interpretive frameworks – a narrative synthesis was employed. Fetal behavioral patterns were mapped against established neurodevelopmental milestones, including subplate remodeling, synaptic density progression, and thalamocortical connectivity. Studies utilizing artificial intelligence (AI) or machine learning for fetal behavior analysis were synthesized separately to assess the technological and ethical implications of algorithmic interpretation. Ethical content was categorized according to the primary frameworks employed, including utilitarian, relational, deontological, and culturally embedded approaches. Interpretations throughout the synthesis remained grounded in structural and functional neurodevelopmental evidence, without inferring conscious awareness.

Overview of included studies

A total of 218 records were identified through database searches. After removing duplicates and screening abstracts, 101 full-text articles were reviewed. Seventy-four studies met the inclusion criteria: they involved human fetuses, employed 4D ultrasound or related neuroimaging modalities, and addressed behavioral or neurodevelopmental features relevant to the research question. Of these, 23 studies were designated as the core analytical set based on methodological quality and thematic relevance, as assessed using the AMSTAR-2 tool. The remaining 51 studies were retained for contextual analysis. The selection process is summarized in the PRISMA diagram (Figure 1).

Figure 1:

PRISMA Flow diagram of study selection process for fetal consciousness review. This diagram illustrates the systematic review process following PRISMA guidelines. A total of n=218 records were identified across four databases (PubMed, Scopus, Embase, Web of Science). After removing duplicates (n=117), n=101 records were screened for relevance. Of these, n=27 were excluded during title/abstract screening, and n=51 full-text articles were excluded for reasons such as lack of behavioral/neurodevelopmental relevance or insufficient data on volitionality, consciousness, or ethics. Ultimately, n=23 studies were included in the final synthesis. Among these, studies were categorized into “core” (high thematic and methodological relevance) and “contextual” (retained for background insight, e.g., AI tools or imaging advances), based on predefined eligibility criteria and quality assessments using AMSTAR-2 and ROBIS tools.

Table 1 presents a comparative summary of the 23 core studies, including their methods, findings, and quality indicators. These studies comprised observational ultrasonography (n=15), functional imaging (fMRI, EEG, fMEG; n=5), and computational approaches including artificial intelligence (AI)-assisted behavior analysis (n=3).

Observed fetal behaviors by gestational age

Across the reviewed literature, a range of fetal behaviors were consistently documented using 4D ultrasonography. These included yawning, blinking, facial movements, sucking, hand-to-face gestures, limb flexion, and head rotation. These behaviors were typically observable by 12–14 weeks of gestation and exhibited increasing complexity, frequency, and symmetry through the third trimester [1–5], 40], 41].

Several studies, notably by Kurjak et al. and collaborators, reported a progressive organization of fetal movement patterns between 24 and 34 weeks, including lateralized limb activity, isolated thumb-sucking, and coordinated facial-muscle engagement [2–6], 8]. Reissland et al. observed the differentiation of facial expressions over time and categorized changes in mouth and eye movements longitudinally [2], 42].

Although the reviewed studies describe these behaviors using terms such as “yawning” or “grimacing,” they uniformly acknowledge that such labels do not imply emotional content and instead serve as morphological descriptors. The temporal appearance and increased coordination of these behaviors suggest correspondence with central nervous system maturation (Table 3).

Neurodevelopmental correlates

Several included studies used functional neuroimaging or postmortem anatomical data to map behavioral features onto neurodevelopmental timelines. Thalamocortical pathways were reported to begin forming between 20 and 24 weeks gestation and to undergo refinement through 30–34 weeks [14], 15]. Subplate formation and synaptogenesis were described as critical preconditions for cortical signal integration, although not direct evidence of conscious experience [43], 44].

Fetal EEG and fMEG studies reported stimulus-evoked responses to auditory input, including changes in heart rate and localized brain activity following exposure to maternal voice or music [16–19]. These studies support the presence of basic sensory encoding and physiological reactivity during late gestation, particularly between 28 and 34 weeks. Observations of fetal habituation to repeated sound stimuli were also noted, suggesting primitive forms of neural plasticity [16], 21].

Table 2 summarizes the relevant structural and functional milestones across gestation, while Table 3 aligns these with sensory system development. Vision, touch, hearing, and proprioception all appear to reach functional thresholds at different points, with touch and hearing showing the earliest responsiveness.

Artificial intelligence in fetal imaging

Three studies in the core set applied AI or machine learning techniques to analyze fetal behavior using 4D ultrasound or MRI data [29–31], 45]. These tools were used to automate the classification of fetal facial movements, quantify gestural symmetry, and track motion sequences. One study employed convolutional neural networks to predict fetal behavioral states in growth-restricted pregnancies [30].

While these applications showed improved resolution and repeatability in gesture classification, none of the reviewed AI models underwent external clinical validation. Furthermore, all authors emphasized the need for interpretive caution, as semantic categories such as “smile” or “grimace” may not correspond to emotional states in fetuses [31], [32], [33]. The studies acknowledged limitations related to small sample sizes, training on postnatal datasets, and a lack of consensus on fetal behavioral coding.

Thus, while AI-enhanced tools may contribute to diagnostic imaging, they currently serve as adjuncts to, rather than replacements for, clinician interpretation.

Ethical reporting in included studies

Of the 74 included articles, 12 addressed ethical or philosophical dimensions of fetal imaging or behavior, typically in brief concluding sections. Topics included the appropriate interpretation of fetal responses in clinical settings [24], 25], the implications of pain perception during fetal surgery [23], 26], and concerns about anthropomorphic mislabeling of fetal movements [25].

These ethical commentaries were primarily utilitarian or precautionary in tone, focusing on minimizing harm in high-risk procedures and avoiding overstatement of behavioral significance. A smaller number of studies referenced cultural frameworks, including Islamic perspectives on fetal development [34], but these were secondary to clinical or imaging-focused discussion.

While Table 4 groups these perspectives for clarity, the ethical analysis drawn from the included studies remains limited in scope and depth. Broader ethical interpretation is addressed in the Discussion.

Fetal behavior and neurodevelopmental integration

This systematic review examined how observable fetal behaviors, as captured through 4D ultrasonography and related neuroimaging modalities, correspond to known stages of neurodevelopment. The findings indicate that behaviors such as yawning, blinking, limb movement, and facial expressions emerge in a temporally structured pattern from the second trimester onward [1], 2], 4–12], 41]. As shown in Table 1 and Table 3, these behaviors appear to increase in complexity and organization between 24 and 34 weeks of gestation, coinciding with significant milestones in thalamocortical connectivity and subplate remodeling [14], 15], 21], 44]. These observations suggest that fetal behavior is not random but may reflect progressive sensorimotor integration. Figure 2 presents a conceptual framework for understanding how structural neurodevelopment and environmental inputs may contribute to the fetus’s readiness for integrative processing – without implying conscious experience.

Figure 2:

Conceptual framework of neurodevelopmental readiness in the fetus. This diagram illustrates a theoretical model of fetal neurodevelopmental readiness, without inferring subjective consciousness. At its core is neural maturation, including cortical differentiation and thalamocortical integration. Surrounding this are dynamic influences that modulate fetal sensory and behavioral development. Sensory exposure begins early in utero and progressively supports integrative responsiveness. Maternal signals – such as hormonal, metabolic, and emotional cues – shape neurophysiological environments. Technological interfaces like ultrasound and AI analytics enhance observation but do not establish awareness. Cultural, ethical, and philosophical frameworks inform how developmental data are interpreted. The model highlights relational and systemic factors without claiming causal thresholds for consciousness. It is a conceptual synthesis to map conditions under which neurobehavioral readiness may emerge.

These observations suggest that fetal behavior is not random but may reflect progressive sensorimotor integration. However, while the correlation between neural maturation and behavioral regularity is well-supported (see Table 2, Figure 4), the current evidence does not support definitive conclusions about fetal awareness or subjective experience. Rather, these behaviors should be understood as signs of increasing neurological complexity, not volitional or conscious acts.

Avoiding anthropomorphic and semantic overreach

A key concern raised in the literature – and echoed by this review’s findings – is the risk of anthropomorphism in interpreting fetal behavior. Terms such as “smile,” “grimace,” or “yawn” are frequently used in the reviewed studies for morphological convenience but can inadvertently project emotional or intentional content where none may exist [3], 5], 6]. This issue is further complicated by the application of artificial intelligence (AI) tools in fetal behavior analysis. While machine learning algorithms have improved the consistency and granularity of behavior detection, these tools often rely on postnatally derived affective labels [28–33].

As noted in the Results and in Table 1, AI-based systems can identify facial movements or symmetry patterns, but the interpretive validity of these outputs remains limited. Without clinical validation and context-sensitive interpretive frameworks, such technologies risk what scholars have called “semantic inflation” – the over-interpretation of structural gestures as affective states [27], 30]. Therefore, both clinicians and researchers must exercise caution to avoid ascribing psychological meaning to morphologically coded behaviors without robust neurological or phenomenological justification.

Ethical and bioethical considerations

Although the reviewed studies were primarily clinical and observational in nature, 12 included explicit discussions of ethical implications. These ranged from fetal pain perception [23], 24], 26], 55] to maternal–fetal relational dynamics [53], 54], 56] and concerns over technological determinism [25]. Table 4 synthesizes these ethical approaches and situates them within broader philosophical traditions.

One underrepresented but crucial area is feminist bioethics, which critiques the ways fetal imaging technologies can marginalize maternal experience and reinforce a disembodied, autonomous view of the fetus. The increasing use of high-resolution imaging and AI overlays risks promoting a visually dominant narrative that separates the fetus from its relational and gestational context. Similarly, perspectives from disability ethics warn against using behavioral “norms” (e.g., expected movement patterns) to infer developmental worth, especially when used to guide decisions about viability or termination.

Cultural and religious traditions further complicate ethical translation. Islamic bioethics often centers fetal moral status around the concept of ensoulment at approximately 120 days (17 weeks), while Confucian and Buddhist frameworks emphasize relational emergence rather than fixed ontological status [34–39]. These perspectives caution against exporting neurobiological findings into universal ethical conclusions without considering sociocultural variability.

In light of these factors, fetal behavior should be interpreted within a broader ethical ecology that includes medical, relational, cultural, and epistemological dimensions – not just imaging data.

Clinical relevance and policy context

Although observational fetal behaviors may reflect developmental maturity, this review finds no evidence that such behaviors constitute indicators of conscious awareness. Therefore, clinical interventions – such as fetal surgery, anesthesia protocols, or counseling on viability – should be guided by a precautionary ethical approach rather than assumptions of subjectivity.

Importantly, this review finds no empirical basis to support the reinterpretation of fetal behavior as a foundation for changes in abortion law or fetal personhood arguments. As Figure 3 illustrates, fetal imaging often enters sociopolitical debates through emotionally compelling visuals that are detached from scientific interpretation. Ethical communication must distinguish between what can be seen and what is known, especially in contexts where reproductive rights and maternal autonomy are at stake.

Figure 3:

Key interdisciplinary domains informing fetal neurodevelopmental awareness. This figure illustrates the key interdisciplinary domains that inform our understanding of fetal neurodevelopmental awareness. In response to reviewer feedback, speculative elements have been removed, focusing instead on empirically supported and clinically relevant fields. Core contributors include fetal anatomy, neuroimaging, electrophysiology, ethics, law, and maternal-fetal medicine. The central balance symbol represents the integration of scientific evidence with ethical and legal interpretation. The circular arrangement avoids implying linear causality, emphasizing the complexity of fetal sensory and neural development. This model does not claim to define fetal consciousness but highlights the necessary conditions for integrative processing. It serves as a conceptual guide grounded in current evidence and cross-disciplinary dialogue.

Figure 4:

Fetal sensory and brain development: Milestones of neural integration and functional readiness. This timeline illustrates the progression of fetal brain and sensory development across key gestational stages. Each milestone represents structural or functional neural changes that contribute to integrative capacity, not the onset of subjective awareness. Early events such as neural tube formation and cranial nerve differentiation lay the anatomical groundwork for sensory reception. Mid-gestational milestones like thalamocortical connectivity and cortical layering enable higher-order sensory integration. Cortical synaptogenesis and EEG differentiation suggest readiness for environmental responsiveness but do not imply the emergence of consciousness. The model emphasizes functional preparedness and neural maturity without inferring internal experiences. This framework supports cautious interpretation of fetal behavior in light of ongoing neurodevelopment.

Limitations of the evidence and this review

Several limitations constrain both the current literature and this systematic review. First, many included studies relied on observational data with small sample sizes and lacked longitudinal follow-up. Second, there was a high degree of heterogeneity in behavioral coding systems, gestational age windows, and interpretive frameworks. Third, while AI-assisted analyses represent a novel advance, they remain under-validated and often use affective terms without cross-disciplinary consensus.

In addition, although this review attempted to integrate bioethical analysis, most included studies addressed ethical concerns only in passing. This limits the depth of ethical synthesis possible. Finally, the literature remains geographically skewed, with limited representation from low-resource settings or non-Western contexts.

Future research directions

Future studies should aim for longitudinal integration of behavioral and neuroimaging data from mid-gestation through early infancy. Standardized fetal behavior coding systems, combined with advanced imaging (e.g., fetal fMRI) and physiological metrics, may offer more precise correlations between structure and function. AI models must be co-developed with ethical oversight and undergo rigorous validation before clinical adoption.

Moreover, research is needed on the public communication of fetal imaging – especially how visual representations influence perceptions of moral status, parental decision-making, and legal interpretation. Interdisciplinary research that includes obstetrics, neuroscience, ethics, communication studies, and law will be critical in advancing responsible discourse. This review finds that fetal behavior becomes increasingly structured and neurodevelopmentally aligned during the second and third trimesters. These behaviors correlate with key neuroanatomical milestones but do not constitute evidence of conscious experience. Interpretive caution is necessary, especially in the context of advanced imaging and AI-enhanced diagnostics.

Ethically, the growing visibility of the fetus presents both opportunities and challenges. Clinicians, researchers, and policymakers must remain sensitive to the limits of current evidence, avoid premature conclusions, and engage in pluralistic, context-aware ethical reflection. Scientific clarity, clinical responsibility, and ethical humility must guide ongoing inquiry into the complex intersection of fetal development, behavior, and meaning.

Strengths

This review offers a comprehensive and interdisciplinary synthesis of research on fetal behavior as observed through four-dimensional (4D) ultrasonography, functional neuroimaging, and emerging artificial intelligence (AI) tools. By systematically mapping behavioral patterns across gestation onto neurodevelopmental milestones (Table 2, Table 3), the study contributes to a growing understanding of fetal sensorimotor organization and neurological maturation.

Methodologically, the review adhered to PRISMA 2020 guidelines, applying a transparent and reproducible selection process (Figure 1) across four major databases and a 25-year span (2000–2025). Its scope included foundational observational studies as well as recent innovations in AI-assisted imaging, deep learning–based segmentation, and fetal fMRI [21], 22] 31–33]. A notable contribution lies in the integration of diverse disciplines – neuroscience, obstetrics, developmental psychology, and bioethics – into a cohesive framework for analyzing fetal behavior within a structured, developmental, and ethically cautious model.

This review also underscores the interpretive potential and limitations of AI-based movement classifiers in prenatal diagnostics. Rather than endorsing speculative claims, it critically examines how technological visibility must be balanced against epistemic humility. The findings reinforce the importance of contextual interpretation, particularly when visual data are used in ethically sensitive domains such as fetal pain, viability, or clinical counseling.

Limitations

Several limitations of the current evidence base and the present review warrant acknowledgment. First, all inferences regarding fetal awareness or perceptual capacity remain indirect. While behavioral regularity and thalamocortical maturation suggest increasing neurodevelopmental organization, they do not constitute proof of conscious experience [14], 15], 21], 43]. Observable behaviors – such as facial movements or hand gestures – should be interpreted cautiously, using morphologically descriptive terms rather than affective or intentional labels [3], 6].

Second, 4D ultrasonography is constrained by technical variability, including fetal positioning, maternal body composition, and operator expertise. AI-enhanced interpretation, while promising, remains subject to misclassification – particularly when algorithms are trained on postnatal or adult datasets not validated for fetal contexts [29], [30], [31], [32], [33, 57]. These challenges raise concerns about anthropomorphic misapplication and normative assumptions embedded in behavioral modeling.

Third, the ethical interpretation of fetal imaging is highly context-dependent. As shown in Figure 3 and discussed in the literature, fetal imagery may be used rhetorically in political or cultural debates that diverge from scientific consensus [25]. This underscores the need for restraint in public-facing communication and careful distinction between visual representation and empirical evidence.

Future directions

Future research should prioritize longitudinal designs that link prenatal behavior to postnatal neurocognitive and emotional outcomes. Establishing behavioral continuity across developmental stages will strengthen the clinical relevance of fetal observations and provide a firmer foundation for risk assessment and intervention.

There is also a need for improved validation of AI-driven analytic tools. Algorithm development should be interdisciplinary and explicitly designed for prenatal application, incorporating ethical review throughout the design pipeline. This includes safeguards against semantic overreach and normative bias in the labeling of fetal behavior [21], 31], 32].

From a clinical ethics perspective, a precautionary approach remains warranted – especially in high-risk pregnancies or surgical contexts. Continued use of fetal anesthesia based on nociceptive thresholds [23], 26] reflects such a stance and should remain guided by evidence, not public sentiment.

Bioethical analysis would benefit from deeper engagement with feminist, disability, and relational ethics, which interrogate how fetal behavior is framed within maternal, cultural, and systemic contexts. Legal and policy translation must also recognize ethical pluralism, especially given differing cultural understandings of personhood, agency, and gestational thresholds [34–39].

Finally, as fetal imaging technologies become more sophisticated and widely disseminated, the need for ethically responsible communication intensifies. Future scholarship should explore how fetal visibility shapes public attitudes, clinical practice, and reproductive policy. Transparent, interdisciplinary discourse – grounded in scientific caution and cultural respect – will be essential to navigating the increasingly complex terrain of fetal development and its ethical implications.